WEBVTT Kind: captions Language: en 00:00:00.160 --> 00:00:02.389 align:start position:0% This<00:00:00.400> is<00:00:00.560> one<00:00:00.719> of<00:00:00.800> the<00:00:00.960> most<00:00:01.199> underrated<00:00:01.920> AI 00:00:02.389 --> 00:00:02.399 align:start position:0% This is one of the most underrated AI 00:00:02.399 --> 00:00:04.870 align:start position:0% This is one of the most underrated AI breakthroughs<00:00:03.200> so<00:00:03.520> far.<00:00:04.080> These<00:00:04.400> researchers 00:00:04.870 --> 00:00:04.880 align:start position:0% breakthroughs so far. These researchers 00:00:04.880 --> 00:00:06.869 align:start position:0% breakthroughs so far. These researchers made<00:00:05.120> an<00:00:05.279> AI<00:00:05.600> that<00:00:05.839> can<00:00:06.080> understand<00:00:06.400> the<00:00:06.640> code 00:00:06.869 --> 00:00:06.879 align:start position:0% made an AI that can understand the code 00:00:06.879 --> 00:00:09.830 align:start position:0% made an AI that can understand the code of<00:00:07.040> life,<00:00:07.520> DNA.<00:00:08.320> In<00:00:08.559> fact,<00:00:08.960> they<00:00:09.200> can<00:00:09.360> use<00:00:09.519> this 00:00:09.830 --> 00:00:09.840 align:start position:0% of life, DNA. In fact, they can use this 00:00:09.840 --> 00:00:12.549 align:start position:0% of life, DNA. In fact, they can use this to<00:00:10.160> detect<00:00:10.559> cancer<00:00:11.040> or<00:00:11.360> other<00:00:11.679> diseases,<00:00:12.320> or 00:00:12.549 --> 00:00:12.559 align:start position:0% to detect cancer or other diseases, or 00:00:12.559 --> 00:00:14.230 align:start position:0% to detect cancer or other diseases, or it<00:00:12.719> can<00:00:12.880> even<00:00:13.120> predict<00:00:13.440> the<00:00:13.679> effects<00:00:14.000> of 00:00:14.230 --> 00:00:14.240 align:start position:0% it can even predict the effects of 00:00:14.240 --> 00:00:16.070 align:start position:0% it can even predict the effects of genetic<00:00:14.639> mutations,<00:00:15.360> and<00:00:15.519> it<00:00:15.759> can<00:00:15.839> even 00:00:16.070 --> 00:00:16.080 align:start position:0% genetic mutations, and it can even 00:00:16.080 --> 00:00:18.710 align:start position:0% genetic mutations, and it can even generate<00:00:16.560> the<00:00:16.880> complete<00:00:17.359> DNA<00:00:17.920> of<00:00:18.160> a<00:00:18.400> living 00:00:18.710 --> 00:00:18.720 align:start position:0% generate the complete DNA of a living 00:00:18.720 --> 00:00:20.870 align:start position:0% generate the complete DNA of a living thing.<00:00:19.279> The<00:00:19.520> implications<00:00:20.080> of<00:00:20.240> this<00:00:20.560> are 00:00:20.870 --> 00:00:20.880 align:start position:0% thing. The implications of this are 00:00:20.880 --> 00:00:22.950 align:start position:0% thing. The implications of this are wild.<00:00:21.600> This<00:00:21.920> technology<00:00:22.560> could<00:00:22.800> help 00:00:22.950 --> 00:00:22.960 align:start position:0% wild. This technology could help 00:00:22.960 --> 00:00:25.269 align:start position:0% wild. This technology could help scientists<00:00:23.519> engineer<00:00:24.080> better<00:00:24.480> crops<00:00:24.960> and 00:00:25.269 --> 00:00:25.279 align:start position:0% scientists engineer better crops and 00:00:25.279 --> 00:00:27.750 align:start position:0% scientists engineer better crops and unlock<00:00:25.680> new<00:00:26.000> solutions<00:00:26.480> for<00:00:26.800> energy<00:00:27.199> and<00:00:27.519> food 00:00:27.750 --> 00:00:27.760 align:start position:0% unlock new solutions for energy and food 00:00:27.760 --> 00:00:29.750 align:start position:0% unlock new solutions for energy and food security.<00:00:28.400> And<00:00:28.560> this<00:00:28.800> is<00:00:28.960> also<00:00:29.199> great<00:00:29.439> for 00:00:29.750 --> 00:00:29.760 align:start position:0% security. And this is also great for 00:00:29.760 --> 00:00:32.069 align:start position:0% security. And this is also great for personalized<00:00:30.560> medicine.<00:00:31.199> We<00:00:31.439> could<00:00:31.599> even<00:00:31.840> use 00:00:32.069 --> 00:00:32.079 align:start position:0% personalized medicine. We could even use 00:00:32.079 --> 00:00:34.790 align:start position:0% personalized medicine. We could even use this<00:00:32.320> to<00:00:32.559> design<00:00:33.040> entirely<00:00:33.600> new<00:00:33.920> species<00:00:34.480> or 00:00:34.790 --> 00:00:34.800 align:start position:0% this to design entirely new species or 00:00:34.800 --> 00:00:36.950 align:start position:0% this to design entirely new species or even<00:00:35.040> make<00:00:35.360> ourselves<00:00:35.920> better.<00:00:36.559> In<00:00:36.800> this 00:00:36.950 --> 00:00:36.960 align:start position:0% even make ourselves better. In this 00:00:36.960 --> 00:00:39.030 align:start position:0% even make ourselves better. In this video,<00:00:37.280> I'm<00:00:37.440> going<00:00:37.520> to<00:00:37.680> break<00:00:37.920> down<00:00:38.399> exactly 00:00:39.030 --> 00:00:39.040 align:start position:0% video, I'm going to break down exactly 00:00:39.040 --> 00:00:41.030 align:start position:0% video, I'm going to break down exactly how<00:00:39.360> they<00:00:39.600> did<00:00:39.760> this,<00:00:40.079> how<00:00:40.320> it<00:00:40.480> works,<00:00:40.800> and 00:00:41.030 --> 00:00:41.040 align:start position:0% how they did this, how it works, and 00:00:41.040 --> 00:00:43.430 align:start position:0% how they did this, how it works, and some<00:00:41.360> insane<00:00:41.920> findings<00:00:42.399> from<00:00:42.719> this<00:00:42.960> project. 00:00:43.430 --> 00:00:43.440 align:start position:0% some insane findings from this project. 00:00:43.440 --> 00:00:45.910 align:start position:0% some insane findings from this project. Now,<00:00:43.680> this<00:00:43.840> is<00:00:43.920> a<00:00:44.160> very<00:00:44.399> technical<00:00:44.960> paper,<00:00:45.600> but 00:00:45.910 --> 00:00:45.920 align:start position:0% Now, this is a very technical paper, but 00:00:45.920 --> 00:00:47.590 align:start position:0% Now, this is a very technical paper, but as<00:00:46.239> always,<00:00:46.640> I'm<00:00:46.800> going<00:00:46.879> to<00:00:46.960> break<00:00:47.200> this<00:00:47.360> down 00:00:47.590 --> 00:00:47.600 align:start position:0% as always, I'm going to break this down 00:00:47.600 --> 00:00:49.590 align:start position:0% as always, I'm going to break this down into<00:00:47.920> simple<00:00:48.239> terms<00:00:48.480> so<00:00:48.719> that<00:00:48.879> anyone<00:00:49.280> can 00:00:49.590 --> 00:00:49.600 align:start position:0% into simple terms so that anyone can 00:00:49.600 --> 00:00:52.229 align:start position:0% into simple terms so that anyone can understand.<00:00:50.320> Let's<00:00:50.640> jump<00:00:50.960> right<00:00:51.200> in.<00:00:51.680> Now,<00:00:51.920> to 00:00:52.229 --> 00:00:52.239 align:start position:0% understand. Let's jump right in. Now, to 00:00:52.239 --> 00:00:54.470 align:start position:0% understand. Let's jump right in. Now, to understand<00:00:52.719> what<00:00:53.039> they<00:00:53.280> did<00:00:53.440> in<00:00:53.680> this<00:00:53.920> paper, 00:00:54.470 --> 00:00:54.480 align:start position:0% understand what they did in this paper, 00:00:54.480 --> 00:00:56.709 align:start position:0% understand what they did in this paper, let's<00:00:54.800> start<00:00:55.039> with<00:00:55.360> the<00:00:55.600> basics.<00:00:56.239> You<00:00:56.480> see, 00:00:56.709 --> 00:00:56.719 align:start position:0% let's start with the basics. You see, 00:00:56.719 --> 00:00:58.709 align:start position:0% let's start with the basics. You see, current<00:00:57.039> AI<00:00:57.440> models<00:00:57.680> that<00:00:57.840> we<00:00:58.000> use<00:00:58.160> today<00:00:58.480> like 00:00:58.709 --> 00:00:58.719 align:start position:0% current AI models that we use today like 00:00:58.719 --> 00:01:01.750 align:start position:0% current AI models that we use today like Chat<00:00:59.039> GPT<00:00:59.760> and<00:01:00.000> Gemini,<00:01:00.800> these<00:01:01.120> are<00:01:01.359> large 00:01:01.750 --> 00:01:01.760 align:start position:0% Chat GPT and Gemini, these are large 00:01:01.760 --> 00:01:03.670 align:start position:0% Chat GPT and Gemini, these are large language<00:01:02.160> models.<00:01:02.719> They're<00:01:03.039> designed<00:01:03.280> to<00:01:03.440> do 00:01:03.670 --> 00:01:03.680 align:start position:0% language models. They're designed to do 00:01:03.680 --> 00:01:05.189 align:start position:0% language models. They're designed to do really<00:01:03.920> good<00:01:04.080> at<00:01:04.320> understanding<00:01:04.879> and 00:01:05.189 --> 00:01:05.199 align:start position:0% really good at understanding and 00:01:05.199 --> 00:01:07.190 align:start position:0% really good at understanding and outputting<00:01:05.760> natural<00:01:06.159> language.<00:01:06.799> That's<00:01:06.960> why 00:01:07.190 --> 00:01:07.200 align:start position:0% outputting natural language. That's why 00:01:07.200 --> 00:01:09.750 align:start position:0% outputting natural language. That's why it<00:01:07.439> can<00:01:07.600> generate<00:01:08.000> a<00:01:08.320> beautiful<00:01:08.799> poem<00:01:09.280> or<00:01:09.520> a 00:01:09.750 --> 00:01:09.760 align:start position:0% it can generate a beautiful poem or a 00:01:09.760 --> 00:01:13.429 align:start position:0% it can generate a beautiful poem or a flawless<00:01:10.400> essay<00:01:11.200> or<00:01:11.600> even<00:01:11.920> do<00:01:12.159> a<00:01:12.400> full<00:01:12.880> medical 00:01:13.429 --> 00:01:13.439 align:start position:0% flawless essay or even do a full medical 00:01:13.439 --> 00:01:15.910 align:start position:0% flawless essay or even do a full medical research<00:01:13.920> report<00:01:14.320> for<00:01:14.560> you.<00:01:15.119> But<00:01:15.360> here's<00:01:15.680> the 00:01:15.910 --> 00:01:15.920 align:start position:0% research report for you. But here's the 00:01:15.920 --> 00:01:17.749 align:start position:0% research report for you. But here's the interesting<00:01:16.400> question.<00:01:16.960> Instead<00:01:17.280> of<00:01:17.439> the 00:01:17.749 --> 00:01:17.759 align:start position:0% interesting question. Instead of the 00:01:17.759 --> 00:01:19.990 align:start position:0% interesting question. Instead of the natural<00:01:18.080> language<00:01:18.560> that<00:01:18.960> you<00:01:19.119> and<00:01:19.280> I<00:01:19.520> use<00:01:19.759> to 00:01:19.990 --> 00:01:20.000 align:start position:0% natural language that you and I use to 00:01:20.000 --> 00:01:21.590 align:start position:0% natural language that you and I use to communicate,<00:01:20.720> you<00:01:20.880> know<00:01:20.960> what<00:01:21.119> else<00:01:21.360> is 00:01:21.590 --> 00:01:21.600 align:start position:0% communicate, you know what else is 00:01:21.600 --> 00:01:24.630 align:start position:0% communicate, you know what else is similar?<00:01:22.240> DNA<00:01:23.040> or<00:01:23.280> the<00:01:23.520> language<00:01:23.920> of<00:01:24.159> life. 00:01:24.630 --> 00:01:24.640 align:start position:0% similar? DNA or the language of life. 00:01:24.640 --> 00:01:26.469 align:start position:0% similar? DNA or the language of life. What<00:01:24.880> if<00:01:24.960> we<00:01:25.119> took<00:01:25.360> the<00:01:25.600> same<00:01:25.840> logic<00:01:26.240> from 00:01:26.469 --> 00:01:26.479 align:start position:0% What if we took the same logic from 00:01:26.479 --> 00:01:28.870 align:start position:0% What if we took the same logic from these<00:01:26.799> large<00:01:27.200> language<00:01:27.600> models<00:01:28.000> and<00:01:28.320> apply<00:01:28.640> it 00:01:28.870 --> 00:01:28.880 align:start position:0% these large language models and apply it 00:01:28.880 --> 00:01:31.510 align:start position:0% these large language models and apply it to<00:01:29.200> train<00:01:29.600> an<00:01:29.840> AI<00:01:30.400> that<00:01:30.720> can<00:01:30.880> understand<00:01:31.280> and 00:01:31.510 --> 00:01:31.520 align:start position:0% to train an AI that can understand and 00:01:31.520 --> 00:01:34.310 align:start position:0% to train an AI that can understand and generate<00:01:32.000> DNA?<00:01:32.799> Well,<00:01:33.040> that's<00:01:33.439> exactly<00:01:34.000> what 00:01:34.310 --> 00:01:34.320 align:start position:0% generate DNA? Well, that's exactly what 00:01:34.320 --> 00:01:36.710 align:start position:0% generate DNA? Well, that's exactly what these<00:01:34.799> researchers<00:01:35.360> did.<00:01:35.840> So,<00:01:36.000> the<00:01:36.159> paper<00:01:36.479> is 00:01:36.710 --> 00:01:36.720 align:start position:0% these researchers did. So, the paper is 00:01:36.720 --> 00:01:38.789 align:start position:0% these researchers did. So, the paper is called<00:01:37.200> genome<00:01:37.600> modeling<00:01:37.920> and<00:01:38.159> design<00:01:38.400> across 00:01:38.789 --> 00:01:38.799 align:start position:0% called genome modeling and design across 00:01:38.799 --> 00:01:41.590 align:start position:0% called genome modeling and design across all<00:01:39.040> domains<00:01:39.439> of<00:01:39.600> life<00:01:39.920> with<00:01:40.320> EVO<00:01:40.880> 2.<00:01:41.200> And<00:01:41.360> this 00:01:41.590 --> 00:01:41.600 align:start position:0% all domains of life with EVO 2. And this 00:01:41.600 --> 00:01:43.990 align:start position:0% all domains of life with EVO 2. And this was<00:01:41.759> published<00:01:42.240> recently<00:01:42.880> in<00:01:43.360> nature<00:01:43.840> which 00:01:43.990 --> 00:01:44.000 align:start position:0% was published recently in nature which 00:01:44.000 --> 00:01:45.749 align:start position:0% was published recently in nature which is<00:01:44.159> like<00:01:44.320> the<00:01:44.479> most<00:01:44.720> prestigious<00:01:45.280> scientific 00:01:45.749 --> 00:01:45.759 align:start position:0% is like the most prestigious scientific 00:01:45.759 --> 00:01:47.590 align:start position:0% is like the most prestigious scientific journal<00:01:46.159> in<00:01:46.320> the<00:01:46.479> world.<00:01:46.960> Now,<00:01:47.200> before<00:01:47.439> we 00:01:47.590 --> 00:01:47.600 align:start position:0% journal in the world. Now, before we 00:01:47.600 --> 00:01:50.469 align:start position:0% journal in the world. Now, before we dive<00:01:47.920> into<00:01:48.640> why<00:01:48.960> this<00:01:49.200> is<00:01:49.360> so<00:01:49.600> important,<00:01:50.240> we 00:01:50.469 --> 00:01:50.479 align:start position:0% dive into why this is so important, we 00:01:50.479 --> 00:01:52.950 align:start position:0% dive into why this is so important, we first<00:01:50.720> need<00:01:50.880> to<00:01:51.040> understand<00:01:51.520> what<00:01:51.920> DNA<00:01:52.479> is. 00:01:52.950 --> 00:01:52.960 align:start position:0% first need to understand what DNA is. 00:01:52.960 --> 00:01:55.190 align:start position:0% first need to understand what DNA is. You<00:01:53.200> see,<00:01:53.439> DNA<00:01:54.000> is<00:01:54.159> like<00:01:54.399> the<00:01:54.640> instruction 00:01:55.190 --> 00:01:55.200 align:start position:0% You see, DNA is like the instruction 00:01:55.200 --> 00:01:57.670 align:start position:0% You see, DNA is like the instruction manual<00:01:55.680> for<00:01:56.000> life.<00:01:56.479> Inside<00:01:56.960> almost<00:01:57.360> every 00:01:57.670 --> 00:01:57.680 align:start position:0% manual for life. Inside almost every 00:01:57.680 --> 00:02:00.630 align:start position:0% manual for life. Inside almost every cell<00:01:58.000> in<00:01:58.320> every<00:01:58.719> living<00:01:59.040> thing,<00:01:59.600> there's<00:02:00.000> DNA. 00:02:00.630 --> 00:02:00.640 align:start position:0% cell in every living thing, there's DNA. 00:02:00.640 --> 00:02:03.109 align:start position:0% cell in every living thing, there's DNA. And<00:02:00.799> it<00:02:00.960> stores<00:02:01.520> information<00:02:02.399> needed<00:02:02.799> to 00:02:03.109 --> 00:02:03.119 align:start position:0% And it stores information needed to 00:02:03.119 --> 00:02:06.069 align:start position:0% And it stores information needed to build<00:02:03.360> and<00:02:03.680> run<00:02:03.920> the<00:02:04.159> organism.<00:02:04.799> Now,<00:02:05.119> DNA<00:02:05.680> is 00:02:06.069 --> 00:02:06.079 align:start position:0% build and run the organism. Now, DNA is 00:02:06.079 --> 00:02:09.990 align:start position:0% build and run the organism. Now, DNA is made<00:02:06.399> of<00:02:06.960> only<00:02:07.360> four<00:02:07.759> letters<00:02:08.479> G,<00:02:08.800> C,<00:02:09.119> A<00:02:09.360> or<00:02:09.599> T. 00:02:09.990 --> 00:02:10.000 align:start position:0% made of only four letters G, C, A or T. 00:02:10.000 --> 00:02:12.869 align:start position:0% made of only four letters G, C, A or T. And<00:02:10.160> these<00:02:10.479> letters<00:02:11.039> bond<00:02:11.440> in<00:02:11.680> pairs.<00:02:12.319> So,<00:02:12.640> G 00:02:12.869 --> 00:02:12.879 align:start position:0% And these letters bond in pairs. So, G 00:02:12.879 --> 00:02:15.750 align:start position:0% And these letters bond in pairs. So, G bonds<00:02:13.200> with<00:02:13.440> C<00:02:13.760> and<00:02:14.000> A<00:02:14.239> bonds<00:02:14.560> with<00:02:14.800> T<00:02:15.120> and<00:02:15.440> vice 00:02:15.750 --> 00:02:15.760 align:start position:0% bonds with C and A bonds with T and vice 00:02:15.760 --> 00:02:18.150 align:start position:0% bonds with C and A bonds with T and vice versa.<00:02:16.239> So<00:02:16.480> from<00:02:16.720> this<00:02:17.040> bonding<00:02:17.520> it<00:02:17.760> creates 00:02:18.150 --> 00:02:18.160 align:start position:0% versa. So from this bonding it creates 00:02:18.160 --> 00:02:20.790 align:start position:0% versa. So from this bonding it creates two<00:02:18.480> strands<00:02:18.959> like<00:02:19.200> this<00:02:19.840> which<00:02:20.400> wind 00:02:20.790 --> 00:02:20.800 align:start position:0% two strands like this which wind 00:02:20.800 --> 00:02:23.190 align:start position:0% two strands like this which wind together<00:02:21.120> in<00:02:21.360> a<00:02:21.520> spiral<00:02:22.000> or<00:02:22.319> double<00:02:22.720> helix 00:02:23.190 --> 00:02:23.200 align:start position:0% together in a spiral or double helix 00:02:23.200 --> 00:02:25.750 align:start position:0% together in a spiral or double helix shape<00:02:23.599> like<00:02:23.920> this.<00:02:24.400> So<00:02:24.640> this<00:02:24.800> is<00:02:24.959> the<00:02:25.200> classic 00:02:25.750 --> 00:02:25.760 align:start position:0% shape like this. So this is the classic 00:02:25.760 --> 00:02:27.670 align:start position:0% shape like this. So this is the classic DNA<00:02:26.319> shape<00:02:26.640> that<00:02:26.879> you<00:02:27.040> might<00:02:27.200> be<00:02:27.360> familiar 00:02:27.670 --> 00:02:27.680 align:start position:0% DNA shape that you might be familiar 00:02:27.680 --> 00:02:30.150 align:start position:0% DNA shape that you might be familiar with.<00:02:28.080> Now<00:02:28.319> the<00:02:28.560> sequence<00:02:29.120> of<00:02:29.440> these<00:02:29.760> letters 00:02:30.150 --> 00:02:30.160 align:start position:0% with. Now the sequence of these letters 00:02:30.160 --> 00:02:32.229 align:start position:0% with. Now the sequence of these letters is<00:02:30.400> what<00:02:30.640> contains<00:02:31.120> the<00:02:31.360> instructions<00:02:32.000> of 00:02:32.229 --> 00:02:32.239 align:start position:0% is what contains the instructions of 00:02:32.239 --> 00:02:34.869 align:start position:0% is what contains the instructions of life.<00:02:32.800> It<00:02:32.959> encodes<00:02:33.599> thousands<00:02:34.080> of<00:02:34.319> traits<00:02:34.640> and 00:02:34.869 --> 00:02:34.879 align:start position:0% life. It encodes thousands of traits and 00:02:34.879 --> 00:02:37.270 align:start position:0% life. It encodes thousands of traits and biological<00:02:35.519> processes.<00:02:36.319> For<00:02:36.480> example,<00:02:36.959> for 00:02:37.270 --> 00:02:37.280 align:start position:0% biological processes. For example, for 00:02:37.280 --> 00:02:39.910 align:start position:0% biological processes. For example, for humans<00:02:37.920> this<00:02:38.160> includes<00:02:38.720> eye<00:02:38.959> color,<00:02:39.440> height, 00:02:39.910 --> 00:02:39.920 align:start position:0% humans this includes eye color, height, 00:02:39.920 --> 00:02:42.390 align:start position:0% humans this includes eye color, height, risk<00:02:40.239> of<00:02:40.480> disease,<00:02:41.040> our<00:02:41.280> immune<00:02:41.680> system,<00:02:42.160> how 00:02:42.390 --> 00:02:42.400 align:start position:0% risk of disease, our immune system, how 00:02:42.400 --> 00:02:44.630 align:start position:0% risk of disease, our immune system, how we<00:02:42.640> metabolize,<00:02:43.519> and<00:02:43.760> more.<00:02:44.160> It's<00:02:44.400> the 00:02:44.630 --> 00:02:44.640 align:start position:0% we metabolize, and more. It's the 00:02:44.640 --> 00:02:46.550 align:start position:0% we metabolize, and more. It's the blueprint<00:02:45.040> for<00:02:45.280> how<00:02:45.519> living<00:02:45.840> things<00:02:46.160> function 00:02:46.550 --> 00:02:46.560 align:start position:0% blueprint for how living things function 00:02:46.560 --> 00:02:48.790 align:start position:0% blueprint for how living things function and<00:02:46.879> stay<00:02:47.120> alive.<00:02:47.840> All<00:02:47.840> right.<00:02:48.160> Now,<00:02:48.319> back<00:02:48.560> to 00:02:48.790 --> 00:02:48.800 align:start position:0% and stay alive. All right. Now, back to 00:02:48.800 --> 00:02:52.309 align:start position:0% and stay alive. All right. Now, back to this<00:02:49.040> paper<00:02:49.599> here.<00:02:49.920> They<00:02:50.239> created<00:02:50.640> this<00:02:51.040> EVO2 00:02:52.309 --> 00:02:52.319 align:start position:0% this paper here. They created this EVO2 00:02:52.319 --> 00:02:55.509 align:start position:0% this paper here. They created this EVO2 biological<00:02:53.280> foundation<00:02:53.840> model<00:02:54.400> trained<00:02:54.800> on<00:02:55.120> 9 00:02:55.509 --> 00:02:55.519 align:start position:0% biological foundation model trained on 9 00:02:55.519 --> 00:02:58.390 align:start position:0% biological foundation model trained on 9 trillion<00:02:56.160> DNA<00:02:56.720> base<00:02:57.040> pairs.<00:02:57.840> So,<00:02:58.000> this<00:02:58.160> is 00:02:58.390 --> 00:02:58.400 align:start position:0% trillion DNA base pairs. So, this is 00:02:58.400 --> 00:03:01.110 align:start position:0% trillion DNA base pairs. So, this is actually<00:02:58.879> very<00:02:59.280> similar<00:02:59.760> to<00:03:00.400> large<00:03:00.720> language 00:03:01.110 --> 00:03:01.120 align:start position:0% actually very similar to large language 00:03:01.120 --> 00:03:03.030 align:start position:0% actually very similar to large language models<00:03:01.360> like<00:03:01.599> Chat<00:03:01.920> GPT.<00:03:02.480> But<00:03:02.640> instead<00:03:02.879> of 00:03:03.030 --> 00:03:03.040 align:start position:0% models like Chat GPT. But instead of 00:03:03.040 --> 00:03:04.710 align:start position:0% models like Chat GPT. But instead of being<00:03:03.280> trained<00:03:03.519> on<00:03:03.760> all<00:03:03.920> the<00:03:04.080> data<00:03:04.319> from<00:03:04.560> the 00:03:04.710 --> 00:03:04.720 align:start position:0% being trained on all the data from the 00:03:04.720 --> 00:03:08.149 align:start position:0% being trained on all the data from the internet,<00:03:05.280> this<00:03:05.599> EVO<00:03:06.159> 2<00:03:06.560> was<00:03:06.879> trained<00:03:07.200> on<00:03:07.760> 9 00:03:08.149 --> 00:03:08.159 align:start position:0% internet, this EVO 2 was trained on 9 00:03:08.159 --> 00:03:11.190 align:start position:0% internet, this EVO 2 was trained on 9 trillion<00:03:08.720> DNA<00:03:09.360> base<00:03:09.599> pairs.<00:03:10.480> Specifically, 00:03:11.190 --> 00:03:11.200 align:start position:0% trillion DNA base pairs. Specifically, 00:03:11.200 --> 00:03:13.670 align:start position:0% trillion DNA base pairs. Specifically, they<00:03:11.440> trained<00:03:11.760> it<00:03:12.000> on<00:03:12.319> a<00:03:12.640> data<00:03:12.879> set<00:03:13.200> called 00:03:13.670 --> 00:03:13.680 align:start position:0% they trained it on a data set called 00:03:13.680 --> 00:03:16.309 align:start position:0% they trained it on a data set called Open<00:03:14.159> Genome<00:03:14.720> 2.<00:03:15.120> And<00:03:15.360> this<00:03:15.519> data<00:03:15.840> set 00:03:16.309 --> 00:03:16.319 align:start position:0% Open Genome 2. And this data set 00:03:16.319 --> 00:03:18.390 align:start position:0% Open Genome 2. And this data set essentially<00:03:16.800> takes<00:03:17.200> millions<00:03:17.680> of<00:03:17.920> diverse 00:03:18.390 --> 00:03:18.400 align:start position:0% essentially takes millions of diverse 00:03:18.400 --> 00:03:20.550 align:start position:0% essentially takes millions of diverse living<00:03:18.800> things<00:03:19.280> from<00:03:19.519> across<00:03:19.840> the<00:03:20.080> entire 00:03:20.550 --> 00:03:20.560 align:start position:0% living things from across the entire 00:03:20.560 --> 00:03:23.190 align:start position:0% living things from across the entire spectrum<00:03:20.959> of<00:03:21.200> life<00:03:21.599> and<00:03:21.840> puts<00:03:22.159> their<00:03:22.319> DNA<00:03:22.800> into 00:03:23.190 --> 00:03:23.200 align:start position:0% spectrum of life and puts their DNA into 00:03:23.200 --> 00:03:25.350 align:start position:0% spectrum of life and puts their DNA into a<00:03:23.440> massive<00:03:23.920> digital<00:03:24.400> library.<00:03:24.959> So,<00:03:25.120> as<00:03:25.280> you 00:03:25.350 --> 00:03:25.360 align:start position:0% a massive digital library. So, as you 00:03:25.360 --> 00:03:27.110 align:start position:0% a massive digital library. So, as you can<00:03:25.440> see,<00:03:25.680> this<00:03:25.840> includes<00:03:26.319> everything<00:03:26.800> from 00:03:27.110 --> 00:03:27.120 align:start position:0% can see, this includes everything from 00:03:27.120 --> 00:03:30.229 align:start position:0% can see, this includes everything from bacteria,<00:03:28.000> plants,<00:03:28.480> fungi,<00:03:29.120> and<00:03:29.519> animals, 00:03:30.229 --> 00:03:30.239 align:start position:0% bacteria, plants, fungi, and animals, 00:03:30.239 --> 00:03:32.309 align:start position:0% bacteria, plants, fungi, and animals, including<00:03:30.640> of<00:03:30.879> course<00:03:31.120> humans.<00:03:31.840> The<00:03:32.000> total 00:03:32.309 --> 00:03:32.319 align:start position:0% including of course humans. The total 00:03:32.319 --> 00:03:35.589 align:start position:0% including of course humans. The total size<00:03:32.560> of<00:03:32.720> this<00:03:32.959> is<00:03:33.200> like<00:03:33.680> 9<00:03:34.080> trillion<00:03:34.799> DNA<00:03:35.360> base 00:03:35.589 --> 00:03:35.599 align:start position:0% size of this is like 9 trillion DNA base 00:03:35.599 --> 00:03:37.830 align:start position:0% size of this is like 9 trillion DNA base pairs.<00:03:36.159> So<00:03:36.400> afterwards<00:03:36.879> they<00:03:37.120> used<00:03:37.360> all<00:03:37.599> this 00:03:37.830 --> 00:03:37.840 align:start position:0% pairs. So afterwards they used all this 00:03:37.840 --> 00:03:40.710 align:start position:0% pairs. So afterwards they used all this DNA<00:03:38.400> to<00:03:38.640> train<00:03:39.040> an<00:03:39.280> AI<00:03:39.680> model.