[0:00]
So, I'm at my in-laws house this weekend on Sunday, and you know, there's food, there's family, there's a child who I am told is mine, two of them, if I recall correctly. And I'm sitting there in the corner like an [ __ ] swiping furiously on my phone, sweating because at 2:00 a.m. the night before, I had a thought, which I should not have had, which is, what if AGI is possible? I posted a membersonly video the other day confessing that I had accidentally wandered into a sketchy part of my brain 
[0:31]
where I entertained the idea that AGI could be possible in theory and is an engineering problem that could be solved in a couple hundred years or so. And from the moment I woke up yesterday to about the time I went to sleep, I was reading reading about AGI and intelligence and what other scientists and authors believe. and I've decided that I'm [ __ ] out. I dipped one foot into the cult of AGI by accident and now I'm noping the [ __ ] back out. I mean, I'm absolutely flabbergasted to find out 
[1:03]
that David Deutsch, the father of quantum computing and the author of the book The Beginning of Infinity, doesn't even believe animals are intelligent or conscious. Like I'm sitting here trying to determine if a Python script can be generally intelligent and this man is like dogs aren't even alive. At that point I was like, "Okay, you know what? I'm out." Like if that's where the line is right now, if there are cases being made that animals aren't intelligent, then how the [ __ ] are we even having a conversation about whether computers are 
[1:34]
intelligent or what the [ __ ] it would mean for them to be generally intelligent? There are no definitions for any of these terms that we are claiming exists or is even possible. There is [ __ ] nothing. AGI is some sort of materialists propaganda wet dream [ __ ] I am absolutely livid because I spent 12 hours reading up on this [ __ ] yesterday and I'm absolutely exhausted. My wife was like, "Honey, can you put down your phone and acknowledge my mother?" And I was like, "Not now, babe. I've got to figure out whether intelligence is possible." The worst 
[2:05]
part is that it's not even clear what problem AGI would solve. Like why would you need AGI to do a coding task? Why would you need a general intelligence to do that? What does that offer on top of current generation LLMs? In fact, for things like coding, you don't want generality. You want the opposite. You want specialization. The mid-level engineer on your team who's been working in your codebase for 10 years writes a lot better code than a PhD level engineer who you just hired who isn't familiar with your code. It's about 
[2:35]
carrying the whole thing in your head for many, many years. It's about specializing in that specific code base. The problem with LLMs today, the reason they suck balls at coding is they can't hold a damn thing in its tiny ass head. And the whole thing not even being the codebase itself, but the codebase in all the intangibles and the whole business context and the politics and the reason your stuckup designer will literally quit if you change the color of the button. It's [ __ ] impossible. LLMs by 
[3:06]
definition are doomed to fail in a codebase. And what the [ __ ] problem would AGI solve here? It's a [ __ ] context problem. It's a [ __ ] understanding problem. It's a being engrossed problem. soaking yourself in the universe of the problem you're trying to solve. An AI cannot do that. An AI cannot understand or feel or soak or engross. David Deutsch argues that because they seem to have arrived in a package at once in the form of a human that it doesn't even make sense to separate intelligence from consciousness 
[3:37]
and personhood. And from now on, I'm going to stop pontificating about the future. I'm going to stop giving the technology the benefit of the doubt and the benefit of progress and extrapolation. I'm going to take the present state of LLM and I'm going to treat that as the only fact that exists. Period. I will behave as though there is no such thing as progress. There's no such thing as the future. There's no such thing as 12 to 18 months from now. There is only [ __ ] now. And right [ __ ] now, LLMs are a trick. They are 
[4:07]
not AI. They trick us to believing they are intelligent. The mechanism matters. The mechanism absolutely matters. David Deutsch talks about this in his book. He says that the touring test isn't just about passing AI through some glory hole and determining if the other person thinks it tastes good or not. On the other end, if you could pass the Turing test by tricking people, then what the [ __ ] is the point of the Turing test? You can polish a lie forever and it never becomes the truth. You can put a billion more parameters on the lie and it's a bigger lie now. Like, 
[4:38]
congratulations. LLMs are an elaborate trick. There is no theory in there. It's brute forcing statistics to trick people for long durations, long enough for you to be tricked into forcing your entire company to think this is intelligent and that they have to use this tool to maximize productivity. The real turing test is whether the human on the other end is [ __ ] stupid enough to replace their entire labor pool with this statistical ass tool that builds in inside the distribution demo and thinks 
[5:08]
it could build anything. The bot doesn't have to be intelligent. It just has to be more convincing than the median CEO is skeptical. Oh, but how do you know the human brain isn't just dumb statistics, too? Go [ __ ] yourself. Go [ __ ] yourself. [ __ ] you. I will never again extrapolate. I will never again speak of progress. I will never again give this technology the benefit of the doubt. I will take the present and evaluate that on its own merit. Just 
[5:38]
like we judge literally everything else we buy, nothing else matters. Everything else is [ __ ] fear, confusion, uncertainty, doubt, and propaganda. Thank you for [ __ ] watching.