William Beauchamp: evaluation

Transcript context

…This last question on my side, and I want to pass this to Alessio, is on just, just multi-modality in general. This actually comes from Justine Moore from A16Z, who's a friend of ours. And a lot of people are trying to do voice image video for AI companions. Yes. You just said voice didn't work. Yep. What would make you revisit? So Steve Jobs, he was very, listen, he was very, very clear on this. There's a habit of engineers who, once they've got some cool technology, they want to find a way to package up the cool technology and sell it to consumers, right? That does not work. So you're free to try and build a startup where you've got your cool tech and you want to find someone to sell it to. That's not what we do at Chai. At Chai, we start with the consumer. What does the consumer want? What is their problem? And how do we solve it? So right now, the number one problems for the users, it's not the audio. That's not the number one problem. It's not the image generation either. That's not their problem either. The number one problem for users in AI is this. All the AI is being generated by middle-aged men in Silicon Valley, right? That's all the content. You're interacting with this AI. You're speaking to it for 90 minutes on average. It's being trained by middle-aged men. The guys out there, they're out there. They're talking to you. They're talking to you. They're like, oh, what should the AI say in this situation, right? What's funny, right? What's cool? What's boring? What's entertaining? That's not the way it should be. The way it should be is that the users should be creating the AI, right? And so the way I speak about it is this. Chai, we have this AI engine in which sits atop a thin layer of UGC. So the thin layer of UGC is absolutely essential, right? It's just prompts. But it's just prompts. It's just an image. It's just a name. It's like we've done 1% of what we could do. So we need to keep thickening up that layer of UGC. It must be the case that the users can train the AI. And if reinforcement learning is powerful and important, they have to be able to do that. And so it's got to be the case that there exists, you know, I say to the team, just as Mr. Beast is able to spend 100 million a year or whatever it is on his production company, and he's got a team building the content, the Mr. Beast company is able to spend 100 million a year on his production company. And he's got a team building the content, which then he shares on the YouTube platform. Until there's a team that's earning 100 million a year or spending 100 million on the content that they're producing for the Chai platform, we're not finished, right? So that's the problem. That's what we're excited to build. And getting too caught up in the tech, I think is a fool's errand. It does not work. As an aside, I saw the Beast Games thing on Amazon Prime. It's not doing well. And I'm…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.