<00:03:40.159> And<00:03:40.319> here's<00:03:40.560> the 00:03:40.710 --> 00:03:40.720 align:start position:0% DNA to train an AI model. And here's the 00:03:40.720 --> 00:03:42.789 align:start position:0% DNA to train an AI model. And here's the technical<00:03:41.200> detail<00:03:41.519> that<00:03:41.840> really<00:03:42.159> stands<00:03:42.480> out 00:03:42.789 --> 00:03:42.799 align:start position:0% technical detail that really stands out 00:03:42.799 --> 00:03:45.830 align:start position:0% technical detail that really stands out here.<00:03:43.040> It<00:03:43.280> says<00:03:43.440> that<00:03:43.760> this<00:03:44.080> EVO2<00:03:44.879> model<00:03:45.280> has<00:03:45.519> a 00:03:45.830 --> 00:03:45.840 align:start position:0% here. It says that this EVO2 model has a 00:03:45.840 --> 00:03:48.949 align:start position:0% here. It says that this EVO2 model has a million<00:03:46.239> token<00:03:46.959> context<00:03:47.440> window<00:03:48.080> with<00:03:48.480> single 00:03:48.949 --> 00:03:48.959 align:start position:0% million token context window with single 00:03:48.959 --> 00:03:51.430 align:start position:0% million token context window with single nucleotide<00:03:50.000> resolution.<00:03:50.640> In<00:03:50.879> simple<00:03:51.120> terms, 00:03:51.430 --> 00:03:51.440 align:start position:0% nucleotide resolution. In simple terms, 00:03:51.440 --> 00:03:55.110 align:start position:0% nucleotide resolution. In simple terms, that<00:03:51.680> means<00:03:52.000> the<00:03:52.239> AI<00:03:52.720> can<00:03:52.959> hold<00:03:53.680> 1<00:03:54.000> million<00:03:54.640> DNA 00:03:55.110 --> 00:03:55.120 align:start position:0% that means the AI can hold 1 million DNA 00:03:55.120 --> 00:03:57.589 align:start position:0% that means the AI can hold 1 million DNA letters<00:03:55.599> in<00:03:55.840> its<00:03:56.080> prompt<00:03:56.560> or<00:03:56.799> working<00:03:57.200> memory 00:03:57.589 --> 00:03:57.599 align:start position:0% letters in its prompt or working memory 00:03:57.599 --> 00:03:59.750 align:start position:0% letters in its prompt or working memory at<00:03:57.840> once.<00:03:58.480> Now,<00:03:58.720> why<00:03:58.959> would<00:03:59.120> we<00:03:59.280> need<00:03:59.439> that? 00:03:59.750 --> 00:03:59.760 align:start position:0% at once. Now, why would we need that? 00:03:59.760 --> 00:04:02.149 align:start position:0% at once. Now, why would we need that? After<00:04:00.000> all,<00:04:00.319> a<00:04:00.560> single<00:04:00.799> gene<00:04:01.200> might<00:04:01.439> only<00:04:01.680> be<00:04:01.840> a 00:04:02.149 --> 00:04:02.159 align:start position:0% After all, a single gene might only be a 00:04:02.159 --> 00:04:03.589 align:start position:0% After all, a single gene might only be a few<00:04:02.239> thousand<00:04:02.640> letters<00:04:03.040> long.<00:04:03.360> Shouldn't 00:04:03.589 --> 00:04:03.599 align:start position:0% few thousand letters long. Shouldn't 00:04:03.599 --> 00:04:05.830 align:start position:0% few thousand letters long. Shouldn't that<00:04:03.840> be<00:04:03.920> enough?<00:04:04.560> The<00:04:04.879> problem<00:04:05.120> is<00:04:05.519> that 00:04:05.830 --> 00:04:05.840 align:start position:0% that be enough? The problem is that 00:04:05.840 --> 00:04:08.550 align:start position:0% that be enough? The problem is that biology<00:04:06.560> almost<00:04:07.040> never<00:04:07.360> works<00:04:07.599> in<00:04:07.920> isolation. 00:04:08.550 --> 00:04:08.560 align:start position:0% biology almost never works in isolation. 00:04:08.560 --> 00:04:10.470 align:start position:0% biology almost never works in isolation. Here's<00:04:08.799> an<00:04:08.959> example.<00:04:09.519> You<00:04:09.680> see,<00:04:09.920> a<00:04:10.159> gene, 00:04:10.470 --> 00:04:10.480 align:start position:0% Here's an example. You see, a gene, 00:04:10.480 --> 00:04:12.470 align:start position:0% Here's an example. You see, a gene, which<00:04:10.720> is<00:04:10.879> just<00:04:11.040> a<00:04:11.280> section<00:04:11.439> of<00:04:11.680> DNA,<00:04:12.239> might 00:04:12.470 --> 00:04:12.480 align:start position:0% which is just a section of DNA, might 00:04:12.480 --> 00:04:14.309 align:start position:0% which is just a section of DNA, might contain<00:04:12.799> the<00:04:13.040> instructions<00:04:13.680> to<00:04:14.080> do 00:04:14.309 --> 00:04:14.319 align:start position:0% contain the instructions to do 00:04:14.319 --> 00:04:16.469 align:start position:0% contain the instructions to do something,<00:04:15.040> but<00:04:15.439> whether<00:04:15.760> that<00:04:16.000> gene<00:04:16.239> is 00:04:16.469 --> 00:04:16.479 align:start position:0% something, but whether that gene is 00:04:16.479 --> 00:04:18.710 align:start position:0% something, but whether that gene is actually<00:04:16.799> activated<00:04:17.359> or<00:04:17.680> how<00:04:18.000> strongly<00:04:18.400> it's 00:04:18.710 --> 00:04:18.720 align:start position:0% actually activated or how strongly it's 00:04:18.720 --> 00:04:20.949 align:start position:0% actually activated or how strongly it's executed<00:04:19.359> or<00:04:19.600> when<00:04:19.759> it<00:04:19.919> turns<00:04:20.160> on<00:04:20.560> often 00:04:20.949 --> 00:04:20.959 align:start position:0% executed or when it turns on often 00:04:20.959 --> 00:04:24.070 align:start position:0% executed or when it turns on often depends<00:04:21.359> on<00:04:21.759> other<00:04:22.160> regions<00:04:22.800> of<00:04:23.040> DNA<00:04:23.680> called 00:04:24.070 --> 00:04:24.080 align:start position:0% depends on other regions of DNA called 00:04:24.080 --> 00:04:26.310 align:start position:0% depends on other regions of DNA called regulatory<00:04:24.880> elements.<00:04:25.440> And<00:04:25.600> these<00:04:25.840> elements 00:04:26.310 --> 00:04:26.320 align:start position:0% regulatory elements. And these elements 00:04:26.320 --> 00:04:28.150 align:start position:0% regulatory elements. And these elements aren't<00:04:26.639> always<00:04:27.040> nearby.<00:04:27.600> They<00:04:27.840> might<00:04:28.000> be 00:04:28.150 --> 00:04:28.160 align:start position:0% aren't always nearby. They might be 00:04:28.160 --> 00:04:30.629 align:start position:0% aren't always nearby. They might be hundreds<00:04:28.560> of<00:04:28.800> thousands<00:04:29.360> of<00:04:29.680> letters<00:04:30.160> away 00:04:30.629 --> 00:04:30.639 align:start position:0% hundreds of thousands of letters away 00:04:30.639 --> 00:04:32.550 align:start position:0% hundreds of thousands of letters away along<00:04:30.960> the<00:04:31.199> DNA<00:04:31.600> strand.<00:04:32.160> So<00:04:32.400> the 00:04:32.550 --> 00:04:32.560 align:start position:0% along the DNA strand. So the 00:04:32.560 --> 00:04:34.390 align:start position:0% along the DNA strand. So the instructions<00:04:33.120> and<00:04:33.360> the<00:04:33.520> controls<00:04:34.000> can<00:04:34.240> be 00:04:34.390 --> 00:04:34.400 align:start position:0% instructions and the controls can be 00:04:34.400 --> 00:04:36.950 align:start position:0% instructions and the controls can be separated<00:04:34.880> by<00:04:35.199> enormous<00:04:35.840> distances.<00:04:36.639> If<00:04:36.800> the 00:04:36.950 --> 00:04:36.960 align:start position:0% separated by enormous distances. If the 00:04:36.960 --> 00:04:39.189 align:start position:0% separated by enormous distances. If the model's<00:04:37.360> context<00:04:37.759> window<00:04:38.160> is<00:04:38.400> small,<00:04:38.960> say 00:04:39.189 --> 00:04:39.199 align:start position:0% model's context window is small, say 00:04:39.199 --> 00:04:41.590 align:start position:0% model's context window is small, say only<00:04:39.600> a<00:04:39.919> few<00:04:40.080> hundred<00:04:40.320> thousand<00:04:40.720> tokens,<00:04:41.440> it 00:04:41.590 --> 00:04:41.600 align:start position:0% only a few hundred thousand tokens, it 00:04:41.600 --> 00:04:43.590 align:start position:0% only a few hundred thousand tokens, it might<00:04:41.840> not<00:04:42.000> be<00:04:42.240> able<00:04:42.479> to<00:04:42.720> capture<00:04:43.120> all<00:04:43.360> the 00:04:43.590 --> 00:04:43.600 align:start position:0% might not be able to capture all the 00:04:43.600 --> 00:04:45.670 align:start position:0% might not be able to capture all the relevant<00:04:44.000> parts<00:04:44.240> of<00:04:44.400> a<00:04:44.560> DNA<00:04:44.960> strand.<00:04:45.440> Because 00:04:45.670 --> 00:04:45.680 align:start position:0% relevant parts of a DNA strand. Because 00:04:45.680 --> 00:04:48.550 align:start position:0% relevant parts of a DNA strand. Because this<00:04:46.000> new<00:04:46.160> EVO2<00:04:46.880> has<00:04:47.040> a<00:04:47.280> much<00:04:47.440> larger<00:04:48.080> million 00:04:48.550 --> 00:04:48.560 align:start position:0% this new EVO2 has a much larger million 00:04:48.560 --> 00:04:50.790 align:start position:0% this new EVO2 has a much larger million token<00:04:48.960> context<00:04:49.440> window,<00:04:49.919> it<00:04:50.160> can<00:04:50.320> understand 00:04:50.790 --> 00:04:50.800 align:start position:0% token context window, it can understand 00:04:50.800 --> 00:04:53.990 align:start position:0% token context window, it can understand context<00:04:51.360> across<00:04:52.160> DNA<00:04:52.720> a<00:04:52.960> lot<00:04:53.199> better.<00:04:53.680> Now, 00:04:53.990 --> 00:04:54.000 align:start position:0% context across DNA a lot better. Now, 00:04:54.000 --> 00:04:55.749 align:start position:0% context across DNA a lot better. Now, holding<00:04:54.320> that<00:04:54.560> much<00:04:54.800> information<00:04:55.199> is<00:04:55.520> one 00:04:55.749 --> 00:04:55.759 align:start position:0% holding that much information is one 00:04:55.759 --> 00:04:57.270 align:start position:0% holding that much information is one thing,<00:04:56.000> but<00:04:56.320> actually<00:04:56.639> being<00:04:56.880> able<00:04:57.040> to 00:04:57.270 --> 00:04:57.280 align:start position:0% thing, but actually being able to 00:04:57.280 --> 00:04:59.510 align:start position:0% thing, but actually being able to understand,<00:04:58.080> remember,<00:04:58.479> and<00:04:58.800> analyze<00:04:59.280> all 00:04:59.510 --> 00:04:59.520 align:start position:0% understand, remember, and analyze all 00:04:59.520 --> 00:05:01.749 align:start position:0% understand, remember, and analyze all this<00:04:59.759> data<00:05:00.080> is<00:05:00.400> something<00:05:00.639> else.<00:05:01.199> To<00:05:01.440> prove 00:05:01.749 --> 00:05:01.759 align:start position:0% this data is something else. To prove 00:05:01.759 --> 00:05:03.830 align:start position:0% this data is something else. To prove their<00:05:02.000> system<00:05:02.400> could<00:05:02.720> actually<00:05:03.040> handle<00:05:03.520> a 00:05:03.830 --> 00:05:03.840 align:start position:0% their system could actually handle a 00:05:03.840 --> 00:05:05.830 align:start position:0% their system could actually handle a million<00:05:04.160> DNA<00:05:04.639> letters<00:05:04.960> at<00:05:05.199> once,<00:05:05.600> the 00:05:05.830 --> 00:05:05.840 align:start position:0% million DNA letters at once, the 00:05:05.840 --> 00:05:07.990 align:start position:0% million DNA letters at once, the researchers<00:05:06.320> ran<00:05:06.639> a<00:05:06.880> simple<00:05:07.199> but<00:05:07.520> brutal 00:05:07.990 --> 00:05:08.000 align:start position:0% researchers ran a simple but brutal 00:05:08.000 --> 00:05:09.830 align:start position:0% researchers ran a simple but brutal test.<00:05:08.479> Something<00:05:08.800> called<00:05:09.039> a<00:05:09.280> needle<00:05:09.600> in 00:05:09.830 --> 00:05:09.840 align:start position:0% test. Something called a needle in 00:05:09.840 --> 00:05:11.749 align:start position:0% test. Something called a needle in haststack<00:05:10.400> test.<00:05:10.880> It's<00:05:11.120> a<00:05:11.280> very<00:05:11.440> standard 00:05:11.749 --> 00:05:11.759 align:start position:0% haststack test. It's a very standard 00:05:11.759 --> 00:05:14.230 align:start position:0% haststack test. It's a very standard test<00:05:12.000> in<00:05:12.160> AI<00:05:12.720> and<00:05:13.039> basically<00:05:13.440> they<00:05:13.680> generated 00:05:14.230 --> 00:05:14.240 align:start position:0% test in AI and basically they generated 00:05:14.240 --> 00:05:16.310 align:start position:0% test in AI and basically they generated a<00:05:14.560> completely<00:05:15.039> random<00:05:15.520> and<00:05:15.759> meaningless 00:05:16.310 --> 00:05:16.320 align:start position:0% a completely random and meaningless 00:05:16.320 --> 00:05:19.270 align:start position:0% a completely random and meaningless sequence<00:05:16.800> of<00:05:17.280> 1<00:05:17.520> million<00:05:18.000> DNA<00:05:18.639> letters. 00:05:19.270 --> 00:05:19.280 align:start position:0% sequence of 1 million DNA letters. 00:05:19.280 --> 00:05:21.350 align:start position:0% sequence of 1 million DNA letters. That's<00:05:19.520> the<00:05:19.680> haystack.<00:05:20.320> Then<00:05:20.560> they<00:05:20.800> took<00:05:21.039> a 00:05:21.350 --> 00:05:21.360 align:start position:0% That's the haystack. Then they took a 00:05:21.360 --> 00:05:24.790 align:start position:0% That's the haystack. Then they took a very<00:05:21.600> specific<00:05:22.400> 100let<00:05:23.280> sequence<00:05:23.919> and<00:05:24.160> hid<00:05:24.479> it 00:05:24.790 --> 00:05:24.800 align:start position:0% very specific 100let sequence and hid it 00:05:24.800 --> 00:05:27.029 align:start position:0% very specific 100let sequence and hid it somewhere<00:05:25.280> inside<00:05:25.840> this<00:05:26.160> massive<00:05:26.639> random 00:05:27.029 --> 00:05:27.039 align:start position:0% somewhere inside this massive random 00:05:27.039 --> 00:05:29.670 align:start position:0% somewhere inside this massive random jumble<00:05:27.440> of<00:05:27.600> letters.<00:05:28.240> This<00:05:28.560> 100let<00:05:29.280> sequence 00:05:29.670 --> 00:05:29.680 align:start position:0% jumble of letters. This 100let sequence 00:05:29.680 --> 00:05:32.070 align:start position:0% jumble of letters. This 100let sequence is<00:05:29.840> the<00:05:30.080> needle.<00:05:30.479> and<00:05:30.639> they<00:05:30.880> asked<00:05:31.039> Evo<00:05:31.600> 2<00:05:31.840> to 00:05:32.070 --> 00:05:32.080 align:start position:0% is the needle. and they asked Evo 2 to 00:05:32.080 --> 00:05:35.350 align:start position:0% is the needle. and they asked Evo 2 to find<00:05:32.320> it.<00:05:32.960> And<00:05:33.120> in<00:05:33.440> short,<00:05:33.759> EVO<00:05:34.240> 2<00:05:34.479> was<00:05:34.639> able<00:05:34.960> to 00:05:35.350 --> 00:05:35.360 align:start position:0% find it. And in short, EVO 2 was able to 00:05:35.360 --> 00:05:37.830 align:start position:0% find it. And in short, EVO 2 was able to find<00:05:35.680> it<00:05:36.000> perfectly.<00:05:36.720> This<00:05:36.960> proves<00:05:37.280> that<00:05:37.520> it's 00:05:37.830 --> 00:05:37.840 align:start position:0% find it perfectly. This proves that it's 00:05:37.840 --> 00:05:39.510 align:start position:0% find it perfectly. This proves that it's not<00:05:38.080> just<00:05:38.320> skimming<00:05:38.720> the<00:05:38.960> sequence<00:05:39.280> or 00:05:39.510 --> 00:05:39.520 align:start position:0% not just skimming the sequence or 00:05:39.520 --> 00:05:41.749 align:start position:0% not just skimming the sequence or forgetting<00:05:40.000> stuff<00:05:40.320> in<00:05:40.560> this<00:05:40.960> 1<00:05:41.199> million 00:05:41.749 --> 00:05:41.759 align:start position:0% forgetting stuff in this 1 million 00:05:41.759 --> 00:05:43.830 align:start position:0% forgetting stuff in this 1 million letter<00:05:42.160> sequence.<00:05:42.880> It's<00:05:43.120> actually<00:05:43.520> reading 00:05:43.830 --> 00:05:43.840 align:start position:0% letter sequence. It's actually reading 00:05:43.840 --> 00:05:46.310 align:start position:0% letter sequence. It's actually reading and<00:05:44.160> retaining<00:05:44.560> the<00:05:44.880> entire<00:05:45.199> thing.<00:05:45.759> So,<00:05:46.000> at 00:05:46.310 --> 00:05:46.320 align:start position:0% and retaining the entire thing. So, at 00:05:46.320 --> 00:05:48.150 align:start position:0% and retaining the entire thing. So, at this<00:05:46.479> point,<00:05:46.720> we<00:05:46.960> know<00:05:47.199> a<00:05:47.440> few<00:05:47.600> things.<00:05:48.000> We 00:05:48.150 --> 00:05:48.160 align:start position:0% this point, we know a few things. We 00:05:48.160 --> 00:05:50.310 align:start position:0% this point, we know a few things. We know<00:05:48.479> this<00:05:48.720> model<00:05:49.039> was<00:05:49.199> trained<00:05:49.520> on<00:05:50.000> 9 00:05:50.310 --> 00:05:50.320 align:start position:0% know this model was trained on 9 00:05:50.320 --> 00:05:53.029 align:start position:0% know this model was trained on 9 trillion<00:05:50.880> DNA<00:05:51.360> letters<00:05:51.840> across<00:05:52.320> the<00:05:52.639> entire 00:05:53.029 --> 00:05:53.039 align:start position:0% trillion DNA letters across the entire 00:05:53.039 --> 00:05:55.189 align:start position:0% trillion DNA letters across the entire spectrum<00:05:53.440> of<00:05:53.600> life<00:05:54.000> from<00:05:54.320> extreme<00:05:54.720> bacteria 00:05:55.189 --> 00:05:55.199 align:start position:0% spectrum of life from extreme bacteria 00:05:55.199 --> 00:05:57.270 align:start position:0% spectrum of life from extreme bacteria to<00:05:55.520> human<00:05:55.919> cells.<00:05:56.400> We<00:05:56.639> know<00:05:56.720> that<00:05:56.960> it<00:05:57.120> can 00:05:57.270 --> 00:05:57.280 align:start position:0% to human cells. We know that it can 00:05:57.280 --> 00:05:58.870 align:start position:0% to human cells. We know that it can remember<00:05:57.600> what<00:05:57.759> it<00:05:57.919> reads<00:05:58.320> even<00:05:58.479> if<00:05:58.639> it's 00:05:58.870 --> 00:05:58.880 align:start position:0% remember what it reads even if it's 00:05:58.880 --> 00:06:01.270 align:start position:0% remember what it reads even if it's given<00:05:59.120> a<00:05:59.440> million<00:05:59.919> letters.<00:06:00.639> But<00:06:00.800> here's<00:06:01.120> the 00:06:01.270 --> 00:06:01.280 align:start position:0% given a million letters. But here's the 00:06:01.280 --> 00:06:03.110 align:start position:0% given a million letters. But here's the deeper<00:06:01.680> question.<00:06:02.160> Can<00:06:02.400> it<00:06:02.560> actually 00:06:03.110 --> 00:06:03.120 align:start position:0% deeper question. Can it actually 00:06:03.120 --> 00:06:05.189 align:start position:0% deeper question. Can it actually recognize<00:06:03.680> patterns?<00:06:04.319> Can<00:06:04.560> it<00:06:04.720> actually 00:06:05.189 --> 00:06:05.199 align:start position:0% recognize patterns? Can it actually 00:06:05.199 --> 00:06:08.070 align:start position:0% recognize patterns? Can it actually understand<00:06:05.919> the<00:06:06.240> language<00:06:06.560> of<00:06:06.800> DNA?<00:06:07.600> You<00:06:07.759> see, 00:06:08.070 --> 00:06:08.080 align:start position:0% understand the language of DNA? You see, 00:06:08.080 --> 00:06:10.629 align:start position:0% understand the language of DNA? You see, when<00:06:08.319> Evo2<00:06:08.960> was<00:06:09.199> trained,<00:06:09.680> it<00:06:09.919> wasn't<00:06:10.319> given 00:06:10.629 --> 00:06:10.639 align:start position:0% when Evo2 was trained, it wasn't given 00:06:10.639 --> 00:06:12.870 align:start position:0% when Evo2 was trained, it wasn't given any<00:06:10.960> labels<00:06:11.360> or<00:06:11.680> answers.<00:06:12.160> It<00:06:12.319> was<00:06:12.479> never<00:06:12.639> told 00:06:12.870 --> 00:06:12.880 align:start position:0% any labels or answers. It was never told 00:06:12.880 --> 00:06:15.029 align:start position:0% any labels or answers. It was never told that<00:06:13.280> this<00:06:13.680> particular<00:06:14.160> sequence<00:06:14.479> of<00:06:14.639> DNA 00:06:15.029 --> 00:06:15.039 align:start position:0% that this particular sequence of DNA 00:06:15.039 --> 00:06:17.350 align:start position:0% that this particular sequence of DNA triggers<00:06:15.360> a<00:06:15.600> disease<00:06:16.000> or<00:06:16.319> this<00:06:16.639> one<00:06:16.960> is<00:06:17.199> a 00:06:17.350 --> 00:06:17.360 align:start position:0% triggers a disease or this one is a 00:06:17.360 --> 00:06:19.350 align:start position:0% triggers a disease or this one is a mutation<00:06:17.840> that<00:06:18.080> could<00:06:18.319> kill<00:06:18.560> the<00:06:18.800> cell<00:06:19.199> or 00:06:19.350 --> 00:06:19.360 align:start position:0% mutation that could kill the cell or 00:06:19.360 --> 00:06:21.670 align:start position:0% mutation that could kill the cell or that<00:06:19.759> this<00:06:20.000> one,<00:06:20.400> for<00:06:20.479> example,<00:06:21.199> controls 00:06:21.670 --> 00:06:21.680 align:start position:0% that this one, for example, controls 00:06:21.680 --> 00:06:23.830 align:start position:0% that this one, for example, controls leaf<00:06:22.000> color.<00:06:22.400> The<00:06:22.639> model<00:06:23.039> never<00:06:23.360> received 00:06:23.830 --> 00:06:23.840 align:start position:0% leaf color. The model never received 00:06:23.840 --> 00:06:26.550 align:start position:0% leaf color. The model never received these<00:06:24.160> labels<00:06:24.639> or<00:06:24.880> answers.<00:06:25.440> It<00:06:25.759> simply<00:06:26.080> read 00:06:26.550 --> 00:06:26.560 align:start position:0% these labels or answers. It simply read 00:06:26.560 --> 00:06:29.909 align:start position:0% these labels or answers. It simply read raw<00:06:26.960> unlabeled<00:06:27.680> DNA<00:06:28.319> like<00:06:28.560> this.<00:06:29.280> So<00:06:29.600> how<00:06:29.759> can 00:06:29.909 --> 00:06:29.919 align:start position:0% raw unlabeled DNA like this. So how can 00:06:29.919 --> 00:06:32.390 align:start position:0% raw unlabeled DNA like this. So how can we<00:06:30.160> tell<00:06:30.560> whether<00:06:30.960> it<00:06:31.199> actually<00:06:31.600> understands 00:06:32.390 --> 00:06:32.400 align:start position:0% we tell whether it actually understands 00:06:32.400 --> 00:06:34.870 align:start position:0% we tell whether it actually understands DNA?<00:06:33.120> This<00:06:33.280> is<00:06:33.360> where<00:06:33.600> a<00:06:33.840> concept<00:06:34.319> called 00:06:34.870 --> 00:06:34.880 align:start position:0% DNA? This is where a concept called 00:06:34.880 --> 00:06:37.749 align:start position:0% DNA? This is where a concept called zeroshot<00:06:35.840> prediction<00:06:36.479> comes<00:06:36.720> in.<00:06:37.120> Zero<00:06:37.520> shot 00:06:37.749 --> 00:06:37.759 align:start position:0% zeroshot prediction comes in. Zero shot 00:06:37.759 --> 00:06:39.189 align:start position:0% zeroshot prediction comes in. Zero shot means<00:06:38.000> the<00:06:38.160> AI<00:06:38.479> can<00:06:38.639> answer<00:06:38.960> something 00:06:39.189 --> 00:06:39.199 align:start position:0% means the AI can answer something 00:06:39.199 --> 00:06:41.830 align:start position:0% means the AI can answer something without<00:06:39.600> being<00:06:40.000> explicitly<00:06:40.800> trained<00:06:41.280> on<00:06:41.520> it. 00:06:41.830 --> 00:06:41.840 align:start position:0% without being explicitly trained on it. 00:06:41.840 --> 00:06:43.990 align:start position:0% without being explicitly trained on it. But<00:06:42.000> that<00:06:42.160> raises<00:06:42.560> an<00:06:42.720> obvious<00:06:43.039> question.<00:06:43.759> If 00:06:43.990 --> 00:06:44.000 align:start position:0% But that raises an obvious question. If 00:06:44.000 --> 00:06:45.990 align:start position:0% But that raises an obvious question. If the<00:06:44.240> model<00:06:44.479> was<00:06:44.880> never<00:06:45.199> taught<00:06:45.600> what<00:06:45.759> a 00:06:45.990 --> 00:06:46.000 align:start position:0% the model was never taught what a 00:06:46.000 --> 00:06:48.230 align:start position:0% the model was never taught what a disease<00:06:46.319> is,<00:06:46.720> how<00:06:46.880> could<00:06:47.039> it<00:06:47.280> possibly<00:06:47.759> know 00:06:48.230 --> 00:06:48.240 align:start position:0% disease is, how could it possibly know 00:06:48.240 --> 00:06:50.230 align:start position:0% disease is, how could it possibly know that<00:06:48.479> a<00:06:48.720> certain<00:06:49.039> snippet<00:06:49.360> of<00:06:49.520> DNA<00:06:50.000> would 00:06:50.230 --> 00:06:50.240 align:start position:0% that a certain snippet of DNA would 00:06:50.240 --> 00:06:52.710 align:start position:0% that a certain snippet of DNA would cause<00:06:50.479> disease?<00:06:51.199> The<00:06:51.440> answer<00:06:51.919> lies<00:06:52.319> in 00:06:52.710 --> 00:06:52.720 align:start position:0% cause disease? The answer lies in 00:06:52.720 --> 00:06:54.950 align:start position:0% cause disease? The answer lies in evolution.<00:06:53.440> If<00:06:53.680> a<00:06:53.840> genetic<00:06:54.240> sequence<00:06:54.720> is 00:06:54.950 --> 00:06:54.960 align:start position:0% evolution. If a genetic sequence is 00:06:54.960 --> 00:06:57.110 align:start position:0% evolution. If a genetic sequence is absolutely<00:06:55.840> essential<00:06:56.240> for<00:06:56.560> life,<00:06:57.039> for 00:06:57.110 --> 00:06:57.120 align:start position:0% absolutely essential for life, for 00:06:57.120 --> 00:06:59.270 align:start position:0% absolutely essential for life, for example,<00:06:57.600> the<00:06:57.919> proteins<00:06:58.319> that<00:06:58.560> allow<00:06:58.800> a<00:06:59.039> cell 00:06:59.270 --> 00:06:59.280 align:start position:0% example, the proteins that allow a cell 00:06:59.280 --> 00:07:01.909 align:start position:0% example, the proteins that allow a cell to<00:06:59.680> generate<00:07:00.080> energy,<00:07:00.720> evolution<00:07:01.360> tends<00:07:01.759> to 00:07:01.909 --> 00:07:01.919 align:start position:0% to generate energy, evolution tends to 00:07:01.919 --> 00:07:04.230 align:start position:0% to generate energy, evolution tends to keep<00:07:02.160> that<00:07:02.400> sequence<00:07:02.960> from<00:07:03.280> change.<00:07:03.919> You<00:07:04.080> see, 00:07:04.230 --> 00:07:04.240 align:start position:0% keep that sequence from change. You see, 00:07:04.240 --> 00:07:06.309 align:start position:0% keep that sequence from change. You see, all<00:07:04.400> the<00:07:04.560> living<00:07:04.880> things<00:07:05.120> in<00:07:05.360> this<00:07:05.599> data<00:07:05.919> set 00:07:06.309 --> 00:07:06.319 align:start position:0% all the living things in this data set 00:07:06.319 --> 00:07:09.189 align:start position:0% all the living things in this data set are<00:07:06.639> alive<00:07:07.120> because<00:07:07.520> its<00:07:07.840> DNA<00:07:08.319> works<00:07:08.639> and<00:07:08.960> is 00:07:09.189 --> 00:07:09.199 align:start position:0% are alive because its DNA works and is 00:07:09.199 --> 00:07:11.189 align:start position:0% are alive because its DNA works and is functional.