[0:00]
You know, it's wild. There's an Uber driver asked me the same question on the way home from Folsom this weekend, and, um, I think we're already there. So the question is, what's the leadership talent threshold beneath which open the eye loses the juice? Are they approaching it soon? I think we already passed it. So what does this mean? Well, all. All of the talents left. Tesla particularly. And especially Andre Carpenter, 
[0:32]
who is. There's no. There's only one of him. The smartest person in the world on AI. Okay. Was the head of AI. Tesla quit to go co found open AI with a lot of other really smart people. And they were working on this really specific concept, which is, if we take this one paper that was published at Google called attention is all you need, and we just, like, 100,000 x what they tried in that paper, that. That would be really cool. 
[1:03]
That is the premise upon which Open Air was founded. And that is Jupiter 3. Okay. Thousand likes it again. You get GPT for. Okay. And then we learn that, like, you really don't have to do that. Like, you can do 10% of that and make it better. And Facebook has, and they've made it free. And so Llama 3 is already better than GPT4. It's been better than GPT3. And it'll run on your phone. You know, 
[1:37]
foundation models are the world's fastest appreciating asset. It costs hundreds of Millions of dollars to train one of these AIs, and then the next day, someone releases a better one, and everything you made is worthless. And that's happening constantly now. The talent that left Tesla took with it all the progress in AI at Tesla. Tesla has made no progress since Andre Carpentry left. 
[2:08]
And then the person that they tried to replace him with left. And Andre Carpentry brought that talent to open AI, and then he left. And in fact, almost everyone from the founding team left. There is almost no one left from the team that builds open AI. And now they're trying to go public. There's a very good reason why Microsoft didn't want equity in open AI. 
[2:38]
They gave them a bunch of money for exclusive rights for short time because that's the only value there was ever gonna be in open AI. So Sam is trying to get a big payday because it's over. Like, they don't have a moat. All the top talent left to start competing. Uh, firms, anthropic, right. All these. All these other AI firms are the talent that left open AI. And when Andre Carpathi left, 
[3:09]
he published on Github instructions for how everyone can copy everything that open AI has accomplished so far. And they don't have the same, anything, anywhere remotely near the same level of competent talent there that they did when they started. And it's because the challenges that we face from here going forward, Are challenges of tooling. We've long since passed the limits of, 
[3:42]
uh, the, the boundaries of, of scale for training that benefit from the same approach. So the approach is today, for example, that people like young lacuna working on with, like, way more complex classifiers, way more complex topologies and stuff like that. But, like, that's enormously expensive, incremental progress that Facebook is, is willing to pay for and then give away for free. 
[4:15]
Because it kills everyone else. It kills all the opening eyes of the world. To have a much bigger and more powerful and more well funded company willing to do whatever it takes to destroy you. Right. So at some point we're going to have these models. I, I would, I would argue we already do. We already have models that can, in effect, accomplish anything with the right tooling. And this is the huge current push with the llama ecosystem is tooling integrations. That's what they're focused on now. 
[4:49]
Right. So getting it to be able to control things and do things in the world and standardizing how that works. So the next generation of tooling is around agents. Right. And we see that in 0 1. Right. But we also see that in old stuff like auto GPT. Like, it's not a new concept. And that tooling is a completely separate project 
[5:20]
from the project of building LMS. We have the LMS. What's next is not different LMS. There will be incremental changes. There will be marginal improvements in the efficiency, but it's the tooling that is the next step. And the tooling doesn't need open AI. And it's already cheaper for people to use alternatives. 
[5:53]
There are already better alternatives people can use if they want to use a cloud provider. They aren't the fastest, they aren't the smartest, they aren't the best at anything, they're just the most popular. Um, and so enterprise is going to continue moving away from providers like Openai and to other providers that have faster, lower latency, lower priced, lower optics alternatives. I think their time is already over. 
[6:24]
That's why they're trying to go public and cash out. So I would say the answer is we've long since passed that line. If you wanna bet on AI today, bet on Facebook, bet on Nvidia, to some degree Microsoft, not Google, and certainly not open AI.