<00:07:09.680> If<00:07:09.840> a<00:07:10.080> mutation<00:07:10.560> messed<00:07:10.880> up<00:07:11.039> the 00:07:11.189 --> 00:07:11.199 align:start position:0% functional. If a mutation messed up the 00:07:11.199 --> 00:07:13.510 align:start position:0% functional. If a mutation messed up the DNA<00:07:11.520> in<00:07:11.680> a<00:07:11.840> way<00:07:12.000> that<00:07:12.240> could<00:07:12.479> cause<00:07:12.880> disease<00:07:13.280> or 00:07:13.510 --> 00:07:13.520 align:start position:0% DNA in a way that could cause disease or 00:07:13.520 --> 00:07:16.070 align:start position:0% DNA in a way that could cause disease or death,<00:07:13.919> well,<00:07:14.400> that<00:07:14.720> organism<00:07:15.360> dies<00:07:15.680> and<00:07:15.919> that 00:07:16.070 --> 00:07:16.080 align:start position:0% death, well, that organism dies and that 00:07:16.080 --> 00:07:17.990 align:start position:0% death, well, that organism dies and that mutation<00:07:16.639> wouldn't<00:07:17.039> appear<00:07:17.280> in<00:07:17.520> this<00:07:17.680> gene 00:07:17.990 --> 00:07:18.000 align:start position:0% mutation wouldn't appear in this gene 00:07:18.000 --> 00:07:19.830 align:start position:0% mutation wouldn't appear in this gene pool.<00:07:18.479> So<00:07:18.720> over<00:07:19.039> millions<00:07:19.360> of<00:07:19.520> years<00:07:19.680> of 00:07:19.830 --> 00:07:19.840 align:start position:0% pool. So over millions of years of 00:07:19.840 --> 00:07:22.309 align:start position:0% pool. So over millions of years of evolution,<00:07:20.560> this<00:07:20.880> process<00:07:21.440> leaves<00:07:21.759> a<00:07:22.000> clear 00:07:22.309 --> 00:07:22.319 align:start position:0% evolution, this process leaves a clear 00:07:22.319 --> 00:07:24.710 align:start position:0% evolution, this process leaves a clear signal<00:07:22.720> in<00:07:22.880> the<00:07:23.039> data.<00:07:23.360> When<00:07:23.599> Evo<00:07:24.080> 2<00:07:24.400> reads 00:07:24.710 --> 00:07:24.720 align:start position:0% signal in the data. When Evo 2 reads 00:07:24.720 --> 00:07:27.830 align:start position:0% signal in the data. When Evo 2 reads these<00:07:25.280> 9<00:07:25.680> trillion<00:07:26.400> DNA<00:07:26.960> letters<00:07:27.280> from<00:07:27.520> across 00:07:27.830 --> 00:07:27.840 align:start position:0% these 9 trillion DNA letters from across 00:07:27.840 --> 00:07:30.150 align:start position:0% these 9 trillion DNA letters from across the<00:07:28.160> entire<00:07:28.479> spectrum<00:07:28.880> of<00:07:29.039> life,<00:07:29.680> it<00:07:29.840> begins 00:07:30.150 --> 00:07:30.160 align:start position:0% the entire spectrum of life, it begins 00:07:30.160 --> 00:07:32.309 align:start position:0% the entire spectrum of life, it begins to<00:07:30.319> notice<00:07:30.720> these<00:07:31.039> patterns.<00:07:31.520> If<00:07:31.680> it<00:07:31.840> notices 00:07:32.309 --> 00:07:32.319 align:start position:0% to notice these patterns. If it notices 00:07:32.319 --> 00:07:35.189 align:start position:0% to notice these patterns. If it notices the<00:07:32.639> exact<00:07:32.960> same<00:07:33.360> sequence<00:07:33.759> of<00:07:34.000> DNA<00:07:34.479> in<00:07:34.880> many 00:07:35.189 --> 00:07:35.199 align:start position:0% the exact same sequence of DNA in many 00:07:35.199 --> 00:07:37.110 align:start position:0% the exact same sequence of DNA in many different<00:07:35.520> species,<00:07:36.160> it<00:07:36.319> would<00:07:36.479> signal<00:07:36.800> that 00:07:37.110 --> 00:07:37.120 align:start position:0% different species, it would signal that 00:07:37.120 --> 00:07:39.110 align:start position:0% different species, it would signal that this<00:07:37.360> sequence<00:07:37.919> is<00:07:38.240> probably<00:07:38.720> very 00:07:39.110 --> 00:07:39.120 align:start position:0% this sequence is probably very 00:07:39.120 --> 00:07:41.270 align:start position:0% this sequence is probably very important.<00:07:39.840> And<00:07:40.080> conversely,<00:07:40.560> if<00:07:40.800> it<00:07:41.039> has 00:07:41.270 --> 00:07:41.280 align:start position:0% important. And conversely, if it has 00:07:41.280 --> 00:07:43.510 align:start position:0% important. And conversely, if it has never<00:07:41.599> seen<00:07:41.919> a<00:07:42.160> sequence<00:07:42.560> before<00:07:42.880> in<00:07:43.199> this 00:07:43.510 --> 00:07:43.520 align:start position:0% never seen a sequence before in this 00:07:43.520 --> 00:07:45.350 align:start position:0% never seen a sequence before in this spectrum<00:07:43.840> of<00:07:44.080> life,<00:07:44.400> it<00:07:44.639> might<00:07:44.800> mean<00:07:45.039> that 00:07:45.350 --> 00:07:45.360 align:start position:0% spectrum of life, it might mean that 00:07:45.360 --> 00:07:47.670 align:start position:0% spectrum of life, it might mean that that<00:07:45.759> sequence<00:07:46.319> could<00:07:46.560> be<00:07:46.720> harmful<00:07:47.039> or<00:07:47.360> cause 00:07:47.670 --> 00:07:47.680 align:start position:0% that sequence could be harmful or cause 00:07:47.680 --> 00:07:49.749 align:start position:0% that sequence could be harmful or cause death.<00:07:48.160> Anyways,<00:07:48.560> the<00:07:48.800> researchers<00:07:49.280> had<00:07:49.520> to 00:07:49.749 --> 00:07:49.759 align:start position:0% death. Anyways, the researchers had to 00:07:49.759 --> 00:07:51.990 align:start position:0% death. Anyways, the researchers had to prove<00:07:50.000> this.<00:07:50.400> They<00:07:50.560> had<00:07:50.720> to<00:07:50.880> show<00:07:50.960> that<00:07:51.199> Evo<00:07:51.680> 2 00:07:51.990 --> 00:07:52.000 align:start position:0% prove this. They had to show that Evo 2 00:07:52.000 --> 00:07:54.469 align:start position:0% prove this. They had to show that Evo 2 actually<00:07:52.400> understood<00:07:52.960> the<00:07:53.199> language<00:07:53.599> of<00:07:53.840> DNA. 00:07:54.469 --> 00:07:54.479 align:start position:0% actually understood the language of DNA. 00:07:54.479 --> 00:07:57.350 align:start position:0% actually understood the language of DNA. So<00:07:54.720> to<00:07:54.960> test<00:07:55.120> it,<00:07:55.680> they<00:07:55.919> took<00:07:56.240> a<00:07:56.479> real<00:07:56.800> DNA 00:07:57.350 --> 00:07:57.360 align:start position:0% So to test it, they took a real DNA 00:07:57.360 --> 00:08:00.309 align:start position:0% So to test it, they took a real DNA sequence<00:07:58.080> and<00:07:58.400> changed<00:07:59.039> one<00:07:59.280> letter<00:07:59.759> and<00:08:00.000> then 00:08:00.309 --> 00:08:00.319 align:start position:0% sequence and changed one letter and then 00:08:00.319 --> 00:08:02.309 align:start position:0% sequence and changed one letter and then asked<00:08:00.639> the<00:08:00.800> model<00:08:01.120> a<00:08:01.360> simple<00:08:01.599> question.<00:08:02.080> Based 00:08:02.309 --> 00:08:02.319 align:start position:0% asked the model a simple question. Based 00:08:02.319 --> 00:08:04.390 align:start position:0% asked the model a simple question. Based on<00:08:02.479> your<00:08:02.720> understanding,<00:08:03.360> how<00:08:03.599> likely<00:08:04.080> will 00:08:04.390 --> 00:08:04.400 align:start position:0% on your understanding, how likely will 00:08:04.400 --> 00:08:07.909 align:start position:0% on your understanding, how likely will this<00:08:04.800> DNA<00:08:05.520> exist<00:08:05.840> in<00:08:06.080> nature?<00:08:06.800> If<00:08:07.039> EVO<00:08:07.520> assigns 00:08:07.909 --> 00:08:07.919 align:start position:0% this DNA exist in nature? If EVO assigns 00:08:07.919 --> 00:08:11.029 align:start position:0% this DNA exist in nature? If EVO assigns an<00:08:08.160> extremely<00:08:08.879> low<00:08:09.360> probability,<00:08:10.240> the<00:08:10.479> model 00:08:11.029 --> 00:08:11.039 align:start position:0% an extremely low probability, the model 00:08:11.039 --> 00:08:13.350 align:start position:0% an extremely low probability, the model essentially<00:08:11.520> is<00:08:11.840> raising<00:08:12.160> a<00:08:12.400> red<00:08:12.639> flag.<00:08:13.120> It's 00:08:13.350 --> 00:08:13.360 align:start position:0% essentially is raising a red flag. It's 00:08:13.360 --> 00:08:15.430 align:start position:0% essentially is raising a red flag. It's saying<00:08:13.599> like<00:08:13.919> this<00:08:14.160> mutation<00:08:14.639> is<00:08:14.960> probably 00:08:15.430 --> 00:08:15.440 align:start position:0% saying like this mutation is probably 00:08:15.440 --> 00:08:17.430 align:start position:0% saying like this mutation is probably harmful<00:08:16.000> because<00:08:16.240> it<00:08:16.479> wouldn't<00:08:16.879> exist<00:08:17.199> in 00:08:17.430 --> 00:08:17.440 align:start position:0% harmful because it wouldn't exist in 00:08:17.440 --> 00:08:19.589 align:start position:0% harmful because it wouldn't exist in nature.<00:08:18.080> Well,<00:08:18.240> when<00:08:18.560> researchers<00:08:19.120> tested 00:08:19.589 --> 00:08:19.599 align:start position:0% nature. Well, when researchers tested 00:08:19.599 --> 00:08:21.830 align:start position:0% nature. Well, when researchers tested EVO<00:08:20.080> 2<00:08:20.240> this<00:08:20.479> way,<00:08:20.800> they<00:08:21.039> found<00:08:21.360> some 00:08:21.830 --> 00:08:21.840 align:start position:0% EVO 2 this way, they found some 00:08:21.840 --> 00:08:23.830 align:start position:0% EVO 2 this way, they found some incredible<00:08:22.560> things.<00:08:23.039> The<00:08:23.280> model<00:08:23.520> was 00:08:23.830 --> 00:08:23.840 align:start position:0% incredible things. The model was 00:08:23.840 --> 00:08:26.390 align:start position:0% incredible things. The model was recognizing<00:08:24.560> very<00:08:24.960> specific<00:08:25.680> biological 00:08:26.390 --> 00:08:26.400 align:start position:0% recognizing very specific biological 00:08:26.400 --> 00:08:29.189 align:start position:0% recognizing very specific biological signals<00:08:26.960> without<00:08:27.440> ever<00:08:27.919> being<00:08:28.240> taught<00:08:28.560> this. 00:08:29.189 --> 00:08:29.199 align:start position:0% signals without ever being taught this. 00:08:29.199 --> 00:08:31.510 align:start position:0% signals without ever being taught this. For<00:08:29.360> example,<00:08:29.919> it<00:08:30.160> correctly<00:08:30.879> identified 00:08:31.510 --> 00:08:31.520 align:start position:0% For example, it correctly identified 00:08:31.520 --> 00:08:33.990 align:start position:0% For example, it correctly identified mutations<00:08:32.240> in<00:08:32.640> start<00:08:33.039> codons<00:08:33.519> and<00:08:33.680> stop 00:08:33.990 --> 00:08:34.000 align:start position:0% mutations in start codons and stop 00:08:34.000 --> 00:08:36.389 align:start position:0% mutations in start codons and stop codons.<00:08:34.560> To<00:08:34.880> understand<00:08:35.360> why<00:08:35.680> this<00:08:36.000> matters, 00:08:36.389 --> 00:08:36.399 align:start position:0% codons. To understand why this matters, 00:08:36.399 --> 00:08:38.550 align:start position:0% codons. To understand why this matters, we<00:08:36.560> need<00:08:36.719> to<00:08:36.880> briefly<00:08:37.360> look<00:08:37.599> at<00:08:37.839> how<00:08:38.240> cells 00:08:38.550 --> 00:08:38.560 align:start position:0% we need to briefly look at how cells 00:08:38.560 --> 00:08:41.350 align:start position:0% we need to briefly look at how cells read<00:08:38.880> DNA.<00:08:39.519> You<00:08:39.680> see,<00:08:39.919> DNA<00:08:40.399> is<00:08:40.560> interpreted<00:08:41.039> in 00:08:41.350 --> 00:08:41.360 align:start position:0% read DNA. You see, DNA is interpreted in 00:08:41.360 --> 00:08:43.670 align:start position:0% read DNA. You see, DNA is interpreted in groups<00:08:41.599> of<00:08:41.839> three<00:08:42.159> letters<00:08:42.719> called<00:08:42.959> codons. 00:08:43.670 --> 00:08:43.680 align:start position:0% groups of three letters called codons. 00:08:43.680 --> 00:08:45.750 align:start position:0% groups of three letters called codons. Now,<00:08:43.839> a<00:08:44.080> start<00:08:44.320> codon,<00:08:44.880> usually<00:08:45.200> the<00:08:45.360> sequence 00:08:45.750 --> 00:08:45.760 align:start position:0% Now, a start codon, usually the sequence 00:08:45.760 --> 00:08:48.710 align:start position:0% Now, a start codon, usually the sequence ATG,<00:08:46.640> tells<00:08:46.880> the<00:08:47.120> cell's<00:08:47.440> machinery<00:08:48.160> where<00:08:48.480> to 00:08:48.710 --> 00:08:48.720 align:start position:0% ATG, tells the cell's machinery where to 00:08:48.720 --> 00:08:51.030 align:start position:0% ATG, tells the cell's machinery where to begin<00:08:49.360> building<00:08:49.680> a<00:08:49.920> protein.<00:08:50.399> It<00:08:50.640> basically 00:08:51.030 --> 00:08:51.040 align:start position:0% begin building a protein. It basically 00:08:51.040 --> 00:08:53.750 align:start position:0% begin building a protein. It basically means<00:08:51.440> start<00:08:51.760> here.<00:08:52.320> A<00:08:52.560> stop<00:08:52.800> codon<00:08:53.360> does<00:08:53.600> the 00:08:53.750 --> 00:08:53.760 align:start position:0% means start here. A stop codon does the 00:08:53.760 --> 00:08:55.990 align:start position:0% means start here. A stop codon does the opposite.<00:08:54.160> It<00:08:54.399> tells<00:08:54.560> the<00:08:54.800> cell<00:08:55.120> when<00:08:55.440> to<00:08:55.600> stop 00:08:55.990 --> 00:08:56.000 align:start position:0% opposite. It tells the cell when to stop 00:08:56.000 --> 00:08:58.550 align:start position:0% opposite. It tells the cell when to stop building.<00:08:56.640> So,<00:08:56.800> if<00:08:56.959> you<00:08:57.120> mutate<00:08:57.760> the<00:08:58.160> start 00:08:58.550 --> 00:08:58.560 align:start position:0% building. So, if you mutate the start 00:08:58.560 --> 00:09:00.230 align:start position:0% building. So, if you mutate the start codon,<00:08:59.040> in<00:08:59.279> other<00:08:59.360> words,<00:08:59.600> if<00:08:59.760> you<00:08:59.920> mess<00:09:00.080> up 00:09:00.230 --> 00:09:00.240 align:start position:0% codon, in other words, if you mess up 00:09:00.240 --> 00:09:02.150 align:start position:0% codon, in other words, if you mess up the<00:09:00.399> letters,<00:09:01.040> the<00:09:01.200> protein<00:09:01.600> is<00:09:01.839> never 00:09:02.150 --> 00:09:02.160 align:start position:0% the letters, the protein is never 00:09:02.160 --> 00:09:04.470 align:start position:0% the letters, the protein is never produced.<00:09:02.800> And<00:09:02.959> if<00:09:03.120> you<00:09:03.360> mutate<00:09:03.839> the<00:09:04.080> stop 00:09:04.470 --> 00:09:04.480 align:start position:0% produced. And if you mutate the stop 00:09:04.480 --> 00:09:06.070 align:start position:0% produced. And if you mutate the stop code<00:09:04.640> on,<00:09:04.959> then<00:09:05.120> the<00:09:05.279> cell<00:09:05.600> just<00:09:05.760> keeps 00:09:06.070 --> 00:09:06.080 align:start position:0% code on, then the cell just keeps 00:09:06.080 --> 00:09:08.150 align:start position:0% code on, then the cell just keeps working<00:09:06.399> and<00:09:06.640> working<00:09:07.120> far<00:09:07.440> past<00:09:07.839> the 00:09:08.150 --> 00:09:08.160 align:start position:0% working and working far past the 00:09:08.160 --> 00:09:10.230 align:start position:0% working and working far past the intended<00:09:08.720> endpoint.<00:09:09.360> It<00:09:09.600> just<00:09:09.760> produces 00:09:10.230 --> 00:09:10.240 align:start position:0% intended endpoint. It just produces 00:09:10.240 --> 00:09:12.070 align:start position:0% intended endpoint. It just produces something<00:09:10.480> that<00:09:10.720> often<00:09:11.040> does<00:09:11.279> not<00:09:11.440> work.<00:09:11.920> So 00:09:12.070 --> 00:09:12.080 align:start position:0% something that often does not work. So 00:09:12.080 --> 00:09:14.230 align:start position:0% something that often does not work. So when<00:09:12.399> scientists<00:09:13.040> added<00:09:13.440> mutations<00:09:14.000> to 00:09:14.230 --> 00:09:14.240 align:start position:0% when scientists added mutations to 00:09:14.240 --> 00:09:17.269 align:start position:0% when scientists added mutations to either<00:09:14.640> the<00:09:14.959> start<00:09:15.279> or<00:09:15.600> stop<00:09:15.920> codons,<00:09:16.480> EVO<00:09:16.959> 2 00:09:17.269 --> 00:09:17.279 align:start position:0% either the start or stop codons, EVO 2 00:09:17.279 --> 00:09:19.990 align:start position:0% either the start or stop codons, EVO 2 was<00:09:17.519> able<00:09:17.839> to<00:09:18.080> correctly<00:09:18.720> flag<00:09:19.120> this<00:09:19.279> mutation 00:09:19.990 --> 00:09:20.000 align:start position:0% was able to correctly flag this mutation 00:09:20.000 --> 00:09:22.630 align:start position:0% was able to correctly flag this mutation as<00:09:20.320> highly<00:09:20.800> destructive,<00:09:21.519> which<00:09:21.839> is<00:09:22.000> correct. 00:09:22.630 --> 00:09:22.640 align:start position:0% as highly destructive, which is correct. 00:09:22.640 --> 00:09:24.870 align:start position:0% as highly destructive, which is correct. And<00:09:22.800> it<00:09:23.040> was<00:09:23.279> even<00:09:23.519> able<00:09:23.839> to<00:09:24.000> understand<00:09:24.560> more 00:09:24.870 --> 00:09:24.880 align:start position:0% And it was even able to understand more 00:09:24.880 --> 00:09:27.269 align:start position:0% And it was even able to understand more subtle<00:09:25.279> things.<00:09:25.760> So<00:09:26.000> here's<00:09:26.320> another<00:09:26.720> insane 00:09:27.269 --> 00:09:27.279 align:start position:0% subtle things. So here's another insane 00:09:27.279 --> 00:09:29.269 align:start position:0% subtle things. So here's another insane finding.<00:09:27.920> The<00:09:28.160> researchers<00:09:28.640> found<00:09:28.880> that<00:09:29.040> the 00:09:29.269 --> 00:09:29.279 align:start position:0% finding. The researchers found that the 00:09:29.279 --> 00:09:31.670 align:start position:0% finding. The researchers found that the model<00:09:29.519> could<00:09:29.920> identify<00:09:30.560> things<00:09:30.880> like<00:09:31.360> the 00:09:31.670 --> 00:09:31.680 align:start position:0% model could identify things like the 00:09:31.680 --> 00:09:34.550 align:start position:0% model could identify things like the Shine<00:09:32.240> Delgarno<00:09:33.040> sequence<00:09:33.519> and<00:09:33.760> the<00:09:33.920> KAC 00:09:34.550 --> 00:09:34.560 align:start position:0% Shine Delgarno sequence and the KAC 00:09:34.560 --> 00:09:36.870 align:start position:0% Shine Delgarno sequence and the KAC sequence.<00:09:35.200> Now,<00:09:35.440> what<00:09:35.760> on<00:09:36.000> earth<00:09:36.240> are<00:09:36.399> these? 00:09:36.870 --> 00:09:36.880 align:start position:0% sequence. Now, what on earth are these? 00:09:36.880 --> 00:09:38.790 align:start position:0% sequence. Now, what on earth are these? Think<00:09:37.120> of<00:09:37.279> them<00:09:37.440> as<00:09:37.600> like<00:09:37.920> landing<00:09:38.240> pads<00:09:38.640> for 00:09:38.790 --> 00:09:38.800 align:start position:0% Think of them as like landing pads for 00:09:38.800 --> 00:09:41.269 align:start position:0% Think of them as like landing pads for the<00:09:39.040> cell's<00:09:39.519> protein<00:09:40.080> building<00:09:40.480> machinery. 00:09:41.269 --> 00:09:41.279 align:start position:0% the cell's protein building machinery. 00:09:41.279 --> 00:09:43.910 align:start position:0% the cell's protein building machinery. You<00:09:41.440> see,<00:09:41.760> before<00:09:42.240> a<00:09:42.560> protein<00:09:43.040> can<00:09:43.279> be<00:09:43.440> made, 00:09:43.910 --> 00:09:43.920 align:start position:0% You see, before a protein can be made, 00:09:43.920 --> 00:09:45.990 align:start position:0% You see, before a protein can be made, something<00:09:44.240> called<00:09:44.480> a<00:09:44.720> ribosome<00:09:45.440> has<00:09:45.680> to 00:09:45.990 --> 00:09:46.000 align:start position:0% something called a ribosome has to 00:09:46.000 --> 00:09:47.829 align:start position:0% something called a ribosome has to attach<00:09:46.399> to<00:09:46.560> it<00:09:46.720> at<00:09:46.959> exactly<00:09:47.360> the<00:09:47.600> right 00:09:47.829 --> 00:09:47.839 align:start position:0% attach to it at exactly the right 00:09:47.839 --> 00:09:49.829 align:start position:0% attach to it at exactly the right location.<00:09:48.560> And<00:09:48.640> if<00:09:48.800> it<00:09:49.040> docks<00:09:49.360> in<00:09:49.519> the<00:09:49.680> wrong 00:09:49.829 --> 00:09:49.839 align:start position:0% location. And if it docks in the wrong 00:09:49.839 --> 00:09:51.910 align:start position:0% location. And if it docks in the wrong place,<00:09:50.160> the<00:09:50.399> entire<00:09:50.720> protein<00:09:51.360> might<00:09:51.600> not<00:09:51.760> be 00:09:51.910 --> 00:09:51.920 align:start position:0% place, the entire protein might not be 00:09:51.920 --> 00:09:54.310 align:start position:0% place, the entire protein might not be built<00:09:52.240> correctly<00:09:52.800> or<00:09:53.040> even<00:09:53.279> at<00:09:53.600> all.<00:09:53.920> So<00:09:54.080> the 00:09:54.310 --> 00:09:54.320 align:start position:0% built correctly or even at all. So the 00:09:54.320 --> 00:09:56.710 align:start position:0% built correctly or even at all. So the cell<00:09:54.560> has<00:09:54.800> to<00:09:55.040> also<00:09:55.440> code<00:09:55.680> up<00:09:56.000> this<00:09:56.320> landing 00:09:56.710 --> 00:09:56.720 align:start position:0% cell has to also code up this landing 00:09:56.720 --> 00:09:59.110 align:start position:0% cell has to also code up this landing pad<00:09:57.040> region.<00:09:57.519> And<00:09:57.839> for<00:09:58.160> bacteria,<00:09:58.880> this 00:09:59.110 --> 00:09:59.120 align:start position:0% pad region. And for bacteria, this 00:09:59.120 --> 00:10:01.350 align:start position:0% pad region. And for bacteria, this region<00:09:59.440> is<00:09:59.600> called<00:09:59.839> the<00:10:00.080> shine<00:10:00.560> delgo 00:10:01.350 --> 00:10:01.360 align:start position:0% region is called the shine delgo 00:10:01.360 --> 00:10:03.350 align:start position:0% region is called the shine delgo sequence.<00:10:01.920> And<00:10:02.080> then<00:10:02.240> for<00:10:02.480> more<00:10:02.800> complex 00:10:03.350 --> 00:10:03.360 align:start position:0% sequence. And then for more complex 00:10:03.360 --> 00:10:05.509 align:start position:0% sequence. And then for more complex organisms<00:10:04.000> such<00:10:04.240> as<00:10:04.480> humans<00:10:04.800> and<00:10:05.040> plants, 00:10:05.509 --> 00:10:05.519 align:start position:0% organisms such as humans and plants, 00:10:05.519 --> 00:10:07.190 align:start position:0% organisms such as humans and plants, this<00:10:05.760> sequence<00:10:06.160> is<00:10:06.320> called<00:10:06.480> the<00:10:06.640> KAC 00:10:07.190 --> 00:10:07.200 align:start position:0% this sequence is called the KAC 00:10:07.200 --> 00:10:09.509 align:start position:0% this sequence is called the KAC sequence.<00:10:07.920> These<00:10:08.160> sequences<00:10:08.720> sit<00:10:09.040> basically 00:10:09.509 --> 00:10:09.519 align:start position:0% sequence. These sequences sit basically 00:10:09.519 --> 00:10:11.350 align:start position:0% sequence. These sequences sit basically next<00:10:09.760> to<00:10:09.920> the<00:10:10.080> start<00:10:10.320> of<00:10:10.480> the<00:10:10.640> gene<00:10:11.040> and 00:10:11.350 --> 00:10:11.360 align:start position:0% next to the start of the gene and 00:10:11.360 --> 00:10:13.910 align:start position:0% next to the start of the gene and basically<00:10:12.080> guide<00:10:12.399> the<00:10:12.640> ribosome<00:10:13.440> to<00:10:13.680> the 00:10:13.910 --> 00:10:13.920 align:start position:0% basically guide the ribosome to the 00:10:13.920 --> 00:10:15.750 align:start position:0% basically guide the ribosome to the correct<00:10:14.320> starting<00:10:14.720> point.<00:10:15.279> Without<00:10:15.600> them, 00:10:15.750 --> 00:10:15.760 align:start position:0% correct starting point. Without them, 00:10:15.760 --> 00:10:17.670 align:start position:0% correct starting point. Without them, the<00:10:15.920> ribosome<00:10:16.480> might<00:10:16.720> simply<00:10:17.120> drift<00:10:17.440> past 00:10:17.670 --> 00:10:17.680 align:start position:0% the ribosome might simply drift past 00:10:17.680 --> 00:10:19.990 align:start position:0% the ribosome might simply drift past this<00:10:17.920> and<00:10:18.160> never<00:10:18.560> latch<00:10:18.800> on<00:10:19.040> to<00:10:19.200> it<00:10:19.360> to<00:10:19.600> make<00:10:19.760> a 00:10:19.990 --> 00:10:20.000 align:start position:0% this and never latch on to it to make a 00:10:20.000 --> 00:10:21.829 align:start position:0% this and never latch on to it to make a protein.<00:10:20.480> What's<00:10:20.800> remarkable<00:10:21.279> is<00:10:21.519> that<00:10:21.680> the 00:10:21.829 --> 00:10:21.839 align:start position:0% protein. What's remarkable is that the 00:10:21.839 --> 00:10:24.310 align:start position:0% protein. What's remarkable is that the AI<00:10:22.320> also<00:10:22.640> figured<00:10:22.959> out<00:10:23.200> these<00:10:23.519> landing<00:10:23.839> pads 00:10:24.310 --> 00:10:24.320 align:start position:0% AI also figured out these landing pads 00:10:24.320 --> 00:10:26.870 align:start position:0% AI also figured out these landing pads purely<00:10:24.800> from<00:10:25.200> context.<00:10:25.839> If<00:10:26.079> the<00:10:26.320> scientists 00:10:26.870 --> 00:10:26.880 align:start position:0% purely from context. If the scientists 00:10:26.880 --> 00:10:29.670 align:start position:0% purely from context. If the scientists added<00:10:27.279> mutations<00:10:27.920> in<00:10:28.240> these<00:10:28.480> areas,<00:10:29.040> the<00:10:29.200> AI 00:10:29.670 --> 00:10:29.680 align:start position:0% added mutations in these areas, the AI 00:10:29.680 --> 00:10:32.389 align:start position:0% added mutations in these areas, the AI was<00:10:30.000> able<00:10:30.320> to<00:10:30.640> successfully<00:10:31.440> flag<00:10:31.839> this<00:10:32.160> as 00:10:32.389 --> 00:10:32.399 align:start position:0% was able to successfully flag this as 00:10:32.399 --> 00:10:34.870 align:start position:0% was able to successfully flag this as highly<00:10:32.800> damaging.<00:10:33.680> And<00:10:33.920> there's<00:10:34.240> more.<00:10:34.640> It 00:10:34.870 --> 00:10:34.880 align:start position:0% highly damaging. And there's more. It 00:10:34.880 --> 00:10:37.030 align:start position:0% highly damaging. And there's more. It also<00:10:35.200> figured<00:10:35.440> out<00:10:35.680> the<00:10:36.000> difference<00:10:36.399> between 00:10:37.030 --> 00:10:37.040 align:start position:0% also figured out the difference between 00:10:37.040 --> 00:10:39.990 align:start position:0% also figured out the difference between what<00:10:37.279> is<00:10:37.440> called<00:10:37.680> a<00:10:38.240> synonymous<00:10:38.959> mutation<00:10:39.760> as 00:10:39.990 --> 00:10:40.000 align:start position:0% what is called a synonymous mutation as 00:10:40.000 --> 00:10:42.630 align:start position:0% what is called a synonymous mutation as well<00:10:40.160> as<00:10:40.480> a<00:10:40.720> frame<00:10:41.120> shift<00:10:41.440> mutation.<00:10:42.320> Let's 00:10:42.630 --> 00:10:42.640 align:start position:0% well as a frame shift mutation. Let's 00:10:42.640 --> 00:10:44.710 align:start position:0% well as a frame shift mutation. Let's explain<00:10:42.959> these<00:10:43.279> in<00:10:43.519> simple<00:10:43.839> terms.<00:10:44.320> So,<00:10:44.560> a 00:10:44.710 --> 00:10:44.720 align:start position:0% explain these in simple terms. So, a 00:10:44.720 --> 00:10:46.870 align:start position:0% explain these in simple terms. So, a synonymous<00:10:45.279> mutation<00:10:45.680> is<00:10:45.839> when<00:10:46.079> you<00:10:46.320> change<00:10:46.640> a 00:10:46.870 --> 00:10:46.880 align:start position:0% synonymous mutation is when you change a 00:10:46.880 --> 00:10:48.790 align:start position:0% synonymous mutation is when you change a single<00:10:47.120> letter<00:10:47.279> of<00:10:47.440> DNA.<00:10:48.079> But<00:10:48.160> because<00:10:48.399> of<00:10:48.560> the 00:10:48.790 --> 00:10:48.800 align:start position:0% single letter of DNA. But because of the 00:10:48.800 --> 00:10:50.790 align:start position:0% single letter of DNA. But because of the redundancy<00:10:49.440> of<00:10:49.680> the<00:10:49.839> genetic<00:10:50.320> code,<00:10:50.640> and 00:10:50.790 --> 00:10:50.800 align:start position:0% redundancy of the genetic code, and 00:10:50.800 --> 00:10:52.949 align:start position:0% redundancy of the genetic code, and because<00:10:51.040> it's<00:10:51.279> just<00:10:51.600> one<00:10:51.920> change,<00:10:52.480> it<00:10:52.720> doesn't 00:10:52.949 --> 00:10:52.959 align:start position:0% because it's just one change, it doesn't 00:10:52.959 --> 00:10:55.430 align:start position:0% because it's just one change, it doesn't really<00:10:53.279> mess<00:10:53.680> anything<00:10:54.079> up.<00:10:54.560> So,<00:10:54.720> when<00:10:54.880> Evo<00:10:55.279> 2 00:10:55.430 --> 00:10:55.440 align:start position:0% really mess anything up. So, when Evo 2 00:10:55.440 --> 00:10:57.430 align:start position:0% really mess anything up. So, when Evo 2 was<00:10:55.600> shown<00:10:55.839> these<00:10:56.160> mutations,<00:10:56.720> it<00:10:56.959> recognized 00:10:57.430 --> 00:10:57.440 align:start position:0% was shown these mutations, it recognized 00:10:57.440 --> 00:10:59.509 align:start position:0% was shown these mutations, it recognized them<00:10:57.600> and<00:10:57.839> said,<00:10:58.160> "This<00:10:58.480> is<00:10:58.640> fine.<00:10:59.120> It's<00:10:59.279> a 00:10:59.509 --> 00:10:59.519 align:start position:0% them and said, "This is fine. It's a 00:10:59.519 --> 00:11:01.829 align:start position:0% them and said, "This is fine. It's a high<00:10:59.680> likelihood<00:11:00.240> of<00:11:00.480> survival."<00:11:01.440> But<00:11:01.600> a 00:11:01.829 --> 00:11:01.839 align:start position:0% high likelihood of survival." But a 00:11:01.839 --> 00:11:03.990 align:start position:0% high likelihood of survival." But a frame<00:11:02.160> shift<00:11:02.480> mutation<00:11:03.040> is<00:11:03.279> different.<00:11:03.760> It's 00:11:03.990 --> 00:11:04.000 align:start position:0% frame shift mutation is different. It's 00:11:04.000 --> 00:11:06.949 align:start position:0% frame shift mutation is different. It's catastrophic.<00:11:05.040> You<00:11:05.200> see,<00:11:05.600> DNA<00:11:06.079> is<00:11:06.320> read<00:11:06.640> in 00:11:06.949 --> 00:11:06.959 align:start position:0% catastrophic. You see, DNA is read in 00:11:06.959 --> 00:11:09.269 align:start position:0% catastrophic. You see, DNA is read in chunks<00:11:07.360> of<00:11:07.680> three.<00:11:08.079> So<00:11:08.240> if<00:11:08.399> you<00:11:08.560> insert<00:11:09.040> or 00:11:09.269 --> 00:11:09.279 align:start position:0% chunks of three. So if you insert or 00:11:09.279 --> 00:11:12.069 align:start position:0% chunks of three. So if you insert or delete<00:11:09.920> just<00:11:10.320> one<00:11:10.640> single<00:11:11.040> letter,<00:11:11.680> every 00:11:12.069 --> 00:11:12.079 align:start position:0% delete just one single letter, every 00:11:12.079 --> 00:11:14.470 align:start position:0% delete just one single letter, every single<00:11:12.560> threelet<00:11:13.200> chunk<00:11:13.519> downstream<00:11:14.160> is 00:11:14.470 --> 00:11:14.480 align:start position:0% single threelet chunk downstream is 00:11:14.480 --> 00:11:16.949 align:start position:0% single threelet chunk downstream is shifted<00:11:14.959> by<00:11:15.200> one.<00:11:15.760> So<00:11:15.920> the<00:11:16.160> entire<00:11:16.640> rest<00:11:16.800> of 00:11:16.949 --> 00:11:16.959 align:start position:0% shifted by one. So the entire rest of 00:11:16.959 --> 00:11:18.710 align:start position:0% shifted by one. So the entire rest of the<00:11:17.120> sequence<00:11:17.519> becomes<00:11:17.839> messed<00:11:18.160> up.<00:11:18.480> Well, 00:11:18.710 --> 00:11:18.720 align:start position:0% the sequence becomes messed up. Well, 00:11:18.720 --> 00:11:20.949 align:start position:0% the sequence becomes messed up. Well, when<00:11:18.959> EVO2<00:11:19.600> was<00:11:19.839> shown<00:11:20.240> these<00:11:20.560> types<00:11:20.800> of 00:11:20.949 --> 00:11:20.959 align:start position:0% when EVO2 was shown these types of 00:11:20.959 --> 00:11:23.670 align:start position:0% when EVO2 was shown these types of mutations,<00:11:21.760> it<00:11:21.920> was<00:11:22.160> also<00:11:22.560> able<00:11:22.880> to<00:11:23.120> correctly 00:11:23.670 --> 00:11:23.680 align:start position:0% mutations, it was also able to correctly 00:11:23.680 --> 00:11:26.150 align:start position:0% mutations, it was also able to correctly say<00:11:24.000> that<00:11:24.399> these<00:11:24.880> would<00:11:25.120> be<00:11:25.519> severely 00:11:26.150 --> 00:11:26.160 align:start position:0% say that these would be severely 00:11:26.160 --> 00:11:27.990 align:start position:0% say that these would be severely damaging.<00:11:26.720> So<00:11:26.880> as<00:11:27.040> you<00:11:27.120> can<00:11:27.200> see,<00:11:27.440> EVO<00:11:27.839> 2 00:11:27.990 --> 00:11:28.000 align:start position:0% damaging. So as you can see, EVO 2 00:11:28.000 --> 00:11:29.670 align:start position:0% damaging. So as you can see, EVO 2 didn't<00:11:28.240> just<00:11:28.399> memorize<00:11:28.959> letters.<00:11:29.440> It 00:11:29.670 --> 00:11:29.680 align:start position:0% didn't just memorize letters. It 00:11:29.680 --> 00:11:31.670 align:start position:0% didn't just memorize letters. It actually<00:11:29.920> reverse<00:11:30.399> engineered<00:11:30.880> how<00:11:31.200> DNA 00:11:31.670 --> 00:11:31.680 align:start position:0% actually reverse engineered how DNA 00:11:31.680 --> 00:11:33.509 align:start position:0% actually reverse engineered how DNA works<00:11:32.079> without<00:11:32.480> ever<00:11:32.800> being<00:11:33.040> taught 00:11:33.509 --> 00:11:33.519 align:start position:0% works without ever being taught 00:11:33.519 --> 00:11:36.230 align:start position:0% works without ever being taught anything.<00:11:34.160> And<00:11:34.320> it<00:11:34.480> gets<00:11:34.800> even<00:11:35.200> crazier.<00:11:36.000> The 00:11:36.230 --> 00:11:36.240 align:start position:0% anything. And it gets even crazier. The 00:11:36.240 --> 00:11:38.310 align:start position:0% anything. And it gets even crazier. The researchers<00:11:36.720> did<00:11:36.959> an<00:11:37.200> even<00:11:37.360> more<00:11:37.760> challenging 00:11:38.310 --> 00:11:38.320 align:start position:0% researchers did an even more challenging 00:11:38.320 --> 00:11:40.870 align:start position:0% researchers did an even more challenging test<00:11:38.640> called<00:11:38.959> the<00:11:39.120> ciliate<00:11:39.680> code<00:11:40.079> test.<00:11:40.640> You 00:11:40.870 --> 00:11:40.880 align:start position:0% test called the ciliate code test. You 00:11:40.880 --> 00:11:42.630 align:start position:0% test called the ciliate code test. You see,<00:11:41.120> for<00:11:41.360> most<00:11:41.680> life<00:11:41.920> on<00:11:42.079> Earth,<00:11:42.399> the 00:11:42.630 --> 00:11:42.640 align:start position:0% see, for most life on Earth, the 00:11:42.640 --> 00:11:45.269 align:start position:0% see, for most life on Earth, the sequence<00:11:43.120> TGA<00:11:43.760> is<00:11:44.000> a<00:11:44.160> stop<00:11:44.560> codon,<00:11:45.120> a 00:11:45.269 --> 00:11:45.279 align:start position:0% sequence TGA is a stop codon, a 00:11:45.279 --> 00:11:47.990 align:start position:0% sequence TGA is a stop codon, a universal<00:11:46.079> stop<00:11:46.399> sign<00:11:46.720> that<00:11:47.040> tells<00:11:47.519> the<00:11:47.760> cell 00:11:47.990 --> 00:11:48.000 align:start position:0% universal stop sign that tells the cell 00:11:48.000 --> 00:11:50.790 align:start position:0% universal stop sign that tells the cell to<00:11:48.240> stop<00:11:48.560> building<00:11:48.880> a<00:11:49.120> protein.<00:11:49.760> But<00:11:50.000> biology 00:11:50.790 --> 00:11:50.800 align:start position:0% to stop building a protein. But biology 00:11:50.800 --> 00:11:52.949 align:start position:0% to stop building a protein. But biology loves<00:11:51.279> exceptions.<00:11:52.079> And<00:11:52.320> there's<00:11:52.480> a<00:11:52.720> group<00:11:52.800> of 00:11:52.949 --> 00:11:52.959 align:start position:0% loves exceptions. And there's a group of 00:11:52.959 --> 00:11:55.509 align:start position:0% loves exceptions. And there's a group of organisms<00:11:53.519> called<00:11:53.920> ciliates<00:11:54.560> that<00:11:54.800> evolved<00:11:55.200> a 00:11:55.509 --> 00:11:55.519 align:start position:0% organisms called ciliates that evolved a 00:11:55.519 --> 00:11:57.509 align:start position:0% organisms called ciliates that evolved a slightly<00:11:55.920> different<00:11:56.240> genetic<00:11:56.720> pattern.<00:11:57.279> In 00:11:57.509 --> 00:11:57.519 align:start position:0% slightly different genetic pattern. In 00:11:57.519 --> 00:12:00.630 align:start position:0% slightly different genetic pattern. In their<00:11:57.760> DNA,<00:11:58.320> TGA<00:11:59.120> doesn't<00:11:59.519> mean<00:11:59.920> stop.<00:12:00.480> It 00:12:00.630 --> 00:12:00.640 align:start position:0% their DNA, TGA doesn't mean stop. It 00:12:00.640 --> 00:12:02.949 align:start position:0% their DNA, TGA doesn't mean stop. It actually<00:12:00.880> means<00:12:01.200> keep<00:12:01.519> going<00:12:02.079> just<00:12:02.320> for<00:12:02.480> them. 00:12:02.949 --> 00:12:02.959 align:start position:0% actually means keep going just for them. 00:12:02.959 --> 00:12:04.870 align:start position:0% actually means keep going just for them. previous<00:12:03.360> AI<00:12:03.760> models<00:12:04.079> that<00:12:04.320> were<00:12:04.480> trained<00:12:04.720> on 00:12:04.870 --> 00:12:04.880 align:start position:0% previous AI models that were trained on 00:12:04.880 --> 00:12:07.590 align:start position:0% previous AI models that were trained on DNA<00:12:05.519> would<00:12:05.839> completely<00:12:06.399> fail<00:12:06.720> this<00:12:06.959> test.<00:12:07.440> If 00:12:07.590 --> 00:12:07.600 align:start position:0% DNA would completely fail this test. If 00:12:07.600 --> 00:12:09.829 align:start position:0% DNA would completely fail this test. If they<00:12:07.760> were<00:12:07.920> given<00:12:08.399> ciliate<00:12:08.959> DNA<00:12:09.440> and<00:12:09.600> asked 00:12:09.829 --> 00:12:09.839 align:start position:0% they were given ciliate DNA and asked 00:12:09.839 --> 00:12:12.550 align:start position:0% they were given ciliate DNA and asked about<00:12:10.160> TGA,<00:12:10.880> the<00:12:11.040> AI<00:12:11.440> models<00:12:11.839> would<00:12:12.160> all<00:12:12.399> say 00:12:12.550 --> 00:12:12.560 align:start position:0% about TGA, the AI models would all say 00:12:12.560 --> 00:12:14.629 align:start position:0% about TGA, the AI models would all say that<00:12:12.880> this<00:12:13.040> is<00:12:13.200> a<00:12:13.360> stop<00:12:13.600> sign,<00:12:14.000> which<00:12:14.240> is<00:12:14.399> not 00:12:14.629 --> 00:12:14.639 align:start position:0% that this is a stop sign, which is not 00:12:14.639 --> 00:12:16.550 align:start position:0% that this is a stop sign, which is not correct.<00:12:15.279> But<00:12:15.440> the<00:12:15.680> remarkable<00:12:16.160> thing<00:12:16.320> about 00:12:16.550 --> 00:12:16.560 align:start position:0% correct. But the remarkable thing about 00:12:16.560 --> 00:12:18.949 align:start position:0% correct. But the remarkable thing about EVO<00:12:17.040> 2<00:12:17.279> is<00:12:17.519> that<00:12:17.760> because<00:12:18.000> it<00:12:18.240> has<00:12:18.399> a<00:12:18.639> million 00:12:18.949 --> 00:12:18.959 align:start position:0% EVO 2 is that because it has a million 00:12:18.959 --> 00:12:21.030 align:start position:0% EVO 2 is that because it has a million token<00:12:19.360> context<00:12:19.839> window<00:12:20.399> because<00:12:20.639> it<00:12:20.880> can 00:12:21.030 --> 00:12:21.040 align:start position:0% token context window because it can 00:12:21.040 --> 00:12:23.910 align:start position:0% token context window because it can understand<00:12:21.440> DNA<00:12:21.920> across<00:12:22.399> all<00:12:22.880> of<00:12:23.120> life,<00:12:23.680> it's 00:12:23.910 --> 00:12:23.920 align:start position:0% understand DNA across all of life, it's 00:12:23.920 --> 00:12:26.710 align:start position:0% understand DNA across all of life, it's able<00:12:24.160> to<00:12:24.399> inspect<00:12:24.880> this<00:12:25.279> entire<00:12:25.839> ciliate<00:12:26.320> DNA 00:12:26.710 --> 00:12:26.720 align:start position:0% able to inspect this entire ciliate DNA 00:12:26.720 --> 00:12:29.190 align:start position:0% able to inspect this entire ciliate DNA first.<00:12:27.279> And<00:12:27.600> given<00:12:27.920> its<00:12:28.240> context,<00:12:28.800> it's<00:12:29.040> able 00:12:29.190 --> 00:12:29.200 align:start position:0% first. And given its context, it's able 00:12:29.200 --> 00:12:31.910 align:start position:0% first. And given its context, it's able to<00:12:29.360> determine<00:12:29.760> that<00:12:30.079> TGA<00:12:30.880> inside<00:12:31.279> this<00:12:31.600> was 00:12:31.910 --> 00:12:31.920 align:start position:0% to determine that TGA inside this was 00:12:31.920 --> 00:12:33.910 align:start position:0% to determine that TGA inside this was not<00:12:32.160> a<00:12:32.399> stop<00:12:32.639> codon.<00:12:33.279> Note<00:12:33.519> that<00:12:33.680> the 00:12:33.910 --> 00:12:33.920 align:start position:0% not a stop codon. Note that the 00:12:33.920 --> 00:12:36.230 align:start position:0% not a stop codon. Note that the researchers<00:12:34.399> never<00:12:34.639> told<00:12:34.880> it<00:12:35.120> that<00:12:35.360> this<00:12:35.680> DNA 00:12:36.230 --> 00:12:36.240 align:start position:0% researchers never told it that this DNA 00:12:36.240 --> 00:12:39.110 align:start position:0% researchers never told it that this DNA belongs<00:12:36.639> to<00:12:37.120> this<00:12:37.519> organism<00:12:38.160> called<00:12:38.399> a<00:12:38.560> cyot. 00:12:39.110 --> 00:12:39.120 align:start position:0% belongs to this organism called a cyot. 00:12:39.120 --> 00:12:41.750 align:start position:0% belongs to this organism called a cyot. The<00:12:39.360> AI<00:12:39.920> also<00:12:40.480> never<00:12:40.880> learned<00:12:41.279> that<00:12:41.519> this 00:12:41.750 --> 00:12:41.760 align:start position:0% The AI also never learned that this 00:12:41.760 --> 00:12:43.829 align:start position:0% The AI also never learned that this belongs<00:12:42.079> to<00:12:42.240> a<00:12:42.399> ciliate<00:12:42.800> or<00:12:43.040> what<00:12:43.200> a<00:12:43.360> ciliate 00:12:43.829 --> 00:12:43.839 align:start position:0% belongs to a ciliate or what a ciliate 00:12:43.839 --> 00:12:46.629 align:start position:0% belongs to a ciliate or what a ciliate even<00:12:44.079> is.<00:12:44.720> Yet,<00:12:44.959> it<00:12:45.200> inferred<00:12:45.680> this<00:12:45.920> from<00:12:46.320> its 00:12:46.629 --> 00:12:46.639 align:start position:0% even is. Yet, it inferred this from its 00:12:46.639 --> 00:12:49.110 align:start position:0% even is. Yet, it inferred this from its understanding<00:12:47.279> of<00:12:47.519> the<00:12:47.760> grammar<00:12:48.160> of<00:12:48.399> DNA. 00:12:49.110 --> 00:12:49.120 align:start position:0% understanding of the grammar of DNA. 00:12:49.120 --> 00:12:51.509 align:start position:0% understanding of the grammar of DNA. Now,<00:12:49.440> so<00:12:49.680> far<00:12:50.079> we've<00:12:50.399> talked<00:12:50.639> about<00:12:50.880> tests<00:12:51.279> on 00:12:51.509 --> 00:12:51.519 align:start position:0% Now, so far we've talked about tests on 00:12:51.519 --> 00:12:54.069 align:start position:0% Now, so far we've talked about tests on bacteria<00:12:52.160> and<00:12:52.399> other<00:12:52.720> microorganisms.<00:12:53.920> But 00:12:54.069 --> 00:12:54.079 align:start position:0% bacteria and other microorganisms. But 00:12:54.079 --> 00:12:56.230 align:start position:0% bacteria and other microorganisms. But if<00:12:54.320> this<00:12:54.560> model<00:12:54.959> truly<00:12:55.440> understands<00:12:56.000> the 00:12:56.230 --> 00:12:56.240 align:start position:0% if this model truly understands the 00:12:56.240 --> 00:12:58.629 align:start position:0% if this model truly understands the language<00:12:56.639> of<00:12:56.880> life,<00:12:57.360> then<00:12:57.600> in<00:12:57.920> theory,<00:12:58.399> it 00:12:58.629 --> 00:12:58.639 align:start position:0% language of life, then in theory, it 00:12:58.639 --> 00:13:01.030 align:start position:0% language of life, then in theory, it should<00:12:58.880> also<00:12:59.200> work<00:12:59.440> with<00:12:59.760> human<00:13:00.079> DNA<00:13:00.560> as<00:13:00.800> well, 00:13:01.030 --> 00:13:01.040 align:start position:0% should also work with human DNA as well, 00:13:01.040 --> 00:13:02.949 align:start position:0% should also work with human DNA as well, right?<00:13:01.519> And<00:13:01.680> that's<00:13:02.000> exactly<00:13:02.480> what<00:13:02.800> the 00:13:02.949 --> 00:13:02.959 align:start position:0% right? And that's exactly what the 00:13:02.959 --> 00:13:05.110 align:start position:0% right? And that's exactly what the researchers<00:13:03.519> tested<00:13:04.000> next.<00:13:04.720> I've<00:13:04.959> been 00:13:05.110 --> 00:13:05.120 align:start position:0% researchers tested next. I've been 00:13:05.120 --> 00:13:06.870 align:start position:0% researchers tested next. I've been covering<00:13:05.440> the<00:13:05.680> best<00:13:06.000> image<00:13:06.320> and<00:13:06.560> video 00:13:06.870 --> 00:13:06.880 align:start position:0% covering the best image and video 00:13:06.880 --> 00:13:08.629 align:start position:0% covering the best image and video generators,<00:13:07.519> and<00:13:07.680> it<00:13:07.920> can<00:13:08.079> get<00:13:08.240> very 00:13:08.629 --> 00:13:08.639 align:start position:0% generators, and it can get very 00:13:08.639 --> 00:13:10.550 align:start position:0% generators, and it can get very overwhelming.<00:13:09.279> Fortunately,<00:13:09.839> Higsfield, 00:13:10.550 --> 00:13:10.560 align:start position:0% overwhelming. Fortunately, Higsfield, 00:13:10.560 --> 00:13:12.150 align:start position:0% overwhelming. Fortunately, Higsfield, the<00:13:10.720> sponsor<00:13:11.040> of<00:13:11.200> this<00:13:11.440> video,<00:13:11.839> brings 00:13:12.150 --> 00:13:12.160 align:start position:0% the sponsor of this video, brings 00:13:12.160 --> 00:13:14.629 align:start position:0% the sponsor of this video, brings everything<00:13:12.560> together<00:13:13.120> all<00:13:13.360> in<00:13:13.680> one<00:13:13.920> place. 00:13:14.629 --> 00:13:14.639 align:start position:0% everything together all in one place. 00:13:14.639 --> 00:13:17.269 align:start position:0% everything together all in one place. From<00:13:15.120> leading<00:13:15.519> image<00:13:16.000> and<00:13:16.320> video<00:13:16.720> generation 00:13:17.269 --> 00:13:17.279 align:start position:0% From leading image and video generation 00:13:17.279 --> 00:13:20.150 align:start position:0% From leading image and video generation models<00:13:17.760> to<00:13:18.079> AI<00:13:18.560> cinematic<00:13:19.120> workflows.<00:13:19.920> The 00:13:20.150 --> 00:13:20.160 align:start position:0% models to AI cinematic workflows. The 00:13:20.160 --> 00:13:22.550 align:start position:0% models to AI cinematic workflows. The idea<00:13:20.480> here<00:13:20.800> is<00:13:21.120> to<00:13:21.680> bring<00:13:22.000> more<00:13:22.240> of<00:13:22.399> a 00:13:22.550 --> 00:13:22.560 align:start position:0% idea here is to bring more of a 00:13:22.560 --> 00:13:24.629 align:start position:0% idea here is to bring more of a filmmaking<00:13:23.120> workflow<00:13:23.680> to<00:13:23.920> AI<00:13:24.320> video 00:13:24.629 --> 00:13:24.639 align:start position:0% filmmaking workflow to AI video 00:13:24.639 --> 00:13:26.550 align:start position:0% filmmaking workflow to AI video creation.<00:13:25.200> Instead<00:13:25.519> of<00:13:25.680> you<00:13:25.920> just<00:13:26.079> typing<00:13:26.399> a 00:13:26.550 --> 00:13:26.560 align:start position:0% creation. Instead of you just typing a 00:13:26.560 --> 00:13:28.389 align:start position:0% creation. Instead of you just typing a prompt<00:13:26.800> and<00:13:27.120> hoping<00:13:27.360> for<00:13:27.519> the<00:13:27.760> best,<00:13:28.160> you 00:13:28.389 --> 00:13:28.399 align:start position:0% prompt and hoping for the best, you 00:13:28.399 --> 00:13:30.389 align:start position:0% prompt and hoping for the best, you actually<00:13:28.800> follow<00:13:29.120> a<00:13:29.360> creative<00:13:29.760> pipeline. 00:13:30.389 --> 00:13:30.399 align:start position:0% actually follow a creative pipeline. 00:13:30.399 --> 00:13:32.550 align:start position:0% actually follow a creative pipeline. Cast<00:13:30.720> your<00:13:30.959> actor,<00:13:31.600> build<00:13:31.839> the<00:13:32.000> scene,<00:13:32.320> and 00:13:32.550 --> 00:13:32.560 align:start position:0% Cast your actor, build the scene, and 00:13:32.560 --> 00:13:34.710 align:start position:0% Cast your actor, build the scene, and direct<00:13:32.880> the<00:13:33.040> shot.<00:13:33.519> It<00:13:33.760> starts<00:13:34.000> with<00:13:34.399> sole 00:13:34.710 --> 00:13:34.720 align:start position:0% direct the shot. It starts with sole 00:13:34.720 --> 00:13:37.110 align:start position:0% direct the shot. It starts with sole cast,<00:13:35.120> which<00:13:35.360> is<00:13:35.519> Higsfield's<00:13:36.240> AI<00:13:36.639> actor 00:13:37.110 --> 00:13:37.120 align:start position:0% cast, which is Higsfield's AI actor 00:13:37.120 --> 00:13:39.030 align:start position:0% cast, which is Higsfield's AI actor creation<00:13:37.680> system.<00:13:38.160> Instead<00:13:38.480> of<00:13:38.560> prompting 00:13:39.030 --> 00:13:39.040 align:start position:0% creation system. Instead of prompting 00:13:39.040 --> 00:13:41.590 align:start position:0% creation system. Instead of prompting random<00:13:39.440> faces<00:13:39.839> every<00:13:40.160> time,<00:13:40.560> you<00:13:40.800> can<00:13:41.040> design 00:13:41.590 --> 00:13:41.600 align:start position:0% random faces every time, you can design 00:13:41.600 --> 00:13:43.750 align:start position:0% random faces every time, you can design fully<00:13:42.000> customizable<00:13:42.720> characters<00:13:43.120> that<00:13:43.440> stay 00:13:43.750 --> 00:13:43.760 align:start position:0% fully customizable characters that stay 00:13:43.760 --> 00:13:46.870 align:start position:0% fully customizable characters that stay visually<00:13:44.399> consistent<00:13:45.120> across<00:13:45.839> your<00:13:46.160> scenes. 00:13:46.870 --> 00:13:46.880 align:start position:0% visually consistent across your scenes. 00:13:46.880 --> 00:13:49.110 align:start position:0% visually consistent across your scenes. You<00:13:47.040> can<00:13:47.200> define<00:13:47.519> their<00:13:47.760> genre,<00:13:48.320> age, 00:13:49.110 --> 00:13:49.120 align:start position:0% You can define their genre, age, 00:13:49.120 --> 00:13:51.030 align:start position:0% You can define their genre, age, identity,<00:13:49.760> physical<00:13:50.160> features,<00:13:50.639> clothing 00:13:51.030 --> 00:13:51.040 align:start position:0% identity, physical features, clothing 00:13:51.040 --> 00:13:54.310 align:start position:0% identity, physical features, clothing style,<00:13:51.519> and<00:13:51.760> more.<00:13:52.320> Next<00:13:52.720> comes<00:13:53.279> Soul<00:13:53.760> Cinema, 00:13:54.310 --> 00:13:54.320 align:start position:0% style, and more. Next comes Soul Cinema, 00:13:54.320 --> 00:13:56.310 align:start position:0% style, and more. Next comes Soul Cinema, which<00:13:54.560> generates<00:13:55.040> cinematic<00:13:55.600> grade<00:13:55.920> images 00:13:56.310 --> 00:13:56.320 align:start position:0% which generates cinematic grade images 00:13:56.320 --> 00:13:58.629 align:start position:0% which generates cinematic grade images to<00:13:56.560> build<00:13:56.880> your<00:13:57.120> scenes.<00:13:57.680> The<00:13:57.920> model<00:13:58.160> focuses 00:13:58.629 --> 00:13:58.639 align:start position:0% to build your scenes. The model focuses 00:13:58.639 --> 00:14:00.629 align:start position:0% to build your scenes. The model focuses heavily<00:13:58.959> on<00:13:59.279> lighting,<00:13:59.760> composition,<00:14:00.399> and 00:14:00.629 --> 00:14:00.639 align:start position:0% heavily on lighting, composition, and 00:14:00.639 --> 00:14:03.509 align:start position:0% heavily on lighting, composition, and texture,<00:14:01.279> so<00:14:01.680> the<00:14:01.920> outputs<00:14:02.480> look<00:14:02.720> like<00:14:03.040> films 00:14:03.509 --> 00:14:03.519 align:start position:0% texture, so the outputs look like films 00:14:03.519 --> 00:14:05.910 align:start position:0% texture, so the outputs look like films instead<00:14:03.839> of<00:14:04.000> AI<00:14:04.399> slop.<00:14:04.959> Once<00:14:05.279> your<00:14:05.519> characters 00:14:05.910 --> 00:14:05.920 align:start position:0% instead of AI slop. Once your characters 00:14:05.920 --> 00:14:07.750 align:start position:0% instead of AI slop. Once your characters and<00:14:06.240> scenes<00:14:06.639> are<00:14:06.800> ready,<00:14:07.199> you<00:14:07.440> can<00:14:07.519> move 00:14:07.750 --> 00:14:07.760 align:start position:0% and scenes are ready, you can move 00:14:07.760 --> 00:14:10.710 align:start position:0% and scenes are ready, you can move everything<00:14:08.240> into<00:14:08.480> Sue<00:14:09.040> Cinema<00:14:09.600> Studio<00:14:10.160> 2, 00:14:10.710 --> 00:14:10.720 align:start position:0% everything into Sue Cinema Studio 2, 00:14:10.720 --> 00:14:12.550 align:start position:0% everything into Sue Cinema Studio 2, which<00:14:10.959> is<00:14:11.120> where<00:14:11.360> the<00:14:11.680> actual<00:14:12.160> video 00:14:12.550 --> 00:14:12.560 align:start position:0% which is where the actual video 00:14:12.560 --> 00:14:14.949 align:start position:0% which is where the actual video direction<00:14:13.199> happens.<00:14:13.920> This<00:14:14.160> works<00:14:14.480> like<00:14:14.720> a 00:14:14.949 --> 00:14:14.959 align:start position:0% direction happens. This works like a 00:14:14.959 --> 00:14:16.949 align:start position:0% direction happens. This works like a miniature<00:14:15.519> film<00:14:15.920> studio,<00:14:16.320> where<00:14:16.639> you<00:14:16.800> can 00:14:16.949 --> 00:14:16.959 align:start position:0% miniature film studio, where you can 00:14:16.959 --> 00:14:19.590 align:start position:0% miniature film studio, where you can create<00:14:17.279> multi-shot<00:14:18.079> sequences,<00:14:18.800> control<00:14:19.279> the 00:14:19.590 --> 00:14:19.600 align:start position:0% create multi-shot sequences, control the 00:14:19.600 --> 00:14:21.829 align:start position:0% create multi-shot sequences, control the speed<00:14:19.920> and<00:14:20.160> the<00:14:20.399> camera<00:14:20.720> motion,<00:14:21.279> or<00:14:21.519> you<00:14:21.760> can 00:14:21.829 --> 00:14:21.839 align:start position:0% speed and the camera motion, or you can 00:14:21.839 --> 00:14:23.990 align:start position:0% speed and the camera motion, or you can even<00:14:22.079> define<00:14:22.480> start<00:14:22.720> and<00:14:23.040> end<00:14:23.279> frames,<00:14:23.760> and 00:14:23.990 --> 00:14:24.000 align:start position:0% even define start and end frames, and 00:14:24.000 --> 00:14:25.750 align:start position:0% even define start and end frames, and the<00:14:24.160> system<00:14:24.480> generates<00:14:24.959> the<00:14:25.199> intermediate 00:14:25.750 --> 00:14:25.760 align:start position:0% the system generates the intermediate 00:14:25.760 --> 00:14:27.430 align:start position:0% the system generates the intermediate motion<00:14:26.160> automatically,<00:14:27.040> essentially 00:14:27.430 --> 00:14:27.440 align:start position:0% motion automatically, essentially 00:14:27.440 --> 00:14:29.189 align:start position:0% motion automatically, essentially filling<00:14:27.760> in<00:14:27.920> the<00:14:28.160> animation.<00:14:28.639> All<00:14:28.880> of<00:14:28.959> this 00:14:29.189 --> 00:14:29.199 align:start position:0% filling in the animation. All of this 00:14:29.199 --> 00:14:30.870 align:start position:0% filling in the animation. All of this makes<00:14:29.360> Higsfield<00:14:30.079> one<00:14:30.240> of<00:14:30.320> the<00:14:30.480> first 00:14:30.870 --> 00:14:30.880 align:start position:0% makes Higsfield one of the first 00:14:30.880 --> 00:14:33.350 align:start position:0% makes Higsfield one of the first platforms<00:14:31.440> to<00:14:31.680> truly<00:14:32.079> connect<00:14:32.560> characters, 00:14:33.350 --> 00:14:33.360 align:start position:0% platforms to truly connect characters, 00:14:33.360 --> 00:14:35.430 align:start position:0% platforms to truly connect characters, cinematic<00:14:33.920> image<00:14:34.240> generation,<00:14:34.800> and<00:14:35.120> video 00:14:35.430 --> 00:14:35.440 align:start position:0% cinematic image generation, and video 00:14:35.440 --> 00:14:37.590 align:start position:0% cinematic image generation, and video direction<00:14:36.000> in<00:14:36.240> one<00:14:36.560> seamless<00:14:37.040> workflow. 00:14:37.590 --> 00:14:37.600 align:start position:0% direction in one seamless workflow. 00:14:37.600 --> 00:14:39.590 align:start position:0% direction in one seamless workflow. Whether<00:14:38.000> you're<00:14:38.480> making<00:14:38.800> short<00:14:39.040> films, 00:14:39.590 --> 00:14:39.600 align:start position:0% Whether you're making short films, 00:14:39.600 --> 00:14:41.910 align:start position:0% Whether you're making short films, cinematic<00:14:40.160> ads,<00:14:40.639> music<00:14:41.040> videos,<00:14:41.440> or<00:14:41.680> any 00:14:41.910 --> 00:14:41.920 align:start position:0% cinematic ads, music videos, or any 00:14:41.920 --> 00:14:44.310 align:start position:0% cinematic ads, music videos, or any other<00:14:42.160> content,<00:14:42.720> the<00:14:43.040> entire<00:14:43.519> process<00:14:44.000> can 00:14:44.310 --> 00:14:44.320 align:start position:0% other content, the entire process can 00:14:44.320 --> 00:14:46.790 align:start position:0% other content, the entire process can happen<00:14:44.720> inside<00:14:45.440> one<00:14:45.760> tool.<00:14:46.320> If<00:14:46.560> you<00:14:46.639> want<00:14:46.720> to 00:14:46.790 --> 00:14:46.800 align:start position:0% happen inside one tool. If you want to 00:14:46.800 --> 00:14:48.710 align:start position:0% happen inside one tool. If you want to try<00:14:46.959> it<00:14:47.120> yourself,<00:14:47.600> check<00:14:47.760> out<00:14:47.920> Higsfield 00:14:48.710 --> 00:14:48.720 align:start position:0% try it yourself, check out Higsfield 00:14:48.720 --> 00:14:51.189 align:start position:0% try it yourself, check out Higsfield using<00:14:48.959> the<00:14:49.199> link<00:14:49.440> in<00:14:49.680> the<00:14:49.839> description<00:14:50.160> below. 00:14:51.189 --> 00:14:51.199 align:start position:0% using the link in the description below. 00:14:51.199 --> 00:14:53.030 align:start position:0% using the link in the description below. This<00:14:51.440> brings<00:14:51.680> us<00:14:51.839> to<00:14:52.000> the<00:14:52.240> third<00:14:52.639> major 00:14:53.030 --> 00:14:53.040 align:start position:0% This brings us to the third major 00:14:53.040 --> 00:14:55.110 align:start position:0% This brings us to the third major section<00:14:53.360> of<00:14:53.600> the<00:14:53.760> paper<00:14:54.160> called<00:14:54.639> human 00:14:55.110 --> 00:14:55.120 align:start position:0% section of the paper called human 00:14:55.120 --> 00:14:57.350 align:start position:0% section of the paper called human variant<00:14:55.839> effect<00:14:56.320> prediction.<00:14:56.959> And<00:14:57.199> this 00:14:57.350 --> 00:14:57.360 align:start position:0% variant effect prediction. And this 00:14:57.360 --> 00:14:59.350 align:start position:0% variant effect prediction. And this could<00:14:57.600> directly<00:14:58.079> affect<00:14:58.720> real<00:14:59.040> human 00:14:59.350 --> 00:14:59.360 align:start position:0% could directly affect real human 00:14:59.360 --> 00:15:01.189 align:start position:0% could directly affect real human patients.<00:15:00.000> Here's<00:15:00.240> what<00:15:00.399> they<00:15:00.560> did.<00:15:00.959> The 00:15:01.189 --> 00:15:01.199 align:start position:0% patients. Here's what they did. The 00:15:01.199 --> 00:15:03.350 align:start position:0% patients. Here's what they did. The researchers<00:15:01.600> took<00:15:01.920> a<00:15:02.160> huge<00:15:02.480> genetics<00:15:03.040> data 00:15:03.350 --> 00:15:03.360 align:start position:0% researchers took a huge genetics data 00:15:03.360 --> 00:15:06.230 align:start position:0% researchers took a huge genetics data set<00:15:03.680> called<00:15:04.240> Clinfair,<00:15:05.360> which<00:15:05.680> is<00:15:05.839> basically 00:15:06.230 --> 00:15:06.240 align:start position:0% set called Clinfair, which is basically 00:15:06.240 --> 00:15:09.110 align:start position:0% set called Clinfair, which is basically a<00:15:06.560> database<00:15:07.120> where<00:15:07.600> doctors<00:15:08.160> record<00:15:08.639> human 00:15:09.110 --> 00:15:09.120 align:start position:0% a database where doctors record human 00:15:09.120 --> 00:15:11.829 align:start position:0% a database where doctors record human DNA<00:15:09.680> variations<00:15:10.320> and<00:15:10.720> classify<00:15:11.199> them.<00:15:11.760> For 00:15:11.829 --> 00:15:11.839 align:start position:0% DNA variations and classify them. For 00:15:11.839 --> 00:15:14.389 align:start position:0% DNA variations and classify them. For example,<00:15:12.560> some<00:15:12.880> DNA<00:15:13.279> variants<00:15:13.680> are<00:15:13.920> labeled 00:15:14.389 --> 00:15:14.399 align:start position:0% example, some DNA variants are labeled 00:15:14.399 --> 00:15:16.470 align:start position:0% example, some DNA variants are labeled pathogenic,<00:15:15.279> meaning<00:15:15.519> they<00:15:15.760> cause<00:15:16.000> disease, 00:15:16.470 --> 00:15:16.480 align:start position:0% pathogenic, meaning they cause disease, 00:15:16.480 --> 00:15:18.550 align:start position:0% pathogenic, meaning they cause disease, and<00:15:16.720> others<00:15:17.040> are<00:15:17.279> labeled<00:15:17.760> benign,<00:15:18.320> meaning 00:15:18.550 --> 00:15:18.560 align:start position:0% and others are labeled benign, meaning 00:15:18.560 --> 00:15:20.629 align:start position:0% and others are labeled benign, meaning they<00:15:18.800> have<00:15:19.040> no<00:15:19.279> measurable<00:15:19.920> impact<00:15:20.320> on 00:15:20.629 --> 00:15:20.639 align:start position:0% they have no measurable impact on 00:15:20.639 --> 00:15:22.310 align:start position:0% they have no measurable impact on health.<00:15:21.120> Now,<00:15:21.519> the<00:15:21.760> researchers 00:15:22.310 --> 00:15:22.320 align:start position:0% health. Now, the researchers 00:15:22.320 --> 00:15:25.110 align:start position:0% health. Now, the researchers specifically<00:15:22.880> targeted<00:15:23.360> the<00:15:23.519> genes<00:15:24.160> BRCA 00:15:25.110 --> 00:15:25.120 align:start position:0% specifically targeted the genes BRCA 00:15:25.120 --> 00:15:27.269 align:start position:0% specifically targeted the genes BRCA from<00:15:25.360> this<00:15:25.600> data<00:15:25.920> set.<00:15:26.240> And<00:15:26.399> these<00:15:26.639> are<00:15:26.880> very 00:15:27.269 --> 00:15:27.279 align:start position:0% from this data set. And these are very 00:15:27.279 --> 00:15:28.870 align:start position:0% from this data set. And these are very important<00:15:27.760> genes,<00:15:28.160> which<00:15:28.399> you<00:15:28.639> might<00:15:28.800> have 00:15:28.870 --> 00:15:28.880 align:start position:0% important genes, which you might have 00:15:28.880 --> 00:15:30.550 align:start position:0% important genes, which you might have heard<00:15:29.120> of<00:15:29.279> before<00:15:29.760> because<00:15:30.160> certain 00:15:30.550 --> 00:15:30.560 align:start position:0% heard of before because certain 00:15:30.560 --> 00:15:33.110 align:start position:0% heard of before because certain mutations<00:15:31.360> within<00:15:31.920> these<00:15:32.320> genes<00:15:32.800> can 00:15:33.110 --> 00:15:33.120 align:start position:0% mutations within these genes can 00:15:33.120 --> 00:15:35.269 align:start position:0% mutations within these genes can drastically<00:15:33.920> increase<00:15:34.160> a<00:15:34.399> person's<00:15:34.800> risk<00:15:35.040> of 00:15:35.269 --> 00:15:35.279 align:start position:0% drastically increase a person's risk of 00:15:35.279 --> 00:15:37.509 align:start position:0% drastically increase a person's risk of breast<00:15:35.680> and<00:15:35.920> ovarian<00:15:36.480> cancer.<00:15:37.199> Imagine 00:15:37.509 --> 00:15:37.519 align:start position:0% breast and ovarian cancer. Imagine 00:15:37.519 --> 00:15:39.590 align:start position:0% breast and ovarian cancer. Imagine you're<00:15:37.680> a<00:15:37.839> doctor<00:15:38.160> and<00:15:38.480> you<00:15:38.720> have<00:15:38.959> the<00:15:39.199> DNA 00:15:39.590 --> 00:15:39.600 align:start position:0% you're a doctor and you have the DNA 00:15:39.600 --> 00:15:41.269 align:start position:0% you're a doctor and you have the DNA report<00:15:40.000> of<00:15:40.240> your<00:15:40.399> patient<00:15:40.880> and<00:15:41.120> you're 00:15:41.269 --> 00:15:41.279 align:start position:0% report of your patient and you're 00:15:41.279 --> 00:15:43.910 align:start position:0% report of your patient and you're looking<00:15:41.519> through<00:15:41.839> the<00:15:42.160> BRCA<00:15:43.040> genes<00:15:43.360> to<00:15:43.519> see<00:15:43.680> if 00:15:43.910 --> 00:15:43.920 align:start position:0% looking through the BRCA genes to see if 00:15:43.920 --> 00:15:46.310 align:start position:0% looking through the BRCA genes to see if there<00:15:44.160> are<00:15:44.320> any<00:15:44.720> mutations.<00:15:45.440> A<00:15:45.760> small<00:15:46.000> subtle 00:15:46.310 --> 00:15:46.320 align:start position:0% there are any mutations. A small subtle 00:15:46.320 --> 00:15:48.550 align:start position:0% there are any mutations. A small subtle change<00:15:46.720> could<00:15:47.040> be<00:15:47.279> completely<00:15:47.680> harmless<00:15:48.240> or 00:15:48.550 --> 00:15:48.560 align:start position:0% change could be completely harmless or 00:15:48.560 --> 00:15:51.110 align:start position:0% change could be completely harmless or it<00:15:48.800> could<00:15:49.040> mean<00:15:49.360> cancer<00:15:50.000> which<00:15:50.240> is<00:15:50.480> obviously 00:15:51.110 --> 00:15:51.120 align:start position:0% it could mean cancer which is obviously 00:15:51.120 --> 00:15:52.629 align:start position:0% it could mean cancer which is obviously life-threatening.<00:15:51.920> But<00:15:52.079> it's<00:15:52.240> really<00:15:52.480> hard 00:15:52.629 --> 00:15:52.639 align:start position:0% life-threatening. But it's really hard 00:15:52.639 --> 00:15:55.189 align:start position:0% life-threatening. But it's really hard for<00:15:52.880> a<00:15:53.040> human<00:15:53.600> to<00:15:53.839> actually<00:15:54.399> figure<00:15:54.639> out<00:15:54.880> which 00:15:55.189 --> 00:15:55.199 align:start position:0% for a human to actually figure out which 00:15:55.199 --> 00:15:57.670 align:start position:0% for a human to actually figure out which mutations<00:15:55.920> are<00:15:56.240> okay,<00:15:56.720> which<00:15:56.959> mutations<00:15:57.519> are 00:15:57.670 --> 00:15:57.680 align:start position:0% mutations are okay, which mutations are 00:15:57.680 --> 00:15:59.910 align:start position:0% mutations are okay, which mutations are not.<00:15:58.160> You<00:15:58.320> see,<00:15:58.639> every<00:15:58.880> human<00:15:59.199> has<00:15:59.519> thousands 00:15:59.910 --> 00:15:59.920 align:start position:0% not. You see, every human has thousands 00:15:59.920 --> 00:16:02.150 align:start position:0% not. You see, every human has thousands of<00:16:00.160> genetic<00:16:00.639> variations.<00:16:01.440> And<00:16:01.600> in<00:16:01.759> a<00:16:01.920> sense, 00:16:02.150 --> 00:16:02.160 align:start position:0% of genetic variations. And in a sense, 00:16:02.160 --> 00:16:04.230 align:start position:0% of genetic variations. And in a sense, we<00:16:02.399> all<00:16:02.639> have<00:16:02.880> different<00:16:03.199> mutations.<00:16:04.000> The 00:16:04.230 --> 00:16:04.240 align:start position:0% we all have different mutations. The 00:16:04.240 --> 00:16:06.150 align:start position:0% we all have different mutations. The overwhelming<00:16:04.880> majority<00:16:05.279> of<00:16:05.440> the<00:16:05.680> time,<00:16:05.920> these 00:16:06.150 --> 00:16:06.160 align:start position:0% overwhelming majority of the time, these 00:16:06.160 --> 00:16:08.389 align:start position:0% overwhelming majority of the time, these mutations<00:16:06.800> don't<00:16:07.120> actually<00:16:07.600> cause<00:16:07.920> anything. 00:16:08.389 --> 00:16:08.399 align:start position:0% mutations don't actually cause anything. 00:16:08.399 --> 00:16:10.870 align:start position:0% mutations don't actually cause anything. But<00:16:08.560> the<00:16:08.720> trick<00:16:08.880> is<00:16:09.120> to<00:16:09.440> identify<00:16:10.240> exactly 00:16:10.870 --> 00:16:10.880 align:start position:0% But the trick is to identify exactly 00:16:10.880 --> 00:16:13.110 align:start position:0% But the trick is to identify exactly which<00:16:11.199> mutations<00:16:12.079> actually<00:16:12.639> increase<00:16:12.880> the 00:16:13.110 --> 00:16:13.120 align:start position:0% which mutations actually increase the 00:16:13.120 --> 00:16:15.110 align:start position:0% which mutations actually increase the risk<00:16:13.360> of<00:16:13.600> cancer,<00:16:14.079> which<00:16:14.320> is<00:16:14.480> a<00:16:14.639> lot<00:16:14.800> harder 00:16:15.110 --> 00:16:15.120 align:start position:0% risk of cancer, which is a lot harder 00:16:15.120 --> 00:16:17.269 align:start position:0% risk of cancer, which is a lot harder said<00:16:15.360> than<00:16:15.600> done.<00:16:16.160> So<00:16:16.320> the<00:16:16.480> researchers<00:16:16.959> gave 00:16:17.269 --> 00:16:17.279 align:start position:0% said than done. So the researchers gave 00:16:17.279 --> 00:16:19.829 align:start position:0% said than done. So the researchers gave EVO<00:16:17.759> 2<00:16:18.160> this<00:16:18.480> data<00:16:18.880> and<00:16:19.120> they<00:16:19.279> asked<00:16:19.519> it<00:16:19.680> to 00:16:19.829 --> 00:16:19.839 align:start position:0% EVO 2 this data and they asked it to 00:16:19.839 --> 00:16:22.310 align:start position:0% EVO 2 this data and they asked it to classify<00:16:20.399> which<00:16:20.639> ones<00:16:20.959> are<00:16:21.199> okay,<00:16:21.839> which<00:16:22.000> ones 00:16:22.310 --> 00:16:22.320 align:start position:0% classify which ones are okay, which ones 00:16:22.320 --> 00:16:24.230 align:start position:0% classify which ones are okay, which ones are<00:16:22.560> damaging.<00:16:23.120> Again,<00:16:23.440> note<00:16:23.600> that<00:16:23.759> the<00:16:23.920> model 00:16:24.230 --> 00:16:24.240 align:start position:0% are damaging. Again, note that the model 00:16:24.240 --> 00:16:26.550 align:start position:0% are damaging. Again, note that the model has<00:16:24.560> never<00:16:24.880> been<00:16:25.040> trained<00:16:25.440> on<00:16:25.759> human<00:16:26.160> medical 00:16:26.550 --> 00:16:26.560 align:start position:0% has never been trained on human medical 00:16:26.560 --> 00:16:29.269 align:start position:0% has never been trained on human medical records.<00:16:27.199> It<00:16:27.440> was<00:16:27.600> never<00:16:28.079> given<00:16:28.480> any<00:16:28.800> answers 00:16:29.269 --> 00:16:29.279 align:start position:0% records. It was never given any answers 00:16:29.279 --> 00:16:31.749 align:start position:0% records. It was never given any answers on<00:16:29.680> which<00:16:30.000> genes<00:16:30.320> cause<00:16:30.639> cancer<00:16:31.040> and<00:16:31.360> which 00:16:31.749 --> 00:16:31.759 align:start position:0% on which genes cause cancer and which 00:16:31.759 --> 00:16:33.430 align:start position:0% on which genes cause cancer and which don't.<00:16:32.240> It<00:16:32.399> doesn't<00:16:32.560> even<00:16:32.720> know<00:16:32.880> what<00:16:33.040> cancer 00:16:33.430 --> 00:16:33.440 align:start position:0% don't. It doesn't even know what cancer 00:16:33.440 --> 00:16:36.310 align:start position:0% don't. It doesn't even know what cancer is.<00:16:34.079> Yet,<00:16:34.320> the<00:16:34.560> results<00:16:35.040> were<00:16:35.440> shocking.<00:16:36.079> The 00:16:36.310 --> 00:16:36.320 align:start position:0% is. Yet, the results were shocking. The 00:16:36.320 --> 00:16:39.030 align:start position:0% is. Yet, the results were shocking. The model<00:16:36.560> was<00:16:36.880> able<00:16:37.120> to<00:16:37.360> directly<00:16:38.079> identify<00:16:38.720> the 00:16:39.030 --> 00:16:39.040 align:start position:0% model was able to directly identify the 00:16:39.040 --> 00:16:41.350 align:start position:0% model was able to directly identify the mutations<00:16:39.600> that<00:16:40.000> don't<00:16:40.240> do<00:16:40.480> anything<00:16:40.880> and<00:16:41.120> the 00:16:41.350 --> 00:16:41.360 align:start position:0% mutations that don't do anything and the 00:16:41.360 --> 00:16:43.590 align:start position:0% mutations that don't do anything and the mutations<00:16:41.920> that<00:16:42.320> cause<00:16:42.639> cancer.<00:16:43.279> So,<00:16:43.440> it 00:16:43.590 --> 00:16:43.600 align:start position:0% mutations that cause cancer. So, it 00:16:43.600 --> 00:16:45.829 align:start position:0% mutations that cause cancer. So, it turns<00:16:43.759> out<00:16:43.920> EVO<00:16:44.480> 2<00:16:44.639> is<00:16:44.800> also<00:16:45.120> a<00:16:45.360> really<00:16:45.600> good 00:16:45.829 --> 00:16:45.839 align:start position:0% turns out EVO 2 is also a really good 00:16:45.839 --> 00:16:48.069 align:start position:0% turns out EVO 2 is also a really good detector<00:16:46.320> of<00:16:46.560> cancer<00:16:47.199> just<00:16:47.440> from<00:16:47.680> looking<00:16:47.920> at 00:16:48.069 --> 00:16:48.079 align:start position:0% detector of cancer just from looking at 00:16:48.079 --> 00:16:50.470 align:start position:0% detector of cancer just from looking at someone's<00:16:48.560> DNA.<00:16:49.199> Now,<00:16:49.440> up<00:16:49.680> to<00:16:49.920> this<00:16:50.160> point, 00:16:50.470 --> 00:16:50.480 align:start position:0% someone's DNA. Now, up to this point, 00:16:50.480 --> 00:16:52.389 align:start position:0% someone's DNA. Now, up to this point, it's<00:16:50.720> all<00:16:50.959> analytical.<00:16:51.759> It's<00:16:51.920> able<00:16:52.160> to 00:16:52.389 --> 00:16:52.399 align:start position:0% it's all analytical. It's able to 00:16:52.399 --> 00:16:54.949 align:start position:0% it's all analytical. It's able to analyze<00:16:52.880> and<00:16:53.199> interpret<00:16:53.680> DNA.<00:16:54.399> But<00:16:54.560> can<00:16:54.800> it 00:16:54.949 --> 00:16:54.959 align:start position:0% analyze and interpret DNA. But can it 00:16:54.959 --> 00:16:56.790 align:start position:0% analyze and interpret DNA. But can it generate<00:16:55.440> new<00:16:55.680> functional<00:16:56.160> DNA<00:16:56.560> from 00:16:56.790 --> 00:16:56.800 align:start position:0% generate new functional DNA from 00:16:56.800 --> 00:16:59.030 align:start position:0% generate new functional DNA from scratch?<00:16:57.279> Not<00:16:57.440> just<00:16:57.600> a<00:16:57.839> single<00:16:58.160> protein<00:16:58.560> or<00:16:58.800> a 00:16:59.030 --> 00:16:59.040 align:start position:0% scratch? Not just a single protein or a 00:16:59.040 --> 00:17:02.069 align:start position:0% scratch? Not just a single protein or a tiny<00:16:59.279> gene,<00:16:59.680> but<00:16:59.920> the<00:17:00.240> entire<00:17:00.959> continuous<00:17:01.600> DNA 00:17:02.069 --> 00:17:02.079 align:start position:0% tiny gene, but the entire continuous DNA 00:17:02.079 --> 00:17:04.069 align:start position:0% tiny gene, but the entire continuous DNA of<00:17:02.240> a<00:17:02.480> living<00:17:02.720> thing?<00:17:03.360> Well,<00:17:03.600> they<00:17:03.839> started 00:17:04.069 --> 00:17:04.079 align:start position:0% of a living thing? Well, they started 00:17:04.079 --> 00:17:06.230 align:start position:0% of a living thing? Well, they started with<00:17:04.400> human<00:17:04.799> mitochondria.<00:17:05.679> If<00:17:05.919> you're<00:17:06.000> not 00:17:06.230 --> 00:17:06.240 align:start position:0% with human mitochondria. If you're not 00:17:06.240 --> 00:17:08.309 align:start position:0% with human mitochondria. If you're not familiar,<00:17:06.640> mitochondria<00:17:07.520> is<00:17:07.760> the<00:17:07.919> part<00:17:08.160> of 00:17:08.309 --> 00:17:08.319 align:start position:0% familiar, mitochondria is the part of 00:17:08.319 --> 00:17:10.470 align:start position:0% familiar, mitochondria is the part of your<00:17:08.640> cell<00:17:08.880> that<00:17:09.199> generates<00:17:09.679> energy.<00:17:10.319> Think 00:17:10.470 --> 00:17:10.480 align:start position:0% your cell that generates energy. Think 00:17:10.480 --> 00:17:12.470 align:start position:0% your cell that generates energy. Think of<00:17:10.640> it<00:17:10.799> as<00:17:10.959> like<00:17:11.199> the<00:17:11.439> powerhouse<00:17:12.000> of<00:17:12.240> your 00:17:12.470 --> 00:17:12.480 align:start position:0% of it as like the powerhouse of your 00:17:12.480 --> 00:17:13.669 align:start position:0% of it as like the powerhouse of your cell.<00:17:12.799> Now,<00:17:12.959> it<00:17:13.120> turns<00:17:13.280> out<00:17:13.439> that 00:17:13.669 --> 00:17:13.679 align:start position:0% cell. Now, it turns out that 00:17:13.679 --> 00:17:15.829 align:start position:0% cell. Now, it turns out that mitochondria<00:17:14.480> have<00:17:14.640> their<00:17:14.880> own<00:17:15.120> DNA<00:17:15.600> that's 00:17:15.829 --> 00:17:15.839 align:start position:0% mitochondria have their own DNA that's 00:17:15.839 --> 00:17:17.829 align:start position:0% mitochondria have their own DNA that's separate<00:17:16.160> from<00:17:16.480> our<00:17:16.720> cell's<00:17:17.039> DNA.<00:17:17.600> And 00:17:17.829 --> 00:17:17.839 align:start position:0% separate from our cell's DNA. And 00:17:17.839 --> 00:17:20.470 align:start position:0% separate from our cell's DNA. And mitochondria<00:17:18.559> DNA<00:17:18.959> is<00:17:19.360> way<00:17:19.679> smaller.<00:17:20.160> It's 00:17:20.470 --> 00:17:20.480 align:start position:0% mitochondria DNA is way smaller. It's 00:17:20.480 --> 00:17:23.590 align:start position:0% mitochondria DNA is way smaller. It's only<00:17:20.799> roughly<00:17:21.520> 16,000<00:17:22.480> letters.<00:17:23.120> So,<00:17:23.360> it's 00:17:23.590 --> 00:17:23.600 align:start position:0% only roughly 16,000 letters. So, it's 00:17:23.600 --> 00:17:25.750 align:start position:0% only roughly 16,000 letters. So, it's tiny<00:17:23.919> and<00:17:24.240> manageable,<00:17:24.720> but<00:17:25.039> still<00:17:25.360> quite 00:17:25.750 --> 00:17:25.760 align:start position:0% tiny and manageable, but still quite 00:17:25.760 --> 00:17:27.909 align:start position:0% tiny and manageable, but still quite complex.<00:17:26.480> Now,<00:17:26.720> what<00:17:26.959> the<00:17:27.120> researchers<00:17:27.679> did 00:17:27.909 --> 00:17:27.919 align:start position:0% complex. Now, what the researchers did 00:17:27.919 --> 00:17:30.470 align:start position:0% complex. Now, what the researchers did was<00:17:28.240> it<00:17:28.559> just<00:17:28.720> gave<00:17:28.960> EVO<00:17:29.440> 2<00:17:29.840> just<00:17:30.080> the<00:17:30.320> first 00:17:30.470 --> 00:17:30.480 align:start position:0% was it just gave EVO 2 just the first 00:17:30.480 --> 00:17:32.549 align:start position:0% was it just gave EVO 2 just the first few<00:17:30.720> letters<00:17:30.960> of<00:17:31.120> a<00:17:31.280> human<00:17:31.679> mitochondria 00:17:32.549 --> 00:17:32.559 align:start position:0% few letters of a human mitochondria 00:17:32.559 --> 00:17:34.150 align:start position:0% few letters of a human mitochondria genome<00:17:33.039> and<00:17:33.280> then<00:17:33.440> they<00:17:33.600> asked<00:17:33.760> it<00:17:34.000> to 00:17:34.150 --> 00:17:34.160 align:start position:0% genome and then they asked it to 00:17:34.160 --> 00:17:36.310 align:start position:0% genome and then they asked it to basically<00:17:34.799> finish<00:17:35.120> this<00:17:35.360> blueprint.<00:17:36.080> It's 00:17:36.310 --> 00:17:36.320 align:start position:0% basically finish this blueprint. It's 00:17:36.320 --> 00:17:38.390 align:start position:0% basically finish this blueprint. It's like<00:17:36.559> giving<00:17:37.039> an<00:17:37.200> AI<00:17:37.520> chatbot<00:17:38.000> the<00:17:38.240> first 00:17:38.390 --> 00:17:38.400 align:start position:0% like giving an AI chatbot the first 00:17:38.400 --> 00:17:40.230 align:start position:0% like giving an AI chatbot the first sentence<00:17:38.720> of<00:17:38.960> an<00:17:39.200> essay<00:17:39.440> and<00:17:39.600> then<00:17:39.760> telling<00:17:40.000> it 00:17:40.230 --> 00:17:40.240 align:start position:0% sentence of an essay and then telling it 00:17:40.240 --> 00:17:42.870 align:start position:0% sentence of an essay and then telling it to<00:17:40.480> finish<00:17:40.640> the<00:17:40.960> essay.<00:17:41.600> And<00:17:41.919> lo<00:17:42.160> and<00:17:42.320> behold, 00:17:42.870 --> 00:17:42.880 align:start position:0% to finish the essay. And lo and behold, 00:17:42.880 --> 00:17:45.909 align:start position:0% to finish the essay. And lo and behold, EVO<00:17:43.360> 2<00:17:43.679> was<00:17:43.919> able<00:17:44.240> to<00:17:44.559> successfully<00:17:45.280> generate 00:17:45.909 --> 00:17:45.919 align:start position:0% EVO 2 was able to successfully generate 00:17:45.919 --> 00:17:48.150 align:start position:0% EVO 2 was able to successfully generate a<00:17:46.240> completely<00:17:46.799> new<00:17:47.039> mitochondria<00:17:47.840> sequence 00:17:48.150 --> 00:17:48.160 align:start position:0% a completely new mitochondria sequence 00:17:48.160 --> 00:17:51.510 align:start position:0% a completely new mitochondria sequence of<00:17:48.400> roughly<00:17:49.200> 16,000<00:17:50.000> letters<00:17:50.480> from<00:17:50.799> scratch. 00:17:51.510 --> 00:17:51.520 align:start position:0% of roughly 16,000 letters from scratch. 00:17:51.520 --> 00:17:53.909 align:start position:0% of roughly 16,000 letters from scratch. Now,<00:17:51.840> generating<00:17:52.320> this<00:17:52.480> DNA<00:17:52.880> is<00:17:53.120> one<00:17:53.360> thing. 00:17:53.909 --> 00:17:53.919 align:start position:0% Now, generating this DNA is one thing. 00:17:53.919 --> 00:17:56.230 align:start position:0% Now, generating this DNA is one thing. How<00:17:54.080> do<00:17:54.160> we<00:17:54.320> know<00:17:54.480> if<00:17:54.799> this<00:17:55.039> DNA<00:17:55.520> is<00:17:55.760> actually 00:17:56.230 --> 00:17:56.240 align:start position:0% How do we know if this DNA is actually 00:17:56.240 --> 00:17:58.390 align:start position:0% How do we know if this DNA is actually legit?<00:17:57.120> How<00:17:57.360> do<00:17:57.440> we<00:17:57.600> know<00:17:57.840> if<00:17:58.080> it's 00:17:58.390 --> 00:17:58.400 align:start position:0% legit? How do we know if it's 00:17:58.400 --> 00:18:01.190 align:start position:0% legit? How do we know if it's biologically<00:17:59.440> viable?<00:18:00.080> So,<00:18:00.320> the<00:18:00.559> researchers 00:18:01.190 --> 00:18:01.200 align:start position:0% biologically viable? So, the researchers 00:18:01.200 --> 00:18:04.310 align:start position:0% biologically viable? So, the researchers used<00:18:01.600> some<00:18:02.080> external<00:18:02.640> validation<00:18:03.280> tools<00:18:04.000> to 00:18:04.310 --> 00:18:04.320 align:start position:0% used some external validation tools to 00:18:04.320 --> 00:18:06.710 align:start position:0% used some external validation tools to assess<00:18:04.960> this<00:18:05.360> generation.<00:18:06.080> They<00:18:06.320> started 00:18:06.710 --> 00:18:06.720 align:start position:0% assess this generation. They started 00:18:06.720 --> 00:18:09.669 align:start position:0% assess this generation. They started with<00:18:06.960> a<00:18:07.200> tool<00:18:07.520> called<00:18:08.000> Maidoz,<00:18:09.200> which<00:18:09.440> is<00:18:09.520> a 00:18:09.669 --> 00:18:09.679 align:start position:0% with a tool called Maidoz, which is a 00:18:09.679 --> 00:18:12.150 align:start position:0% with a tool called Maidoz, which is a wellestablished<00:18:10.720> software<00:18:11.200> for<00:18:11.520> analyzing 00:18:12.150 --> 00:18:12.160 align:start position:0% wellestablished software for analyzing 00:18:12.160 --> 00:18:14.470 align:start position:0% wellestablished software for analyzing this<00:18:12.400> mitochondria<00:18:13.200> DNA.<00:18:13.760> And<00:18:13.919> the<00:18:14.080> results 00:18:14.470 --> 00:18:14.480 align:start position:0% this mitochondria DNA. And the results 00:18:14.480 --> 00:18:17.270 align:start position:0% this mitochondria DNA. And the results were<00:18:14.799> astonishing.<00:18:15.600> Mitoz<00:18:16.400> confirmed<00:18:16.880> that 00:18:17.270 --> 00:18:17.280 align:start position:0% were astonishing. Mitoz confirmed that 00:18:17.280 --> 00:18:19.750 align:start position:0% were astonishing. Mitoz confirmed that EVO<00:18:17.760> 2's<00:18:18.320> madeup<00:18:18.799> generation<00:18:19.440> of 00:18:19.750 --> 00:18:19.760 align:start position:0% EVO 2's madeup generation of 00:18:19.760 --> 00:18:22.310 align:start position:0% EVO 2's madeup generation of mitochondria<00:18:20.640> DNA<00:18:21.360> actually<00:18:21.679> contained<00:18:22.080> the 00:18:22.310 --> 00:18:22.320 align:start position:0% mitochondria DNA actually contained the 00:18:22.320 --> 00:18:24.549 align:start position:0% mitochondria DNA actually contained the correct<00:18:22.640> number<00:18:22.880> of<00:18:23.200> protein<00:18:23.760> coding<00:18:24.080> genes 00:18:24.549 --> 00:18:24.559 align:start position:0% correct number of protein coding genes 00:18:24.559 --> 00:18:28.310 align:start position:0% correct number of protein coding genes as<00:18:24.720> well<00:18:24.880> as<00:18:25.120> tRNA<00:18:25.840> genes<00:18:26.160> and<00:18:26.559> rRNA<00:18:27.280> genes<00:18:28.000> as 00:18:28.310 --> 00:18:28.320 align:start position:0% as well as tRNA genes and rRNA genes as 00:18:28.320 --> 00:18:31.270 align:start position:0% as well as tRNA genes and rRNA genes as expected<00:18:28.880> in<00:18:29.440> real<00:18:29.919> human<00:18:30.240> mitochondria.<00:18:31.039> Now 00:18:31.270 --> 00:18:31.280 align:start position:0% expected in real human mitochondria. Now 00:18:31.280 --> 00:18:33.430 align:start position:0% expected in real human mitochondria. Now what<00:18:31.520> exactly<00:18:31.919> do<00:18:32.160> all<00:18:32.320> of<00:18:32.400> these<00:18:32.640> terms<00:18:32.960> mean? 00:18:33.430 --> 00:18:33.440 align:start position:0% what exactly do all of these terms mean? 00:18:33.440 --> 00:18:36.150 align:start position:0% what exactly do all of these terms mean? Let's<00:18:33.679> break<00:18:33.919> this<00:18:34.160> down.<00:18:34.559> So<00:18:34.880> DNA<00:18:35.360> is,<00:18:35.760> as<00:18:36.000> you 00:18:36.150 --> 00:18:36.160 align:start position:0% Let's break this down. So DNA is, as you 00:18:36.160 --> 00:18:38.870 align:start position:0% Let's break this down. So DNA is, as you know,<00:18:36.320> the<00:18:36.559> master<00:18:37.120> blueprint<00:18:37.760> of<00:18:38.000> life.<00:18:38.640> It 00:18:38.870 --> 00:18:38.880 align:start position:0% know, the master blueprint of life. It 00:18:38.880 --> 00:18:40.950 align:start position:0% know, the master blueprint of life. It contains<00:18:39.360> instructions<00:18:39.919> that<00:18:40.240> tell<00:18:40.400> a<00:18:40.640> cell 00:18:40.950 --> 00:18:40.960 align:start position:0% contains instructions that tell a cell 00:18:40.960 --> 00:18:43.430 align:start position:0% contains instructions that tell a cell how<00:18:41.200> to<00:18:41.440> build<00:18:41.679> and<00:18:41.919> operate<00:18:42.320> itself.<00:18:43.039> Now 00:18:43.430 --> 00:18:43.440 align:start position:0% how to build and operate itself. Now 00:18:43.440 --> 00:18:45.350 align:start position:0% how to build and operate itself. Now part<00:18:43.600> of<00:18:43.760> these<00:18:44.000> instructions<00:18:44.640> include<00:18:45.120> how 00:18:45.350 --> 00:18:45.360 align:start position:0% part of these instructions include how 00:18:45.360 --> 00:18:47.350 align:start position:0% part of these instructions include how to<00:18:45.520> build<00:18:45.919> proteins<00:18:46.480> which<00:18:46.720> are<00:18:46.880> like<00:18:47.039> the 00:18:47.350 --> 00:18:47.360 align:start position:0% to build proteins which are like the 00:18:47.360 --> 00:18:49.430 align:start position:0% to build proteins which are like the machines<00:18:47.919> and<00:18:48.160> tools<00:18:48.559> that<00:18:48.720> are<00:18:48.880> used<00:18:49.039> for<00:18:49.280> a 00:18:49.430 --> 00:18:49.440 align:start position:0% machines and tools that are used for a 00:18:49.440 --> 00:18:51.510 align:start position:0% machines and tools that are used for a cell<00:18:49.679> to<00:18:49.919> function.<00:18:50.320> Now<00:18:50.480> in<00:18:50.720> order<00:18:50.880> to<00:18:51.120> create 00:18:51.510 --> 00:18:51.520 align:start position:0% cell to function. Now in order to create 00:18:51.520 --> 00:18:54.710 align:start position:0% cell to function. Now in order to create these<00:18:52.080> proteins<00:18:52.720> DNA<00:18:53.280> also<00:18:53.600> needs<00:18:53.840> to<00:18:54.160> code<00:18:54.400> up 00:18:54.710 --> 00:18:54.720 align:start position:0% these proteins DNA also needs to code up 00:18:54.720 --> 00:18:58.230 align:start position:0% these proteins DNA also needs to code up stuff<00:18:55.039> called<00:18:55.520> ribosomal<00:18:56.400> RNAs<00:18:57.039> or<00:18:57.360> rRNAs 00:18:58.230 --> 00:18:58.240 align:start position:0% stuff called ribosomal RNAs or rRNAs 00:18:58.240 --> 00:18:59.990 align:start position:0% stuff called ribosomal RNAs or rRNAs which<00:18:58.480> are<00:18:58.640> like<00:18:58.799> the<00:18:59.039> factory<00:18:59.440> workers. 00:18:59.990 --> 00:19:00.000 align:start position:0% which are like the factory workers. 00:19:00.000 --> 00:19:01.590 align:start position:0% which are like the factory workers. These<00:19:00.320> are<00:19:00.480> in<00:19:00.720> charge<00:19:00.880> of<00:19:01.120> actually 00:19:01.590 --> 00:19:01.600 align:start position:0% These are in charge of actually 00:19:01.600 --> 00:19:03.909 align:start position:0% These are in charge of actually assembling<00:19:02.240> the<00:19:02.559> proteins.<00:19:03.360> And<00:19:03.520> then<00:19:03.679> there 00:19:03.909 --> 00:19:03.919 align:start position:0% assembling the proteins. And then there 00:19:03.919 --> 00:19:06.630 align:start position:0% assembling the proteins. And then there are<00:19:04.320> transfer<00:19:04.960> RNAs<00:19:05.679> which<00:19:05.919> you<00:19:06.160> can<00:19:06.240> think<00:19:06.400> of 00:19:06.630 --> 00:19:06.640 align:start position:0% are transfer RNAs which you can think of 00:19:06.640 --> 00:19:09.029 align:start position:0% are transfer RNAs which you can think of as<00:19:06.880> like<00:19:07.360> delivery<00:19:07.840> trucks<00:19:08.240> that<00:19:08.559> bring<00:19:08.799> the 00:19:09.029 --> 00:19:09.039 align:start position:0% as like delivery trucks that bring the 00:19:09.039 --> 00:19:11.190 align:start position:0% as like delivery trucks that bring the right<00:19:09.200> raw<00:19:09.520> materials<00:19:10.160> to<00:19:10.480> the<00:19:10.720> assembly 00:19:11.190 --> 00:19:11.200 align:start position:0% right raw materials to the assembly 00:19:11.200 --> 00:19:14.470 align:start position:0% right raw materials to the assembly line.<00:19:11.760> Each<00:19:12.160> tRNA<00:19:12.960> carries<00:19:13.360> a<00:19:13.600> specific<00:19:14.080> amino 00:19:14.470 --> 00:19:14.480 align:start position:0% line. Each tRNA carries a specific amino 00:19:14.480 --> 00:19:16.789 align:start position:0% line. Each tRNA carries a specific amino acid<00:19:14.799> and<00:19:15.120> places<00:19:15.440> it<00:19:15.679> precisely<00:19:16.320> where<00:19:16.640> it 00:19:16.789 --> 00:19:16.799 align:start position:0% acid and places it precisely where it 00:19:16.799 --> 00:19:18.310 align:start position:0% acid and places it precisely where it needs<00:19:17.039> to<00:19:17.200> go<00:19:17.520> according<00:19:17.919> to<00:19:18.080> the 00:19:18.310 --> 00:19:18.320 align:start position:0% needs to go according to the 00:19:18.320 --> 00:19:20.470 align:start position:0% needs to go according to the instructions.<00:19:19.120> So<00:19:19.280> DNA<00:19:19.760> basically<00:19:20.160> needs<00:19:20.320> to 00:19:20.470 --> 00:19:20.480 align:start position:0% instructions. So DNA basically needs to 00:19:20.480 --> 00:19:23.350 align:start position:0% instructions. So DNA basically needs to code<00:19:20.720> up<00:19:21.120> everything<00:19:21.760> for<00:19:22.320> this<00:19:22.640> factory<00:19:23.039> to 00:19:23.350 --> 00:19:23.360 align:start position:0% code up everything for this factory to 00:19:23.360 --> 00:19:25.029 align:start position:0% code up everything for this factory to work.<00:19:23.600> It<00:19:23.840> needs<00:19:24.000> to<00:19:24.080> code<00:19:24.240> up<00:19:24.480> what<00:19:24.799> to 00:19:25.029 --> 00:19:25.039 align:start position:0% work. It needs to code up what to 00:19:25.039 --> 00:19:26.950 align:start position:0% work. It needs to code up what to create.<00:19:25.440> It<00:19:25.600> needs<00:19:25.760> to<00:19:25.919> code<00:19:26.080> up<00:19:26.240> the<00:19:26.480> delivery 00:19:26.950 --> 00:19:26.960 align:start position:0% create. It needs to code up the delivery 00:19:26.960 --> 00:19:28.870 align:start position:0% create. It needs to code up the delivery system.<00:19:27.440> Plus,<00:19:27.760> it<00:19:27.919> needs<00:19:28.080> to<00:19:28.160> code<00:19:28.400> up<00:19:28.640> the 00:19:28.870 --> 00:19:28.880 align:start position:0% system. Plus, it needs to code up the 00:19:28.880 --> 00:19:31.510 align:start position:0% system. Plus, it needs to code up the factory<00:19:29.440> workers.<00:19:30.160> And<00:19:30.400> all<00:19:30.720> three<00:19:31.039> of<00:19:31.200> these 00:19:31.510 --> 00:19:31.520 align:start position:0% factory workers. And all three of these 00:19:31.520 --> 00:19:34.390 align:start position:0% factory workers. And all three of these things<00:19:32.000> are<00:19:32.559> crucial<00:19:33.039> for<00:19:33.360> an<00:19:33.600> organism<00:19:34.080> to 00:19:34.390 --> 00:19:34.400 align:start position:0% things are crucial for an organism to 00:19:34.400 --> 00:19:36.310 align:start position:0% things are crucial for an organism to function<00:19:34.720> and<00:19:35.039> stay<00:19:35.280> alive.<00:19:35.840> Now,<00:19:36.080> what's 00:19:36.310 --> 00:19:36.320 align:start position:0% function and stay alive. Now, what's 00:19:36.320 --> 00:19:38.950 align:start position:0% function and stay alive. Now, what's remarkable<00:19:36.799> about<00:19:37.120> EVO<00:19:37.600> 2's<00:19:38.000> generation<00:19:38.720> of 00:19:38.950 --> 00:19:38.960 align:start position:0% remarkable about EVO 2's generation of 00:19:38.960 --> 00:19:41.510 align:start position:0% remarkable about EVO 2's generation of this<00:19:39.280> imaginary<00:19:39.919> mitochondria<00:19:40.799> is<00:19:40.960> that<00:19:41.200> it 00:19:41.510 --> 00:19:41.520 align:start position:0% this imaginary mitochondria is that it 00:19:41.520 --> 00:19:43.590 align:start position:0% this imaginary mitochondria is that it includes<00:19:42.080> all<00:19:42.320> of<00:19:42.480> these<00:19:42.799> features.<00:19:43.440> It 00:19:43.590 --> 00:19:43.600 align:start position:0% includes all of these features. It 00:19:43.600 --> 00:19:46.150 align:start position:0% includes all of these features. It includes<00:19:44.080> code<00:19:44.559> on<00:19:44.880> what<00:19:45.200> proteins<00:19:45.679> to<00:19:45.840> build 00:19:46.150 --> 00:19:46.160 align:start position:0% includes code on what proteins to build 00:19:46.160 --> 00:19:48.150 align:start position:0% includes code on what proteins to build and<00:19:46.400> also<00:19:46.640> the<00:19:46.880> workers<00:19:47.280> in<00:19:47.520> order<00:19:47.679> to<00:19:48.000> build 00:19:48.150 --> 00:19:48.160 align:start position:0% and also the workers in order to build 00:19:48.160 --> 00:19:50.310 align:start position:0% and also the workers in order to build these<00:19:48.480> proteins<00:19:49.039> as<00:19:49.280> well<00:19:49.360> as<00:19:49.520> the<00:19:49.840> delivery 00:19:50.310 --> 00:19:50.320 align:start position:0% these proteins as well as the delivery 00:19:50.320 --> 00:19:53.590 align:start position:0% these proteins as well as the delivery system.<00:19:50.799> EVO<00:19:51.360> 2<00:19:51.600> was<00:19:51.840> able<00:19:52.160> to<00:19:52.880> generate<00:19:53.280> a 00:19:53.590 --> 00:19:53.600 align:start position:0% system. EVO 2 was able to generate a 00:19:53.600 --> 00:19:56.710 align:start position:0% system. EVO 2 was able to generate a fully<00:19:54.000> working<00:19:54.640> DNA<00:19:55.200> for<00:19:55.440> mitochondria.<00:19:56.559> But 00:19:56.710 --> 00:19:56.720 align:start position:0% fully working DNA for mitochondria. But 00:19:56.720 --> 00:19:58.549 align:start position:0% fully working DNA for mitochondria. But the<00:19:56.880> researchers<00:19:57.360> didn't<00:19:57.679> stop<00:19:57.919> there.<00:19:58.320> To 00:19:58.549 --> 00:19:58.559 align:start position:0% the researchers didn't stop there. To 00:19:58.559 --> 00:20:00.390 align:start position:0% the researchers didn't stop there. To make<00:19:58.640> sure<00:19:58.799> these<00:19:59.120> proteins<00:19:59.679> and<00:20:00.000> everything 00:20:00.390 --> 00:20:00.400 align:start position:0% make sure these proteins and everything 00:20:00.400 --> 00:20:03.110 align:start position:0% make sure these proteins and everything else<00:20:00.799> actually<00:20:01.280> work,<00:20:01.760> they<00:20:02.000> next<00:20:02.240> fed<00:20:02.799> this 00:20:03.110 --> 00:20:03.120 align:start position:0% else actually work, they next fed this 00:20:03.120 --> 00:20:06.230 align:start position:0% else actually work, they next fed this generation<00:20:03.840> through<00:20:04.160> AlphaFold<00:20:05.120> 3.<00:20:05.840> This<00:20:06.080> is 00:20:06.230 --> 00:20:06.240 align:start position:0% generation through AlphaFold 3. This is 00:20:06.240 --> 00:20:08.470 align:start position:0% generation through AlphaFold 3. This is the<00:20:06.480> legendary<00:20:07.039> AI<00:20:07.440> from<00:20:07.679> Google<00:20:08.000> DeepMind 00:20:08.470 --> 00:20:08.480 align:start position:0% the legendary AI from Google DeepMind 00:20:08.480 --> 00:20:10.950 align:start position:0% the legendary AI from Google DeepMind that<00:20:08.720> predicts<00:20:09.039> how<00:20:09.280> proteins<00:20:09.760> fold<00:20:10.160> into<00:20:10.480> 3D 00:20:10.950 --> 00:20:10.960 align:start position:0% that predicts how proteins fold into 3D 00:20:10.960 --> 00:20:12.310 align:start position:0% that predicts how proteins fold into 3D shapes.<00:20:11.360> By<00:20:11.520> the<00:20:11.600> way,<00:20:11.760> let<00:20:11.919> me<00:20:12.000> know<00:20:12.080> in<00:20:12.240> the 00:20:12.310 --> 00:20:12.320 align:start position:0% shapes. By the way, let me know in the 00:20:12.320 --> 00:20:14.230 align:start position:0% shapes. By the way, let me know in the comments<00:20:12.559> below<00:20:12.799> if<00:20:13.039> you<00:20:13.120> want<00:20:13.280> me<00:20:13.440> to<00:20:13.679> make<00:20:13.919> an 00:20:14.230 --> 00:20:14.240 align:start position:0% comments below if you want me to make an 00:20:14.240 --> 00:20:16.630 align:start position:0% comments below if you want me to make an explainer<00:20:14.880> video<00:20:15.200> on<00:20:15.440> Alphafold.<00:20:16.160> I<00:20:16.320> might<00:20:16.480> do 00:20:16.630 --> 00:20:16.640 align:start position:0% explainer video on Alphafold. I might do 00:20:16.640 --> 00:20:17.909 align:start position:0% explainer video on Alphafold. I might do one<00:20:16.799> if<00:20:16.880> there<00:20:17.039> is<00:20:17.200> enough<00:20:17.440> interest. 00:20:17.909 --> 00:20:17.919 align:start position:0% one if there is enough interest. 00:20:17.919 --> 00:20:20.710 align:start position:0% one if there is enough interest. Anyways,<00:20:18.400> in<00:20:18.559> short,<00:20:18.960> Alphafold<00:20:19.760> 3<00:20:20.240> also 00:20:20.710 --> 00:20:20.720 align:start position:0% Anyways, in short, Alphafold 3 also 00:20:20.720 --> 00:20:23.190 align:start position:0% Anyways, in short, Alphafold 3 also confirmed<00:20:21.440> that<00:20:21.760> Evo<00:20:22.240> 2's<00:20:22.559> generated 00:20:23.190 --> 00:20:23.200 align:start position:0% confirmed that Evo 2's generated 00:20:23.200 --> 00:20:25.430 align:start position:0% confirmed that Evo 2's generated proteins<00:20:23.840> actually<00:20:24.240> folded<00:20:24.720> correctly.<00:20:25.280> And 00:20:25.430 --> 00:20:25.440 align:start position:0% proteins actually folded correctly. And 00:20:25.440 --> 00:20:28.230 align:start position:0% proteins actually folded correctly. And what's<00:20:25.760> even<00:20:26.000> more<00:20:26.400> important<00:20:26.960> is<00:20:27.200> that<00:20:27.600> these 00:20:28.230 --> 00:20:28.240 align:start position:0% what's even more important is that these 00:20:28.240 --> 00:20:30.630 align:start position:0% what's even more important is that these different<00:20:28.880> generated<00:20:29.440> proteins<00:20:30.080> physically 00:20:30.630 --> 00:20:30.640 align:start position:0% different generated proteins physically 00:20:30.640 --> 00:20:32.950 align:start position:0% different generated proteins physically interlocked<00:20:31.520> with<00:20:31.840> each<00:20:32.080> other<00:20:32.320> like<00:20:32.640> pieces 00:20:32.950 --> 00:20:32.960 align:start position:0% interlocked with each other like pieces 00:20:32.960 --> 00:20:35.590 align:start position:0% interlocked with each other like pieces of<00:20:33.120> a<00:20:33.280> puzzle<00:20:33.840> exactly<00:20:34.559> as<00:20:34.880> they<00:20:35.120> need<00:20:35.280> to<00:20:35.440> do 00:20:35.590 --> 00:20:35.600 align:start position:0% of a puzzle exactly as they need to do 00:20:35.600 --> 00:20:37.909 align:start position:0% of a puzzle exactly as they need to do in<00:20:35.919> real<00:20:36.240> human<00:20:36.640> cells<00:20:36.960> in<00:20:37.280> order<00:20:37.360> to<00:20:37.600> produce 00:20:37.909 --> 00:20:37.919 align:start position:0% in real human cells in order to produce 00:20:37.919 --> 00:20:40.870 align:start position:0% in real human cells in order to produce energy.<00:20:38.559> So<00:20:38.720> this<00:20:38.960> confirms<00:20:39.360> that<00:20:39.760> this<00:20:40.159> DNA 00:20:40.870 --> 00:20:40.880 align:start position:0% energy. So this confirms that this DNA 00:20:40.880 --> 00:20:43.669 align:start position:0% energy. So this confirms that this DNA which<00:20:41.200> was<00:20:41.520> just<00:20:41.919> generated<00:20:42.480> from<00:20:42.799> EVO<00:20:43.280> 2 00:20:43.669 --> 00:20:43.679 align:start position:0% which was just generated from EVO 2 00:20:43.679 --> 00:20:45.909 align:start position:0% which was just generated from EVO 2 actually<00:20:44.080> works<00:20:44.720> just<00:20:44.960> like<00:20:45.200> in<00:20:45.520> real 00:20:45.909 --> 00:20:45.919 align:start position:0% actually works just like in real 00:20:45.919 --> 00:20:47.990 align:start position:0% actually works just like in real mitochondria.<00:20:46.880> Okay,<00:20:47.120> so<00:20:47.280> we've<00:20:47.600> proven<00:20:47.840> that 00:20:47.990 --> 00:20:48.000 align:start position:0% mitochondria. Okay, so we've proven that 00:20:48.000 --> 00:20:50.470 align:start position:0% mitochondria. Okay, so we've proven that EVO2<00:20:48.720> can<00:20:48.880> generate<00:20:49.200> the<00:20:49.360> DNA<00:20:49.760> of<00:20:50.159> this 00:20:50.470 --> 00:20:50.480 align:start position:0% EVO2 can generate the DNA of this 00:20:50.480 --> 00:20:52.710 align:start position:0% EVO2 can generate the DNA of this organel.<00:20:51.280> What<00:20:51.440> if<00:20:51.600> we<00:20:51.760> step<00:20:51.919> it<00:20:52.080> up<00:20:52.240> a<00:20:52.400> notch? 00:20:52.710 --> 00:20:52.720 align:start position:0% organel. What if we step it up a notch? 00:20:52.720 --> 00:20:54.549 align:start position:0% organel. What if we step it up a notch? Can<00:20:52.960> we<00:20:53.120> also<00:20:53.360> get<00:20:53.520> it<00:20:53.679> to<00:20:53.840> generate<00:20:54.240> the 00:20:54.549 --> 00:20:54.559 align:start position:0% Can we also get it to generate the 00:20:54.559 --> 00:20:57.029 align:start position:0% Can we also get it to generate the complete<00:20:54.960> DNA<00:20:55.440> of<00:20:55.600> a<00:20:55.840> living<00:20:56.080> thing?<00:20:56.720> Well, 00:20:57.029 --> 00:20:57.039 align:start position:0% complete DNA of a living thing? Well, 00:20:57.039 --> 00:20:58.710 align:start position:0% complete DNA of a living thing? Well, next<00:20:57.280> the<00:20:57.440> researchers<00:20:57.919> got<00:20:58.080> it<00:20:58.240> to<00:20:58.400> generate 00:20:58.710 --> 00:20:58.720 align:start position:0% next the researchers got it to generate 00:20:58.720 --> 00:21:01.830 align:start position:0% next the researchers got it to generate the<00:20:59.039> full<00:20:59.360> genome<00:20:59.919> of<00:21:00.159> a<00:21:00.400> bacteria<00:21:01.039> called 00:21:01.830 --> 00:21:01.840 align:start position:0% the full genome of a bacteria called 00:21:01.840 --> 00:21:04.870 align:start position:0% the full genome of a bacteria called myopplasma<00:21:03.120> genitalia<00:21:04.080> which<00:21:04.400> looks<00:21:04.720> like 00:21:04.870 --> 00:21:04.880 align:start position:0% myopplasma genitalia which looks like 00:21:04.880 --> 00:21:06.950 align:start position:0% myopplasma genitalia which looks like this<00:21:05.120> for<00:21:05.360> your<00:21:05.520> reference.<00:21:06.000> Now<00:21:06.159> the<00:21:06.320> DNA<00:21:06.720> of 00:21:06.950 --> 00:21:06.960 align:start position:0% this for your reference. Now the DNA of 00:21:06.960 --> 00:21:09.990 align:start position:0% this for your reference. Now the DNA of this<00:21:07.280> is<00:21:07.600> roughly<00:21:08.320> 580,000 00:21:09.990 --> 00:21:10.000 align:start position:0% this is roughly 580,000 00:21:10.000 --> 00:21:11.750 align:start position:0% this is roughly 580,000 letters.<00:21:10.640> And<00:21:10.720> again<00:21:10.960> the<00:21:11.120> researchers<00:21:11.600> gave 00:21:11.750 --> 00:21:11.760 align:start position:0% letters. And again the researchers gave 00:21:11.760 --> 00:21:13.669 align:start position:0% letters. And again the researchers gave it<00:21:12.000> just<00:21:12.159> like<00:21:12.400> the<00:21:12.640> first<00:21:12.880> few<00:21:13.120> letters<00:21:13.520> of 00:21:13.669 --> 00:21:13.679 align:start position:0% it just like the first few letters of 00:21:13.679 --> 00:21:15.830 align:start position:0% it just like the first few letters of this<00:21:13.919> DNA<00:21:14.320> and<00:21:14.559> asked<00:21:14.799> it<00:21:15.039> to<00:21:15.280> generate<00:21:15.679> the 00:21:15.830 --> 00:21:15.840 align:start position:0% this DNA and asked it to generate the 00:21:15.840 --> 00:21:18.549 align:start position:0% this DNA and asked it to generate the rest.<00:21:16.400> And<00:21:16.559> in<00:21:16.799> short<00:21:17.120> it<00:21:17.360> also<00:21:17.760> succeeded.<00:21:18.320> it 00:21:18.549 --> 00:21:18.559 align:start position:0% rest. And in short it also succeeded. it 00:21:18.559 --> 00:21:20.710 align:start position:0% rest. And in short it also succeeded. it was<00:21:18.640> able<00:21:18.880> to<00:21:19.039> generate<00:21:19.440> a<00:21:19.840> complete<00:21:20.320> and 00:21:20.710 --> 00:21:20.720 align:start position:0% was able to generate a complete and 00:21:20.720 --> 00:21:23.430 align:start position:0% was able to generate a complete and fully<00:21:21.200> functional<00:21:21.919> genome<00:21:22.559> that<00:21:22.880> resembles 00:21:23.430 --> 00:21:23.440 align:start position:0% fully functional genome that resembles 00:21:23.440 --> 00:21:26.070 align:start position:0% fully functional genome that resembles this<00:21:23.840> bacteria<00:21:24.320> in<00:21:24.640> real<00:21:24.880> life.<00:21:25.440> Now,<00:21:25.760> instead 00:21:26.070 --> 00:21:26.080 align:start position:0% this bacteria in real life. Now, instead 00:21:26.080 --> 00:21:28.149 align:start position:0% this bacteria in real life. Now, instead of<00:21:26.320> just<00:21:26.640> bacteria,<00:21:27.280> what<00:21:27.440> if<00:21:27.520> we<00:21:27.679> got<00:21:27.840> it<00:21:28.000> to 00:21:28.149 --> 00:21:28.159 align:start position:0% of just bacteria, what if we got it to 00:21:28.159 --> 00:21:30.390 align:start position:0% of just bacteria, what if we got it to generate<00:21:28.559> something<00:21:28.960> even<00:21:29.280> more<00:21:29.600> complex? 00:21:30.390 --> 00:21:30.400 align:start position:0% generate something even more complex? 00:21:30.400 --> 00:21:32.950 align:start position:0% generate something even more complex? So,<00:21:30.559> next,<00:21:30.880> the<00:21:31.120> scientists<00:21:31.600> asked<00:21:31.919> Evo<00:21:32.400> 2<00:21:32.640> to 00:21:32.950 --> 00:21:32.960 align:start position:0% So, next, the scientists asked Evo 2 to 00:21:32.960 --> 00:21:35.750 align:start position:0% So, next, the scientists asked Evo 2 to generate<00:21:33.360> the<00:21:33.679> entire<00:21:34.159> genome<00:21:34.640> of<00:21:35.120> this<00:21:35.440> other 00:21:35.750 --> 00:21:35.760 align:start position:0% generate the entire genome of this other 00:21:35.760 --> 00:21:39.750 align:start position:0% generate the entire genome of this other organism<00:21:36.480> called<00:21:36.960> Sakaromyces<00:21:38.240> Sur<00:21:38.799> Visay. 00:21:39.750 --> 00:21:39.760 align:start position:0% organism called Sakaromyces Sur Visay. 00:21:39.760 --> 00:21:42.070 align:start position:0% organism called Sakaromyces Sur Visay. This<00:21:39.919> is<00:21:40.159> basically<00:21:40.720> a<00:21:40.960> yeast,<00:21:41.440> which<00:21:41.679> is<00:21:41.840> very 00:21:42.070 --> 00:21:42.080 align:start position:0% This is basically a yeast, which is very 00:21:42.080 --> 00:21:44.149 align:start position:0% This is basically a yeast, which is very different<00:21:42.240> from<00:21:42.559> bacteria.<00:21:43.200> In<00:21:43.360> fact,<00:21:43.760> this 00:21:44.149 --> 00:21:44.159 align:start position:0% different from bacteria. In fact, this 00:21:44.159 --> 00:21:46.390 align:start position:0% different from bacteria. In fact, this organism<00:21:44.880> is<00:21:45.120> more<00:21:45.360> related<00:21:45.679> to<00:21:45.840> humans<00:21:46.240> than 00:21:46.390 --> 00:21:46.400 align:start position:0% organism is more related to humans than 00:21:46.400 --> 00:21:48.470 align:start position:0% organism is more related to humans than it<00:21:46.559> is<00:21:46.720> to<00:21:46.960> bacteria.<00:21:47.520> And<00:21:47.760> again,<00:21:48.000> they<00:21:48.320> just 00:21:48.470 --> 00:21:48.480 align:start position:0% it is to bacteria. And again, they just 00:21:48.480 --> 00:21:50.630 align:start position:0% it is to bacteria. And again, they just gave<00:21:48.720> it<00:21:48.880> the<00:21:49.120> first<00:21:49.280> few<00:21:49.440> letters<00:21:49.840> of<00:21:50.000> its<00:21:50.240> DNA 00:21:50.630 --> 00:21:50.640 align:start position:0% gave it the first few letters of its DNA 00:21:50.640 --> 00:21:52.789 align:start position:0% gave it the first few letters of its DNA and<00:21:50.799> then<00:21:50.960> asked<00:21:51.200> it<00:21:51.440> to<00:21:51.600> generate<00:21:52.080> the<00:21:52.400> entire 00:21:52.789 --> 00:21:52.799 align:start position:0% and then asked it to generate the entire 00:21:52.799 --> 00:21:55.350 align:start position:0% and then asked it to generate the entire thing.<00:21:53.200> And<00:21:53.360> lo<00:21:53.520> and<00:21:53.760> behold,<00:21:54.159> it<00:21:54.480> was<00:21:54.640> able<00:21:54.960> to 00:21:55.350 --> 00:21:55.360 align:start position:0% thing. And lo and behold, it was able to 00:21:55.360 --> 00:21:57.830 align:start position:0% thing. And lo and behold, it was able to successfully<00:21:56.159> generate<00:21:56.720> a<00:21:57.039> fully<00:21:57.520> working 00:21:57.830 --> 00:21:57.840 align:start position:0% successfully generate a fully working 00:21:57.840 --> 00:22:00.470 align:start position:0% successfully generate a fully working DNA<00:21:58.480> of<00:21:58.720> this<00:21:58.880> yeast,<00:21:59.440> complete<00:21:59.919> with<00:22:00.240> the 00:22:00.470 --> 00:22:00.480 align:start position:0% DNA of this yeast, complete with the 00:22:00.480 --> 00:22:02.549 align:start position:0% DNA of this yeast, complete with the protein<00:22:00.880> designs,<00:22:01.440> the<00:22:01.679> factory<00:22:02.000> workers, 00:22:02.549 --> 00:22:02.559 align:start position:0% protein designs, the factory workers, 00:22:02.559 --> 00:22:04.870 align:start position:0% protein designs, the factory workers, delivery<00:22:03.120> systems,<00:22:03.679> and<00:22:03.919> everything<00:22:04.400> else<00:22:04.640> it 00:22:04.870 --> 00:22:04.880 align:start position:0% delivery systems, and everything else it 00:22:04.880 --> 00:22:06.789 align:start position:0% delivery systems, and everything else it needs<00:22:05.120> to<00:22:05.360> function.<00:22:05.760> So,<00:22:06.000> this<00:22:06.159> proves<00:22:06.480> that 00:22:06.789 --> 00:22:06.799 align:start position:0% needs to function. So, this proves that 00:22:06.799 --> 00:22:10.149 align:start position:0% needs to function. So, this proves that EVO<00:22:07.280> 2<00:22:07.600> can<00:22:07.919> fluently<00:22:08.480> generate<00:22:08.960> DNA<00:22:09.600> for 00:22:10.149 --> 00:22:10.159 align:start position:0% EVO 2 can fluently generate DNA for 00:22:10.159 --> 00:22:12.149 align:start position:0% EVO 2 can fluently generate DNA for multiple<00:22:10.960> different<00:22:11.360> types<00:22:11.679> of<00:22:11.840> living 00:22:12.149 --> 00:22:12.159 align:start position:0% multiple different types of living 00:22:12.159 --> 00:22:15.110 align:start position:0% multiple different types of living things,<00:22:12.640> spanning<00:22:13.120> organels<00:22:13.679> to<00:22:14.000> bacteria<00:22:14.720> to 00:22:15.110 --> 00:22:15.120 align:start position:0% things, spanning organels to bacteria to 00:22:15.120 --> 00:22:17.669 align:start position:0% things, spanning organels to bacteria to even<00:22:15.600> yeast.<00:22:16.240> For<00:22:16.480> now,<00:22:16.799> you<00:22:17.039> might<00:22:17.280> have<00:22:17.440> a 00:22:17.669 --> 00:22:17.679 align:start position:0% even yeast. For now, you might have a 00:22:17.679 --> 00:22:19.990 align:start position:0% even yeast. For now, you might have a burning<00:22:18.080> question<00:22:18.320> in<00:22:18.559> your<00:22:18.720> head.<00:22:19.120> If<00:22:19.280> Evo<00:22:19.760> 2 00:22:19.990 --> 00:22:20.000 align:start position:0% burning question in your head. If Evo 2 00:22:20.000 --> 00:22:22.549 align:start position:0% burning question in your head. If Evo 2 can<00:22:20.320> generate<00:22:20.640> the<00:22:20.880> DNA<00:22:21.360> of<00:22:21.600> existing<00:22:22.159> stuff, 00:22:22.549 --> 00:22:22.559 align:start position:0% can generate the DNA of existing stuff, 00:22:22.559 --> 00:22:24.390 align:start position:0% can generate the DNA of existing stuff, can<00:22:22.720> we<00:22:22.799> get<00:22:22.960> it<00:22:23.120> to<00:22:23.280> generate<00:22:23.600> a<00:22:23.840> completely 00:22:24.390 --> 00:22:24.400 align:start position:0% can we get it to generate a completely 00:22:24.400 --> 00:22:26.789 align:start position:0% can we get it to generate a completely new<00:22:24.720> species?<00:22:25.520> What<00:22:25.679> if<00:22:25.840> we<00:22:26.080> got<00:22:26.240> it<00:22:26.480> to 00:22:26.789 --> 00:22:26.799 align:start position:0% new species? What if we got it to 00:22:26.799 --> 00:22:28.870 align:start position:0% new species? What if we got it to generate<00:22:27.200> a<00:22:27.440> new<00:22:27.600> type<00:22:27.760> of<00:22:27.919> mouse<00:22:28.320> or<00:22:28.640> even 00:22:28.870 --> 00:22:28.880 align:start position:0% generate a new type of mouse or even 00:22:28.880 --> 00:22:31.110 align:start position:0% generate a new type of mouse or even generate<00:22:29.360> Pokémon?<00:22:30.000> Or<00:22:30.159> even<00:22:30.400> crazier,<00:22:30.960> what 00:22:31.110 --> 00:22:31.120 align:start position:0% generate Pokémon? Or even crazier, what 00:22:31.120 --> 00:22:33.110 align:start position:0% generate Pokémon? Or even crazier, what if<00:22:31.200> we<00:22:31.360> got<00:22:31.440> it<00:22:31.600> to<00:22:31.760> generate<00:22:32.080> a<00:22:32.320> new<00:22:32.480> type<00:22:32.799> of 00:22:33.110 --> 00:22:33.120 align:start position:0% if we got it to generate a new type of 00:22:33.120 --> 00:22:35.590 align:start position:0% if we got it to generate a new type of human?<00:22:33.600> Well,<00:22:33.919> this<00:22:34.080> raises<00:22:34.640> a<00:22:34.960> ton<00:22:35.280> of 00:22:35.590 --> 00:22:35.600 align:start position:0% human? Well, this raises a ton of 00:22:35.600 --> 00:22:37.830 align:start position:0% human? Well, this raises a ton of different<00:22:36.000> bioeththic<00:22:36.799> and<00:22:37.039> biocurity 00:22:37.830 --> 00:22:37.840 align:start position:0% different bioeththic and biocurity 00:22:37.840 --> 00:22:39.909 align:start position:0% different bioeththic and biocurity concerns.<00:22:38.480> So,<00:22:38.640> at<00:22:38.720> least<00:22:38.960> for<00:22:39.200> now,<00:22:39.600> at<00:22:39.600> least 00:22:39.909 --> 00:22:39.919 align:start position:0% concerns. So, at least for now, at least 00:22:39.919 --> 00:22:41.669 align:start position:0% concerns. So, at least for now, at least the<00:22:40.159> researchers<00:22:40.559> weren't<00:22:40.960> crazy<00:22:41.280> enough<00:22:41.440> to 00:22:41.669 --> 00:22:41.679 align:start position:0% the researchers weren't crazy enough to 00:22:41.679 --> 00:22:43.750 align:start position:0% the researchers weren't crazy enough to proceed<00:22:42.000> with<00:22:42.240> that.<00:22:42.799> But<00:22:43.039> speaking<00:22:43.360> of 00:22:43.750 --> 00:22:43.760 align:start position:0% proceed with that. But speaking of 00:22:43.760 --> 00:22:45.830 align:start position:0% proceed with that. But speaking of security,<00:22:44.400> here's<00:22:44.720> another<00:22:44.960> concern.<00:22:45.600> If 00:22:45.830 --> 00:22:45.840 align:start position:0% security, here's another concern. If 00:22:45.840 --> 00:22:47.990 align:start position:0% security, here's another concern. If this<00:22:46.080> can<00:22:46.240> generate<00:22:46.720> anything,<00:22:47.360> couldn't<00:22:47.679> we 00:22:47.990 --> 00:22:48.000 align:start position:0% this can generate anything, couldn't we 00:22:48.000 --> 00:22:49.590 align:start position:0% this can generate anything, couldn't we prompt<00:22:48.240> it<00:22:48.480> to<00:22:48.640> generate<00:22:48.960> a<00:22:49.200> highly 00:22:49.590 --> 00:22:49.600 align:start position:0% prompt it to generate a highly 00:22:49.600 --> 00:22:51.830 align:start position:0% prompt it to generate a highly contagious<00:22:50.320> human<00:22:50.640> virus?<00:22:51.360> Well,<00:22:51.520> the<00:22:51.679> good 00:22:51.830 --> 00:22:51.840 align:start position:0% contagious human virus? Well, the good 00:22:51.840 --> 00:22:54.230 align:start position:0% contagious human virus? Well, the good news<00:22:52.080> is<00:22:52.400> the<00:22:52.720> researchers<00:22:53.360> also<00:22:53.840> thought<00:22:54.000> of 00:22:54.230 --> 00:22:54.240 align:start position:0% news is the researchers also thought of 00:22:54.240 --> 00:22:56.630 align:start position:0% news is the researchers also thought of this<00:22:54.559> beforehand<00:22:55.440> before<00:22:55.840> they<00:22:56.080> trained<00:22:56.400> the 00:22:56.630 --> 00:22:56.640 align:start position:0% this beforehand before they trained the 00:22:56.640 --> 00:22:59.270 align:start position:0% this beforehand before they trained the model.<00:22:57.120> So,<00:22:57.280> they<00:22:57.520> made<00:22:57.760> a<00:22:58.159> very<00:22:58.640> important 00:22:59.270 --> 00:22:59.280 align:start position:0% model. So, they made a very important 00:22:59.280 --> 00:23:01.590 align:start position:0% model. So, they made a very important safety<00:22:59.760> decision<00:23:00.400> right<00:23:00.640> from<00:23:00.880> the<00:23:01.120> start. 00:23:01.590 --> 00:23:01.600 align:start position:0% safety decision right from the start. 00:23:01.600 --> 00:23:04.789 align:start position:0% safety decision right from the start. You<00:23:01.840> see,<00:23:02.080> when<00:23:02.400> they<00:23:02.640> fed<00:23:02.960> it<00:23:03.120> this<00:23:03.440> huge<00:23:04.000> DNA 00:23:04.789 --> 00:23:04.799 align:start position:0% You see, when they fed it this huge DNA 00:23:04.799 --> 00:23:08.070 align:start position:0% You see, when they fed it this huge DNA data<00:23:05.120> set<00:23:05.520> of<00:23:06.000> 9<00:23:06.320> trillion<00:23:06.880> base<00:23:07.200> pairs,<00:23:07.760> they 00:23:08.070 --> 00:23:08.080 align:start position:0% data set of 9 trillion base pairs, they 00:23:08.080 --> 00:23:11.350 align:start position:0% data set of 9 trillion base pairs, they intentionally<00:23:09.200> excluded<00:23:10.080> any<00:23:10.400> sequences<00:23:10.960> of 00:23:11.350 --> 00:23:11.360 align:start position:0% intentionally excluded any sequences of 00:23:11.360 --> 00:23:14.310 align:start position:0% intentionally excluded any sequences of ukarotic<00:23:12.080> viruses,<00:23:12.960> meaning<00:23:13.280> any<00:23:13.520> virus<00:23:14.000> that 00:23:14.310 --> 00:23:14.320 align:start position:0% ukarotic viruses, meaning any virus that 00:23:14.320 --> 00:23:16.630 align:start position:0% ukarotic viruses, meaning any virus that can<00:23:14.480> affect<00:23:14.880> humans,<00:23:15.360> animals,<00:23:15.760> or<00:23:16.080> plants. 00:23:16.630 --> 00:23:16.640 align:start position:0% can affect humans, animals, or plants. 00:23:16.640 --> 00:23:18.789 align:start position:0% can affect humans, animals, or plants. So,<00:23:16.880> this<00:23:17.200> includes<00:23:17.600> the<00:23:17.840> most<00:23:18.240> dangerous 00:23:18.789 --> 00:23:18.799 align:start position:0% So, this includes the most dangerous 00:23:18.799 --> 00:23:21.830 align:start position:0% So, this includes the most dangerous pathogens<00:23:19.440> listed<00:23:19.840> in<00:23:20.080> the<00:23:20.400> USDA<00:23:21.440> select 00:23:21.830 --> 00:23:21.840 align:start position:0% pathogens listed in the USDA select 00:23:21.840 --> 00:23:24.310 align:start position:0% pathogens listed in the USDA select agents<00:23:22.320> and<00:23:22.559> toxins<00:23:23.120> list.<00:23:23.600> These<00:23:23.840> are<00:23:24.000> like 00:23:24.310 --> 00:23:24.320 align:start position:0% agents and toxins list. These are like 00:23:24.320 --> 00:23:26.230 align:start position:0% agents and toxins list. These are like really<00:23:24.640> contagious<00:23:25.200> things<00:23:25.360> that<00:23:25.600> can<00:23:25.840> kill<00:23:26.000> a 00:23:26.230 --> 00:23:26.240 align:start position:0% really contagious things that can kill a 00:23:26.240 --> 00:23:28.149 align:start position:0% really contagious things that can kill a ton<00:23:26.400> of<00:23:26.480> people.<00:23:26.880> So,<00:23:27.120> all<00:23:27.360> this<00:23:27.520> data<00:23:27.919> was 00:23:28.149 --> 00:23:28.159 align:start position:0% ton of people. So, all this data was 00:23:28.159 --> 00:23:30.230 align:start position:0% ton of people. So, all this data was excluded<00:23:28.799> from<00:23:29.120> its<00:23:29.440> data<00:23:29.679> set.<00:23:29.919> In<00:23:30.159> other 00:23:30.230 --> 00:23:30.240 align:start position:0% excluded from its data set. In other 00:23:30.240 --> 00:23:33.110 align:start position:0% excluded from its data set. In other words,<00:23:30.559> the<00:23:30.799> AI<00:23:31.200> was<00:23:31.440> never<00:23:31.840> allowed<00:23:32.240> to<00:23:32.559> learn 00:23:33.110 --> 00:23:33.120 align:start position:0% words, the AI was never allowed to learn 00:23:33.120 --> 00:23:34.950 align:start position:0% words, the AI was never allowed to learn what<00:23:33.360> these<00:23:33.679> viruses<00:23:34.159> even<00:23:34.480> looked<00:23:34.640> like. 00:23:34.950 --> 00:23:34.960 align:start position:0% what these viruses even looked like. 00:23:34.960 --> 00:23:37.350 align:start position:0% what these viruses even looked like. It's<00:23:35.200> like<00:23:35.440> if<00:23:35.679> you<00:23:35.919> were<00:23:36.240> being<00:23:36.480> trained<00:23:36.880> on<00:23:37.120> a 00:23:37.350 --> 00:23:37.360 align:start position:0% It's like if you were being trained on a 00:23:37.360 --> 00:23:39.669 align:start position:0% It's like if you were being trained on a human<00:23:37.600> language,<00:23:38.240> but<00:23:38.480> all<00:23:38.720> the<00:23:38.880> swear<00:23:39.200> words 00:23:39.669 --> 00:23:39.679 align:start position:0% human language, but all the swear words 00:23:39.679 --> 00:23:41.510 align:start position:0% human language, but all the swear words were<00:23:40.000> omitted.<00:23:40.559> You<00:23:40.720> would<00:23:40.880> have<00:23:41.039> no<00:23:41.280> idea 00:23:41.510 --> 00:23:41.520 align:start position:0% were omitted. You would have no idea 00:23:41.520 --> 00:23:43.590 align:start position:0% were omitted. You would have no idea what<00:23:41.760> a<00:23:41.919> swear<00:23:42.240> word<00:23:42.559> even<00:23:42.799> is.<00:23:43.280> Now, 00:23:43.590 --> 00:23:43.600 align:start position:0% what a swear word even is. Now, 00:23:43.600 --> 00:23:45.190 align:start position:0% what a swear word even is. Now, excluding<00:23:44.000> this<00:23:44.240> training<00:23:44.559> data<00:23:44.799> is<00:23:45.039> one 00:23:45.190 --> 00:23:45.200 align:start position:0% excluding this training data is one 00:23:45.200 --> 00:23:47.270 align:start position:0% excluding this training data is one thing.<00:23:45.600> How<00:23:45.760> do<00:23:45.919> they<00:23:46.080> verify<00:23:46.559> that<00:23:46.960> this 00:23:47.270 --> 00:23:47.280 align:start position:0% thing. How do they verify that this 00:23:47.280 --> 00:23:49.590 align:start position:0% thing. How do they verify that this safety<00:23:47.600> filter<00:23:48.080> actually<00:23:48.400> works?<00:23:49.039> That<00:23:49.280> it 00:23:49.590 --> 00:23:49.600 align:start position:0% safety filter actually works? That it 00:23:49.600 --> 00:23:52.470 align:start position:0% safety filter actually works? That it can't<00:23:50.000> generate<00:23:50.480> viruses.<00:23:51.360> So,<00:23:51.600> they<00:23:51.919> used<00:23:52.240> a 00:23:52.470 --> 00:23:52.480 align:start position:0% can't generate viruses. So, they used a 00:23:52.480 --> 00:23:54.549 align:start position:0% can't generate viruses. So, they used a really<00:23:52.799> standard<00:23:53.200> AI<00:23:53.600> metric<00:23:54.159> called 00:23:54.549 --> 00:23:54.559 align:start position:0% really standard AI metric called 00:23:54.559 --> 00:23:56.470 align:start position:0% really standard AI metric called perplexity.<00:23:55.440> No,<00:23:55.760> it's<00:23:56.000> not<00:23:56.159> this 00:23:56.470 --> 00:23:56.480 align:start position:0% perplexity. No, it's not this 00:23:56.480 --> 00:23:58.789 align:start position:0% perplexity. No, it's not this perplexity.<00:23:57.200> Perplexity<00:23:57.840> is<00:23:58.080> actually<00:23:58.480> a 00:23:58.789 --> 00:23:58.799 align:start position:0% perplexity. Perplexity is actually a 00:23:58.799 --> 00:24:01.190 align:start position:0% perplexity. Perplexity is actually a metric<00:23:59.280> which<00:23:59.520> looks<00:23:59.760> at<00:24:00.000> how<00:24:00.320> confused<00:24:00.960> a 00:24:01.190 --> 00:24:01.200 align:start position:0% metric which looks at how confused a 00:24:01.200 --> 00:24:03.029 align:start position:0% metric which looks at how confused a model<00:24:01.520> is<00:24:01.679> when<00:24:01.919> it<00:24:02.159> encounters<00:24:02.720> something 00:24:03.029 --> 00:24:03.039 align:start position:0% model is when it encounters something 00:24:03.039 --> 00:24:05.270 align:start position:0% model is when it encounters something that<00:24:03.360> it<00:24:03.520> hasn't<00:24:03.760> seen<00:24:04.000> before.<00:24:04.640> So<00:24:04.880> after 00:24:05.270 --> 00:24:05.280 align:start position:0% that it hasn't seen before. So after 00:24:05.280 --> 00:24:07.590 align:start position:0% that it hasn't seen before. So after training<00:24:05.679> EVO<00:24:06.159> 2,<00:24:06.320> the<00:24:06.480> researchers<00:24:07.120> also 00:24:07.590 --> 00:24:07.600 align:start position:0% training EVO 2, the researchers also 00:24:07.600 --> 00:24:10.630 align:start position:0% training EVO 2, the researchers also gave<00:24:07.840> it<00:24:08.080> the<00:24:08.320> DNA<00:24:08.960> of<00:24:09.200> some<00:24:09.600> human<00:24:09.919> viruses 00:24:10.630 --> 00:24:10.640 align:start position:0% gave it the DNA of some human viruses 00:24:10.640 --> 00:24:13.269 align:start position:0% gave it the DNA of some human viruses from<00:24:11.279> this<00:24:11.600> list<00:24:12.159> and<00:24:12.400> basically<00:24:12.960> its 00:24:13.269 --> 00:24:13.279 align:start position:0% from this list and basically its 00:24:13.279 --> 00:24:15.669 align:start position:0% from this list and basically its perplexity<00:24:14.240> soared.<00:24:14.960> It<00:24:15.200> was<00:24:15.360> really 00:24:15.669 --> 00:24:15.679 align:start position:0% perplexity soared. It was really 00:24:15.679 --> 00:24:17.990 align:start position:0% perplexity soared. It was really confused.<00:24:16.320> It<00:24:16.559> had<00:24:16.720> no<00:24:17.039> understanding<00:24:17.679> of 00:24:17.990 --> 00:24:18.000 align:start position:0% confused. It had no understanding of 00:24:18.000 --> 00:24:19.990 align:start position:0% confused. It had no understanding of this,<00:24:18.480> no<00:24:18.720> internal<00:24:19.200> framework<00:24:19.600> to<00:24:19.840> make 00:24:19.990 --> 00:24:20.000 align:start position:0% this, no internal framework to make 00:24:20.000 --> 00:24:22.630 align:start position:0% this, no internal framework to make sense<00:24:20.240> of<00:24:20.640> what<00:24:20.880> these<00:24:21.200> sequences<00:24:21.919> are.<00:24:22.400> Now 00:24:22.630 --> 00:24:22.640 align:start position:0% sense of what these sequences are. Now 00:24:22.640 --> 00:24:24.310 align:start position:0% sense of what these sequences are. Now the<00:24:22.799> researchers<00:24:23.279> went<00:24:23.440> a<00:24:23.679> step<00:24:23.919> further. 00:24:24.310 --> 00:24:24.320 align:start position:0% the researchers went a step further. 00:24:24.320 --> 00:24:26.549 align:start position:0% the researchers went a step further. They<00:24:24.559> also<00:24:24.799> tried<00:24:25.120> to<00:24:25.279> get<00:24:25.440> it<00:24:25.679> to<00:24:26.000> generate 00:24:26.549 --> 00:24:26.559 align:start position:0% They also tried to get it to generate 00:24:26.559 --> 00:24:29.430 align:start position:0% They also tried to get it to generate the<00:24:26.799> DNA<00:24:27.200> of<00:24:27.440> some<00:24:27.840> dangerous<00:24:28.400> human<00:24:28.720> viruses 00:24:29.430 --> 00:24:29.440 align:start position:0% the DNA of some dangerous human viruses 00:24:29.440 --> 00:24:31.990 align:start position:0% the DNA of some dangerous human viruses and<00:24:29.679> see<00:24:30.000> what<00:24:30.240> it<00:24:30.480> would<00:24:30.720> do.<00:24:31.440> Thankfully, 00:24:31.990 --> 00:24:32.000 align:start position:0% and see what it would do. Thankfully, 00:24:32.000 --> 00:24:34.310 align:start position:0% and see what it would do. Thankfully, the<00:24:32.240> model<00:24:32.559> failed<00:24:33.039> completely.<00:24:33.919> Now,<00:24:34.080> it 00:24:34.310 --> 00:24:34.320 align:start position:0% the model failed completely. Now, it 00:24:34.320 --> 00:24:35.909 align:start position:0% the model failed completely. Now, it still<00:24:34.480> generated<00:24:35.039> something,<00:24:35.360> but<00:24:35.679> because 00:24:35.909 --> 00:24:35.919 align:start position:0% still generated something, but because 00:24:35.919 --> 00:24:38.470 align:start position:0% still generated something, but because it<00:24:36.240> didn't<00:24:36.400> have<00:24:36.640> this<00:24:37.039> virus<00:24:37.600> data<00:24:38.080> during 00:24:38.470 --> 00:24:38.480 align:start position:0% it didn't have this virus data during 00:24:38.480 --> 00:24:40.549 align:start position:0% it didn't have this virus data during training,<00:24:39.039> it<00:24:39.200> just<00:24:39.360> ended<00:24:39.600> up<00:24:39.840> hallucinating 00:24:40.549 --> 00:24:40.559 align:start position:0% training, it just ended up hallucinating 00:24:40.559 --> 00:24:42.710 align:start position:0% training, it just ended up hallucinating and<00:24:40.880> generating<00:24:41.360> gibberish.<00:24:42.159> It<00:24:42.400> was<00:24:42.480> not 00:24:42.710 --> 00:24:42.720 align:start position:0% and generating gibberish. It was not 00:24:42.720 --> 00:24:45.909 align:start position:0% and generating gibberish. It was not able<00:24:43.039> to<00:24:43.600> successfully<00:24:44.400> generate<00:24:44.880> the<00:24:45.120> DNA<00:24:45.600> of 00:24:45.909 --> 00:24:45.919 align:start position:0% able to successfully generate the DNA of 00:24:45.919 --> 00:24:48.549 align:start position:0% able to successfully generate the DNA of a<00:24:46.159> legit<00:24:46.640> human<00:24:47.039> virus.<00:24:47.679> Thank<00:24:48.000> goodness<00:24:48.320> for 00:24:48.549 --> 00:24:48.559 align:start position:0% a legit human virus. Thank goodness for 00:24:48.559 --> 00:24:50.950 align:start position:0% a legit human virus. Thank goodness for that.<00:24:49.120> Now,<00:24:49.520> the<00:24:49.840> really<00:24:50.159> cool<00:24:50.400> thing<00:24:50.640> is 00:24:50.950 --> 00:24:50.960 align:start position:0% that. Now, the really cool thing is 00:24:50.960 --> 00:24:53.510 align:start position:0% that. Now, the really cool thing is they've<00:24:51.440> actually<00:24:52.000> opensourced<00:24:52.960> everything. 00:24:53.510 --> 00:24:53.520 align:start position:0% they've actually opensourced everything. 00:24:53.520 --> 00:24:55.510 align:start position:0% they've actually opensourced everything. So<00:24:53.840> on<00:24:54.000> this<00:24:54.240> GitHub<00:24:54.640> repo,<00:24:55.039> which<00:24:55.200> I'll<00:24:55.360> link 00:24:55.510 --> 00:24:55.520 align:start position:0% So on this GitHub repo, which I'll link 00:24:55.520 --> 00:24:57.110 align:start position:0% So on this GitHub repo, which I'll link to<00:24:55.679> in<00:24:55.840> the<00:24:56.000> description<00:24:56.320> below,<00:24:56.799> you<00:24:56.960> can 00:24:57.110 --> 00:24:57.120 align:start position:0% to in the description below, you can 00:24:57.120 --> 00:24:59.830 align:start position:0% to in the description below, you can actually<00:24:57.520> download<00:24:57.919> and<00:24:58.240> use<00:24:58.559> EVO<00:24:59.120> 2.<00:24:59.440> So<00:24:59.679> here 00:24:59.830 --> 00:24:59.840 align:start position:0% actually download and use EVO 2. So here 00:24:59.840 --> 00:25:01.909 align:start position:0% actually download and use EVO 2. So here it<00:25:00.080> contains<00:25:00.400> all<00:25:00.559> the<00:25:00.720> instructions<00:25:01.360> on<00:25:01.679> how 00:25:01.909 --> 00:25:01.919 align:start position:0% it contains all the instructions on how 00:25:01.919 --> 00:25:04.310 align:start position:0% it contains all the instructions on how to<00:25:02.080> set<00:25:02.240> this<00:25:02.480> up.<00:25:02.799> Now,<00:25:03.120> in<00:25:03.279> addition<00:25:03.679> to<00:25:04.080> just 00:25:04.310 --> 00:25:04.320 align:start position:0% to set this up. Now, in addition to just 00:25:04.320 --> 00:25:06.070 align:start position:0% to set this up. Now, in addition to just releasing<00:25:04.720> the<00:25:04.880> models,<00:25:05.279> they've<00:25:05.600> also 00:25:06.070 --> 00:25:06.080 align:start position:0% releasing the models, they've also 00:25:06.080 --> 00:25:08.789 align:start position:0% releasing the models, they've also released<00:25:06.720> the<00:25:07.039> data<00:25:07.360> set<00:25:07.840> excluding<00:25:08.480> human 00:25:08.789 --> 00:25:08.799 align:start position:0% released the data set excluding human 00:25:08.799 --> 00:25:11.510 align:start position:0% released the data set excluding human viruses,<00:25:09.440> plus<00:25:09.760> also<00:25:10.400> the<00:25:10.799> training<00:25:11.200> and 00:25:11.510 --> 00:25:11.520 align:start position:0% viruses, plus also the training and 00:25:11.520 --> 00:25:13.909 align:start position:0% viruses, plus also the training and fine-tuning<00:25:12.240> code.<00:25:12.799> So<00:25:13.039> everything<00:25:13.279> is<00:25:13.520> here. 00:25:13.909 --> 00:25:13.919 align:start position:0% fine-tuning code. So everything is here. 00:25:13.919 --> 00:25:15.269 align:start position:0% fine-tuning code. So everything is here. If<00:25:14.159> you<00:25:14.320> have<00:25:14.480> enough<00:25:14.720> compute<00:25:15.039> and 00:25:15.269 --> 00:25:15.279 align:start position:0% If you have enough compute and 00:25:15.279 --> 00:25:16.630 align:start position:0% If you have enough compute and infrastructure<00:25:15.840> at<00:25:16.080> home,<00:25:16.320> you<00:25:16.480> could 00:25:16.630 --> 00:25:16.640 align:start position:0% infrastructure at home, you could 00:25:16.640 --> 00:25:18.310 align:start position:0% infrastructure at home, you could potentially<00:25:17.039> download<00:25:17.520> everything<00:25:18.080> and 00:25:18.310 --> 00:25:18.320 align:start position:0% potentially download everything and 00:25:18.320 --> 00:25:20.710 align:start position:0% potentially download everything and train<00:25:18.640> or<00:25:18.960> fine-tune<00:25:19.600> your<00:25:19.919> own<00:25:20.080> AI<00:25:20.480> model 00:25:20.710 --> 00:25:20.720 align:start position:0% train or fine-tune your own AI model 00:25:20.720 --> 00:25:23.190 align:start position:0% train or fine-tune your own AI model that<00:25:21.039> understands<00:25:21.440> and<00:25:21.760> can<00:25:21.919> generate<00:25:22.480> DNA. 00:25:23.190 --> 00:25:23.200 align:start position:0% that understands and can generate DNA. 00:25:23.200 --> 00:25:25.190 align:start position:0% that understands and can generate DNA. Now,<00:25:23.440> if<00:25:23.600> I<00:25:23.679> click<00:25:23.840> on<00:25:24.000> hugging<00:25:24.400> face<00:25:24.720> for<00:25:24.960> the 00:25:25.190 --> 00:25:25.200 align:start position:0% Now, if I click on hugging face for the 00:25:25.200 --> 00:25:28.310 align:start position:0% Now, if I click on hugging face for the 40<00:25:25.520> billion<00:25:25.840> parameter<00:25:26.320> model,<00:25:26.640> it<00:25:26.880> is<00:25:27.200> 82<00:25:27.760> GB 00:25:28.310 --> 00:25:28.320 align:start position:0% 40 billion parameter model, it is 82 GB 00:25:28.320 --> 00:25:30.390 align:start position:0% 40 billion parameter model, it is 82 GB in<00:25:28.559> size.<00:25:29.039> You'll<00:25:29.279> need<00:25:29.440> quite<00:25:29.679> a<00:25:29.919> high-end 00:25:30.390 --> 00:25:30.400 align:start position:0% in size. You'll need quite a high-end 00:25:30.400 --> 00:25:33.590 align:start position:0% in size. You'll need quite a high-end GPU<00:25:30.880> to<00:25:31.120> run<00:25:31.279> this.<00:25:31.760> Anyways,<00:25:32.240> back<00:25:32.400> to<00:25:32.640> EVO<00:25:33.120> 2. 00:25:33.590 --> 00:25:33.600 align:start position:0% GPU to run this. Anyways, back to EVO 2. 00:25:33.600 --> 00:25:35.590 align:start position:0% GPU to run this. Anyways, back to EVO 2. This<00:25:33.840> actually<00:25:34.240> opens<00:25:34.559> up<00:25:34.799> a<00:25:35.039> ton<00:25:35.360> of 00:25:35.590 --> 00:25:35.600 align:start position:0% This actually opens up a ton of 00:25:35.600 --> 00:25:37.350 align:start position:0% This actually opens up a ton of possibilities,<00:25:36.400> and<00:25:36.640> this<00:25:36.799> is<00:25:36.960> just<00:25:37.120> the 00:25:37.350 --> 00:25:37.360 align:start position:0% possibilities, and this is just the 00:25:37.360 --> 00:25:39.350 align:start position:0% possibilities, and this is just the beginning.<00:25:37.919> You<00:25:38.159> see,<00:25:38.400> if<00:25:38.559> we<00:25:38.640> have<00:25:38.720> an<00:25:38.960> AI 00:25:39.350 --> 00:25:39.360 align:start position:0% beginning. You see, if we have an AI 00:25:39.360 --> 00:25:41.430 align:start position:0% beginning. You see, if we have an AI that<00:25:39.600> can<00:25:39.760> understand<00:25:40.240> DNA<00:25:40.640> and<00:25:40.960> also 00:25:41.430 --> 00:25:41.440 align:start position:0% that can understand DNA and also 00:25:41.440 --> 00:25:44.070 align:start position:0% that can understand DNA and also generate<00:25:42.080> new<00:25:42.400> DNA,<00:25:43.039> we<00:25:43.200> could<00:25:43.440> use<00:25:43.600> it<00:25:43.760> to 00:25:44.070 --> 00:25:44.080 align:start position:0% generate new DNA, we could use it to 00:25:44.080 --> 00:25:46.950 align:start position:0% generate new DNA, we could use it to design<00:25:44.640> better<00:25:45.120> genetically<00:25:45.919> modified<00:25:46.400> crops 00:25:46.950 --> 00:25:46.960 align:start position:0% design better genetically modified crops 00:25:46.960 --> 00:25:48.470 align:start position:0% design better genetically modified crops to<00:25:47.279> create<00:25:47.520> plants<00:25:47.840> that<00:25:48.000> are<00:25:48.159> more 00:25:48.470 --> 00:25:48.480 align:start position:0% to create plants that are more 00:25:48.480 --> 00:25:51.269 align:start position:0% to create plants that are more nutritious<00:25:49.120> or<00:25:49.440> maybe<00:25:49.840> more<00:25:50.159> resilient<00:25:50.960> or 00:25:51.269 --> 00:25:51.279 align:start position:0% nutritious or maybe more resilient or 00:25:51.279 --> 00:25:53.830 align:start position:0% nutritious or maybe more resilient or more<00:25:51.679> efficient<00:25:52.080> at<00:25:52.400> producing<00:25:53.039> energy<00:25:53.520> for 00:25:53.830 --> 00:25:53.840 align:start position:0% more efficient at producing energy for 00:25:53.840 --> 00:25:56.070 align:start position:0% more efficient at producing energy for bofuels.<00:25:54.720> The<00:25:54.960> implications<00:25:55.440> for<00:25:55.760> food 00:25:56.070 --> 00:25:56.080 align:start position:0% bofuels. The implications for food 00:25:56.080 --> 00:25:58.149 align:start position:0% bofuels. The implications for food security<00:25:56.400> and<00:25:56.720> sustainable<00:25:57.279> energy<00:25:57.760> are 00:25:58.149 --> 00:25:58.159 align:start position:0% security and sustainable energy are 00:25:58.159 --> 00:25:59.990 align:start position:0% security and sustainable energy are huge.<00:25:58.799> But<00:25:58.880> it<00:25:59.120> doesn't<00:25:59.279> just<00:25:59.520> stop<00:25:59.760> at 00:25:59.990 --> 00:26:00.000 align:start position:0% huge. But it doesn't just stop at 00:26:00.000 --> 00:26:01.990 align:start position:0% huge. But it doesn't just stop at plants.<00:26:00.559> If<00:26:00.799> we<00:26:00.960> were<00:26:01.120> to<00:26:01.360> really<00:26:01.679> stretch 00:26:01.990 --> 00:26:02.000 align:start position:0% plants. If we were to really stretch 00:26:02.000 --> 00:26:04.149 align:start position:0% plants. If we were to really stretch this<00:26:02.240> out,<00:26:02.559> maybe<00:26:02.799> we<00:26:03.120> could<00:26:03.279> also<00:26:03.679> get<00:26:03.919> this 00:26:04.149 --> 00:26:04.159 align:start position:0% this out, maybe we could also get this 00:26:04.159 --> 00:26:07.190 align:start position:0% this out, maybe we could also get this to<00:26:04.480> analyze<00:26:05.120> the<00:26:05.440> human<00:26:05.760> genome<00:26:06.320> at<00:26:06.559> a<00:26:06.720> level 00:26:07.190 --> 00:26:07.200 align:start position:0% to analyze the human genome at a level 00:26:07.200 --> 00:26:09.110 align:start position:0% to analyze the human genome at a level much<00:26:07.520> deeper<00:26:07.919> than<00:26:08.159> we<00:26:08.480> currently 00:26:09.110 --> 00:26:09.120 align:start position:0% much deeper than we currently 00:26:09.120 --> 00:26:11.269 align:start position:0% much deeper than we currently understand.<00:26:10.000> Maybe<00:26:10.240> it<00:26:10.480> can<00:26:10.640> help<00:26:10.799> us<00:26:11.039> figure 00:26:11.269 --> 00:26:11.279 align:start position:0% understand. Maybe it can help us figure 00:26:11.279 --> 00:26:13.430 align:start position:0% understand. Maybe it can help us figure out<00:26:11.440> which<00:26:11.760> parts<00:26:12.000> of<00:26:12.159> our<00:26:12.320> DNA<00:26:12.720> we<00:26:12.960> can<00:26:13.120> tweak 00:26:13.430 --> 00:26:13.440 align:start position:0% out which parts of our DNA we can tweak 00:26:13.440 --> 00:26:15.510 align:start position:0% out which parts of our DNA we can tweak to<00:26:13.679> make<00:26:13.840> us<00:26:14.080> better.<00:26:14.559> For<00:26:14.640> example,<00:26:15.200> changing 00:26:15.510 --> 00:26:15.520 align:start position:0% to make us better. For example, changing 00:26:15.520 --> 00:26:17.029 align:start position:0% to make us better. For example, changing a<00:26:15.679> certain<00:26:15.919> gene<00:26:16.240> could<00:26:16.400> make<00:26:16.559> us<00:26:16.799> more 00:26:17.029 --> 00:26:17.039 align:start position:0% a certain gene could make us more 00:26:17.039 --> 00:26:18.950 align:start position:0% a certain gene could make us more productive.<00:26:17.679> Changing<00:26:18.080> another<00:26:18.400> gene<00:26:18.720> could 00:26:18.950 --> 00:26:18.960 align:start position:0% productive. Changing another gene could 00:26:18.960 --> 00:26:21.750 align:start position:0% productive. Changing another gene could make<00:26:19.120> us<00:26:19.520> stronger<00:26:20.159> or<00:26:20.480> smarter,<00:26:21.120> or<00:26:21.360> changing 00:26:21.750 --> 00:26:21.760 align:start position:0% make us stronger or smarter, or changing 00:26:21.760 --> 00:26:23.750 align:start position:0% make us stronger or smarter, or changing another<00:26:22.159> gene<00:26:22.400> could<00:26:22.640> make<00:26:22.799> us<00:26:22.960> immune<00:26:23.360> to<00:26:23.600> a 00:26:23.750 --> 00:26:23.760 align:start position:0% another gene could make us immune to a 00:26:23.760 --> 00:26:25.750 align:start position:0% another gene could make us immune to a certain<00:26:24.080> disease.<00:26:24.720> And<00:26:24.880> this<00:26:25.039> is<00:26:25.200> also<00:26:25.520> great 00:26:25.750 --> 00:26:25.760 align:start position:0% certain disease. And this is also great 00:26:25.760 --> 00:26:28.149 align:start position:0% certain disease. And this is also great for<00:26:26.000> personalized<00:26:26.799> medicine.<00:26:27.520> If<00:26:27.760> you<00:26:27.919> have<00:26:28.000> a 00:26:28.149 --> 00:26:28.159 align:start position:0% for personalized medicine. If you have a 00:26:28.159 --> 00:26:30.310 align:start position:0% for personalized medicine. If you have a patient<00:26:28.480> with<00:26:28.720> a<00:26:28.880> disease<00:26:29.440> or<00:26:29.679> something<00:26:29.919> that 00:26:30.310 --> 00:26:30.320 align:start position:0% patient with a disease or something that 00:26:30.320 --> 00:26:32.390 align:start position:0% patient with a disease or something that doctors<00:26:30.799> cannot<00:26:31.200> diagnose,<00:26:31.840> you<00:26:32.080> can<00:26:32.159> just 00:26:32.390 --> 00:26:32.400 align:start position:0% doctors cannot diagnose, you can just 00:26:32.400 --> 00:26:34.549 align:start position:0% doctors cannot diagnose, you can just extract<00:26:32.799> their<00:26:33.039> DNA<00:26:33.520> and<00:26:33.760> then<00:26:34.080> plug<00:26:34.400> it 00:26:34.549 --> 00:26:34.559 align:start position:0% extract their DNA and then plug it 00:26:34.559 --> 00:26:36.470 align:start position:0% extract their DNA and then plug it through<00:26:34.799> this<00:26:34.960> AI<00:26:35.440> to<00:26:35.679> potentially<00:26:36.240> figure 00:26:36.470 --> 00:26:36.480 align:start position:0% through this AI to potentially figure 00:26:36.480 --> 00:26:38.950 align:start position:0% through this AI to potentially figure out<00:26:36.799> what's<00:26:37.200> going<00:26:37.360> on.<00:26:37.840> Each<00:26:38.159> person's<00:26:38.559> DNA 00:26:38.950 --> 00:26:38.960 align:start position:0% out what's going on. Each person's DNA 00:26:38.960 --> 00:26:41.510 align:start position:0% out what's going on. Each person's DNA is<00:26:39.200> slightly<00:26:39.679> different.<00:26:40.159> So,<00:26:40.400> with<00:26:40.720> this<00:26:40.880> AI, 00:26:41.510 --> 00:26:41.520 align:start position:0% is slightly different. So, with this AI, 00:26:41.520 --> 00:26:44.310 align:start position:0% is slightly different. So, with this AI, we<00:26:41.760> can<00:26:41.919> use<00:26:42.080> it<00:26:42.320> to<00:26:42.559> help<00:26:42.880> prescribe<00:26:43.600> bespoke 00:26:44.310 --> 00:26:44.320 align:start position:0% we can use it to help prescribe bespoke 00:26:44.320 --> 00:26:46.470 align:start position:0% we can use it to help prescribe bespoke personalized<00:26:45.200> healthcare.<00:26:46.000> Or<00:26:46.159> here's 00:26:46.470 --> 00:26:46.480 align:start position:0% personalized healthcare. Or here's 00:26:46.480 --> 00:26:48.230 align:start position:0% personalized healthcare. Or here's another<00:26:46.799> crazy<00:26:47.120> thing.<00:26:47.440> If<00:26:47.600> you're<00:26:47.840> expecting 00:26:48.230 --> 00:26:48.240 align:start position:0% another crazy thing. If you're expecting 00:26:48.240 --> 00:26:50.789 align:start position:0% another crazy thing. If you're expecting a<00:26:48.559> baby,<00:26:49.200> maybe<00:26:49.440> you<00:26:49.679> can<00:26:49.840> take<00:26:50.000> its<00:26:50.159> DNA<00:26:50.559> and 00:26:50.789 --> 00:26:50.799 align:start position:0% a baby, maybe you can take its DNA and 00:26:50.799 --> 00:26:53.190 align:start position:0% a baby, maybe you can take its DNA and plug<00:26:50.960> it<00:26:51.120> through<00:26:51.360> this<00:26:51.600> AI<00:26:52.080> to<00:26:52.320> detect<00:26:52.720> a<00:26:53.039> ton 00:26:53.190 --> 00:26:53.200 align:start position:0% plug it through this AI to detect a ton 00:26:53.200 --> 00:26:55.430 align:start position:0% plug it through this AI to detect a ton of<00:26:53.360> things.<00:26:53.919> for<00:26:54.080> example,<00:26:54.559> your<00:26:54.799> baby's<00:26:55.200> hair 00:26:55.430 --> 00:26:55.440 align:start position:0% of things. for example, your baby's hair 00:26:55.440 --> 00:26:58.149 align:start position:0% of things. for example, your baby's hair color,<00:26:55.840> eye<00:26:56.159> color,<00:26:56.559> athleticism,<00:26:57.600> risk<00:26:57.919> of 00:26:58.149 --> 00:26:58.159 align:start position:0% color, eye color, athleticism, risk of 00:26:58.159 --> 00:27:00.390 align:start position:0% color, eye color, athleticism, risk of disease,<00:26:58.799> even<00:26:59.120> personality,<00:26:59.679> and<00:26:59.919> a<00:27:00.080> ton<00:27:00.240> of 00:27:00.390 --> 00:27:00.400 align:start position:0% disease, even personality, and a ton of 00:27:00.400 --> 00:27:03.029 align:start position:0% disease, even personality, and a ton of other<00:27:00.559> stuff,<00:27:01.039> even<00:27:01.360> before<00:27:01.760> it's<00:27:02.159> born.<00:27:02.720> Now, 00:27:03.029 --> 00:27:03.039 align:start position:0% other stuff, even before it's born. Now, 00:27:03.039 --> 00:27:04.950 align:start position:0% other stuff, even before it's born. Now, of<00:27:03.200> course,<00:27:03.520> all<00:27:03.679> of<00:27:03.840> this<00:27:04.159> comes<00:27:04.559> with 00:27:04.950 --> 00:27:04.960 align:start position:0% of course, all of this comes with 00:27:04.960 --> 00:27:07.909 align:start position:0% of course, all of this comes with enormous<00:27:05.679> ethical<00:27:06.159> and<00:27:06.480> societal<00:27:07.120> questions. 00:27:07.909 --> 00:27:07.919 align:start position:0% enormous ethical and societal questions. 00:27:07.919 --> 00:27:10.710 align:start position:0% enormous ethical and societal questions. Just<00:27:08.240> because<00:27:08.559> we<00:27:09.039> can<00:27:09.600> do<00:27:09.840> this<00:27:10.159> doesn't<00:27:10.480> mean 00:27:10.710 --> 00:27:10.720 align:start position:0% Just because we can do this doesn't mean 00:27:10.720 --> 00:27:13.190 align:start position:0% Just because we can do this doesn't mean we<00:27:10.960> should.<00:27:11.760> In<00:27:11.919> fact,<00:27:12.240> open<00:27:12.559> sourcing<00:27:12.960> this 00:27:13.190 --> 00:27:13.200 align:start position:0% we should. In fact, open sourcing this 00:27:13.200 --> 00:27:14.950 align:start position:0% we should. In fact, open sourcing this model<00:27:13.520> so<00:27:13.679> that<00:27:13.919> anyone<00:27:14.320> can<00:27:14.559> play<00:27:14.720> around 00:27:14.950 --> 00:27:14.960 align:start position:0% model so that anyone can play around 00:27:14.960 --> 00:27:17.350 align:start position:0% model so that anyone can play around with<00:27:15.120> it<00:27:15.360> is<00:27:15.679> quite<00:27:16.080> risky<00:27:16.559> to<00:27:16.720> be<00:27:16.880> honest.<00:27:17.200> I 00:27:17.350 --> 00:27:17.360 align:start position:0% with it is quite risky to be honest. I 00:27:17.360 --> 00:27:18.870 align:start position:0% with it is quite risky to be honest. I want<00:27:17.440> to<00:27:17.600> bring<00:27:17.679> up<00:27:17.840> the<00:27:18.000> famous<00:27:18.320> Spider-Man 00:27:18.870 --> 00:27:18.880 align:start position:0% want to bring up the famous Spider-Man 00:27:18.880 --> 00:27:21.430 align:start position:0% want to bring up the famous Spider-Man quote,<00:27:19.360> "With<00:27:19.760> great<00:27:20.080> power<00:27:20.480> comes<00:27:20.960> great 00:27:21.430 --> 00:27:21.440 align:start position:0% quote, "With great power comes great 00:27:21.440 --> 00:27:23.430 align:start position:0% quote, "With great power comes great responsibility."<00:27:22.400> Even<00:27:22.640> though<00:27:22.799> this<00:27:23.039> model 00:27:23.430 --> 00:27:23.440 align:start position:0% responsibility." Even though this model 00:27:23.440 --> 00:27:25.750 align:start position:0% responsibility." Even though this model wasn't<00:27:23.840> trained<00:27:24.159> on<00:27:24.400> any<00:27:24.720> virus<00:27:25.200> data, 00:27:25.750 --> 00:27:25.760 align:start position:0% wasn't trained on any virus data, 00:27:25.760 --> 00:27:27.669 align:start position:0% wasn't trained on any virus data, because<00:27:26.080> you<00:27:26.159> know<00:27:26.320> this<00:27:26.559> model<00:27:27.039> plus<00:27:27.440> the 00:27:27.669 --> 00:27:27.679 align:start position:0% because you know this model plus the 00:27:27.679 --> 00:27:29.750 align:start position:0% because you know this model plus the training<00:27:28.000> code<00:27:28.320> plus<00:27:28.640> the<00:27:28.880> data<00:27:29.120> set<00:27:29.440> was 00:27:29.750 --> 00:27:29.760 align:start position:0% training code plus the data set was 00:27:29.760 --> 00:27:31.269 align:start position:0% training code plus the data set was released,<00:27:30.400> maybe<00:27:30.720> someone<00:27:31.039> could 00:27:31.269 --> 00:27:31.279 align:start position:0% released, maybe someone could 00:27:31.279 --> 00:27:33.110 align:start position:0% released, maybe someone could potentially<00:27:31.760> take<00:27:31.919> this<00:27:32.159> model<00:27:32.400> and<00:27:32.640> retrain 00:27:33.110 --> 00:27:33.120 align:start position:0% potentially take this model and retrain 00:27:33.120 --> 00:27:36.149 align:start position:0% potentially take this model and retrain it<00:27:33.279> on<00:27:33.440> a<00:27:33.600> ton<00:27:33.760> of<00:27:34.000> human<00:27:34.320> virus<00:27:34.960> DNA<00:27:35.440> data<00:27:35.840> and 00:27:36.149 --> 00:27:36.159 align:start position:0% it on a ton of human virus DNA data and 00:27:36.159 --> 00:27:37.669 align:start position:0% it on a ton of human virus DNA data and create<00:27:36.320> a<00:27:36.559> model<00:27:36.720> that<00:27:36.960> can<00:27:37.120> generate<00:27:37.440> a 00:27:37.669 --> 00:27:37.679 align:start position:0% create a model that can generate a 00:27:37.679 --> 00:27:39.909 align:start position:0% create a model that can generate a pandemic<00:27:38.320> beyond<00:27:38.720> our<00:27:39.039> imagination. 00:27:39.909 --> 00:27:39.919 align:start position:0% pandemic beyond our imagination. 00:27:39.919 --> 00:27:42.310 align:start position:0% pandemic beyond our imagination. Anyways,<00:27:40.480> that<00:27:40.720> was<00:27:40.960> quite<00:27:41.200> the<00:27:41.440> deep<00:27:41.679> dive. 00:27:42.310 --> 00:27:42.320 align:start position:0% Anyways, that was quite the deep dive. 00:27:42.320 --> 00:27:44.310 align:start position:0% Anyways, that was quite the deep dive. This<00:27:42.480> paper<00:27:42.799> is<00:27:42.960> just<00:27:43.200> jam-packed<00:27:43.840> with<00:27:44.080> stuff 00:27:44.310 --> 00:27:44.320 align:start position:0% This paper is just jam-packed with stuff 00:27:44.320 --> 00:27:46.149 align:start position:0% This paper is just jam-packed with stuff and<00:27:44.559> there's<00:27:44.799> a<00:27:45.039> ton<00:27:45.279> of<00:27:45.440> technical<00:27:45.919> things 00:27:46.149 --> 00:27:46.159 align:start position:0% and there's a ton of technical things 00:27:46.159 --> 00:27:48.390 align:start position:0% and there's a ton of technical things that<00:27:46.480> I<00:27:46.720> didn't<00:27:47.039> even<00:27:47.360> touch<00:27:47.520> on,<00:27:47.840> but<00:27:48.080> that's 00:27:48.390 --> 00:27:48.400 align:start position:0% that I didn't even touch on, but that's 00:27:48.400 --> 00:27:51.110 align:start position:0% that I didn't even touch on, but that's basically<00:27:48.880> the<00:27:49.120> gist<00:27:49.520> of<00:27:49.760> this<00:27:50.000> paper.<00:27:50.559> Hope<00:27:50.880> I 00:27:51.110 --> 00:27:51.120 align:start position:0% basically the gist of this paper. Hope I 00:27:51.120 --> 00:27:53.269 align:start position:0% basically the gist of this paper. Hope I was<00:27:51.200> able<00:27:51.520> to<00:27:51.840> make<00:27:52.080> it<00:27:52.320> easy<00:27:52.559> enough<00:27:52.880> for<00:27:53.120> you 00:27:53.269 --> 00:27:53.279 align:start position:0% was able to make it easy enough for you 00:27:53.279 --> 00:27:55.190 align:start position:0% was able to make it easy enough for you to<00:27:53.520> understand<00:27:54.080> and<00:27:54.320> also<00:27:54.559> show<00:27:54.720> you<00:27:54.880> how 00:27:55.190 --> 00:27:55.200 align:start position:0% to understand and also show you how 00:27:55.200 --> 00:27:57.510 align:start position:0% to understand and also show you how mind-blowing<00:27:55.919> this<00:27:56.159> AI<00:27:56.559> is.<00:27:56.960> Let<00:27:57.120> me<00:27:57.200> know<00:27:57.279> in 00:27:57.510 --> 00:27:57.520 align:start position:0% mind-blowing this AI is. Let me know in 00:27:57.520 --> 00:27:58.950 align:start position:0% mind-blowing this AI is. Let me know in the<00:27:57.600> comments<00:27:57.840> what<00:27:58.080> you<00:27:58.240> think<00:27:58.399> of<00:27:58.640> all<00:27:58.880> of 00:27:58.950 --> 00:27:58.960 align:start position:0% the comments what you think of all of 00:27:58.960 --> 00:28:01.350 align:start position:0% the comments what you think of all of this.<00:27:59.600> As<00:27:59.919> always,<00:28:00.399> I<00:28:00.640> will<00:28:00.799> be<00:28:00.960> on<00:28:01.120> the 00:28:01.350 --> 00:28:01.360 align:start position:0% this. As always, I will be on the 00:28:01.360 --> 00:28:04.310 align:start position:0% this. As always, I will be on the lookout<00:28:01.760> for<00:28:02.000> the<00:28:02.240> top<00:28:02.559> AI<00:28:03.039> news<00:28:03.360> and<00:28:03.679> tools<00:28:04.080> to 00:28:04.310 --> 00:28:04.320 align:start position:0% lookout for the top AI news and tools to 00:28:04.320 --> 00:28:06.230 align:start position:0% lookout for the top AI news and tools to share<00:28:04.559> with<00:28:04.720> you.<00:28:05.120> So,<00:28:05.279> if<00:28:05.520> you<00:28:05.679> enjoyed<00:28:06.000> this 00:28:06.230 --> 00:28:06.240 align:start position:0% share with you. So, if you enjoyed this 00:28:06.240 --> 00:28:07.990 align:start position:0% share with you. So, if you enjoyed this video,<00:28:06.720> remember<00:28:07.039> to<00:28:07.279> like,<00:28:07.600> share, 00:28:07.990 --> 00:28:08.000 align:start position:0% video, remember to like, share, 00:28:08.000 --> 00:28:09.990 align:start position:0% video, remember to like, share, subscribe,<00:28:08.559> and<00:28:08.799> stay<00:28:09.120> tuned<00:28:09.440> for<00:28:09.679> more 00:28:09.990 --> 00:28:10.000 align:start position:0% subscribe, and stay tuned for more 00:28:10.000 --> 00:28:12.310 align:start position:0% subscribe, and stay tuned for more content.<00:28:10.720> Also,<00:28:11.120> there's<00:28:11.440> just<00:28:11.760> so<00:28:12.080> much 00:28:12.310 --> 00:28:12.320 align:start position:0% content. Also, there's just so much 00:28:12.320 --> 00:28:14.149 align:start position:0% content. Also, there's just so much happening<00:28:12.559> in<00:28:12.720> the<00:28:12.880> world<00:28:13.039> of<00:28:13.200> AI<00:28:13.600> every<00:28:13.840> week. 00:28:14.149 --> 00:28:14.159 align:start position:0% happening in the world of AI every week. 00:28:14.159 --> 00:28:16.630 align:start position:0% happening in the world of AI every week. I<00:28:14.399> can't<00:28:14.720> possibly<00:28:15.200> cover<00:28:15.600> everything<00:28:16.080> on<00:28:16.399> my 00:28:16.630 --> 00:28:16.640 align:start position:0% I can't possibly cover everything on my 00:28:16.640 --> 00:28:18.789 align:start position:0% I can't possibly cover everything on my YouTube<00:28:16.960> channel.<00:28:17.520> So,<00:28:17.760> to<00:28:18.080> really<00:28:18.559> stay 00:28:18.789 --> 00:28:18.799 align:start position:0% YouTube channel. So, to really stay 00:28:18.799 --> 00:28:21.750 align:start position:0% YouTube channel. So, to really stay uptodate<00:28:19.600> with<00:28:19.919> all<00:28:20.159> that's<00:28:20.480> going<00:28:20.640> on<00:28:20.799> in<00:28:21.120> AI, 00:28:21.750 --> 00:28:21.760 align:start position:0% uptodate with all that's going on in AI, 00:28:21.760 --> 00:28:24.230 align:start position:0% uptodate with all that's going on in AI, be<00:28:21.919> sure<00:28:22.080> to<00:28:22.399> subscribe<00:28:22.799> to<00:28:23.120> my<00:28:23.440> free<00:28:23.840> weekly 00:28:24.230 --> 00:28:24.240 align:start position:0% be sure to subscribe to my free weekly 00:28:24.240 --> 00:28:26.070 align:start position:0% be sure to subscribe to my free weekly newsletter.<00:28:24.880> The<00:28:25.120> link<00:28:25.200> to<00:28:25.360> that<00:28:25.600> will<00:28:25.760> be<00:28:25.919> in 00:28:26.070 --> 00:28:26.080 align:start position:0% newsletter. The link to that will be in 00:28:26.080 --> 00:28:27.750 align:start position:0% newsletter. The link to that will be in the<00:28:26.240> description<00:28:26.720> below.<00:28:27.360> Thanks<00:28:27.600> for 00:28:27.750 --> 00:28:27.760 align:start position:0% the description below. Thanks for 00:28:27.760 --> 00:28:29.750 align:start position:0% the description below. Thanks for watching<00:28:28.159> and<00:28:28.480> I'll<00:28:28.799> see<00:28:28.960> you<00:28:29.200> in<00:28:29.440> the<00:28:29.600> next 00:28:29.750 --> 00:28:29.760 align:start position:0% watching and I'll see you in the next 00:28:29.760 --> 00:28:32.080 align:start position:0% watching and I'll see you in the next one.