Unsupervised Learning with Jacob Effron
00:46:44 · Erik Bernhardsson
[00:46:44] S1: So so some of them, I think, will probably work out really well. And... You know? But I imagine in the grands... You know, like, some of them will not work out. And if they don't work out, well, now they're sitting on a ton of GPUs. Like, you know, let's resell those GPUs. Right? Like, especially if there's ever, like, a scenario where, like, VC funding pulls back. Because I I think what... What's what's kind of a scary thing about some of these commitments is, like, you know, effectively, like, what what happens today when you make a GPU commitment is you only have to pay about 30% upfront, sometimes a little bit more, sometimes a little bit less. But what that means is like you can raise a billion dollars. You can go out and and get GPUs for $3,000,000,000 and only pay a billion dollars upfront. I mean, not exact math. But the problem is like, you know, in a year, now you're on the hook another for few 100,000,000, and, like, every month, now you have to pay. So so I I I think that creates, like, a little bit of, like, implicit leverage in a lot of these companies and and and and a and a dependence on subsequent fundraising. And, you know, that works as long as they can go and raise VC dollars to fund the the GPU bills. But at some point, it might not work. And and in that case, I think you're gonna see it a little bit more of a a secondary market where some of these neo labs might, you know, pivot to doing a little bit more sort of neo cloud style, like, we have a bunch of GPUs or or just resell it back to the cloud or I don't know. I'm I'm very bullish on AI as a whole, but, know, I always, like, look for these, like, sort of structural, like, risks. So overall, I think this could be, like, incredible application, incredible demand. But, like, you know, is a 100% of the Neolabs gonna succeed and, you know, be a, you know, trillion dollar company? Like, probably not.
00:48:15 · Jacob Effron
[00:48:15] S0: You know, I I feel like on the AIM for side, one big category that has gotten a lot of buzz is like this RL as a service market and kind of a bunch of folks, you know, almost forward deploying and, like, helping companies build their own models. And, you know, you have had folks like Satya saying, hey. Every company is gonna have their own model. Like, do you subscribe to that? Like, what do you kinda think of this... Of of that overall market?
00:48:35 · Erik Bernhardsson
[00:48:35] S1: No. I think some companies will have their own models. I I I think the... Like, weighted by dollar revenue, like, I I think most models will probably be post trained. Right? Like... But but, like, weighted by company, like, in terms of, like, numbers of companies, like, most companies will not have the the need to, like, post... Do their own post training. Right? But, like, the large ones, right, like, Cognition is training their own code models and and and and CursorDate and and and all these companies are. And, like, those the ones who also served a lot of tokens. So I I I think when you look at, like, reinforcement learning and post training, like, certain types of, you know, parts of it is very attractive. Right? Like, you can work with companies and and and serve, you know, better models and and do it at lower cost and and higher throughput and all of this stuff. Some of it looks like... It's kind of like just a service business. So I I think a lot of the sort of tension over the next few years when you look at post training is gonna be, can you turn this into product? Like, can you sell that product to these companies? This is something we think a lot about is, like, how do we create this, like, amazing product you can just kinda sell and and not turn it into just, a service business.
00:49:38 · Jacob Effron
[00:49:38] S0: And then I guess last one before we move to the quick fire round would just be like, I think there's this question about how many tools live under like a single infra provider in AI world versus our separate categories. I've always struck when we talk to you about like just how much you wanna go build and all these kind of building blocks. I'm wondering if there's categories of AI infra, you know, that you think will be unique and distinct from the stuff you're building that you're optimistic on and that will kind of, like, live as as still large standalone opportunities.
00:50:05 · Erik Bernhardsson
[00:50:05] S1: Yeah. I mean, we're not gonna go into model training, I think. Like, so I I think TypeSafe with Java is gonna be... You know, this can probably gonna be a lot of these, like, custom models. I'm sure in other verticals too, you know, other mode... Modalities, like audio and video and music and, like, obviously, the application side, there's gonna be a lot of incredible companies. And and then there's probably gonna be a lot of companies, like, focusing more on, like, software development and and and stuff like that. And and, you know, I... So I I don't know. Like, my general view on these things is, like, there's always, like, more boxes on the market map than you expect. But I also do think that overall, every company that is successful in the end to some extent has to become a multiproduct company too. So I I don't know if that's attention or not. Like, I I I think, you know, we obviously wanna find, like I said, a cluster of, like, very adjacent things and and do all of it in a way that, like, you know, maximizes the the the revenue and and the shareholder value. But I I... You know, by no means, you know, are we gonna try to expand into everything. I I think people kinda make this, like, false assumption also, like, you know, you look at, like, Anthropic or OpenAI, and they're like, why don't they move up to stock and launch their own law firm and launch their own... Whatever. Like, they're very good at, like, serving these models and doing that at scale. That's how they achieve scale. Right? Like... And they're very good at doing that. Like, I I I don't think it makes sense for every company to do everything. And so similar for us, like, we're not gonna do everything, but we're gonna do a very large cluster of adjacent things and do it exceptionally well. And I think there's gonna be a lot of room in the market to create, you know, a bunch of, you know, very large, you know, 100,000,000,000 or trillion dollar opportunities in the market.
00:51:40 · Jacob Effron, Erik Bernhardsson, Patrick Chase
[00:51:40] S0: We'll end on our quick fire round where we we stuff probably a similar set of questions what we asked you three years ago, maybe a little bit of an evolution. To start, maybe I'd love just one thing you've changed your mind on in AI in the last year. [00:51:52] S1: Well, the opportunity size of sandboxes has obviously been kind of crazy. Right? Like, I I didn't think it would be anywhere close to, you know, I mean I [00:51:59] S0: thought you knew that in '23, and that's why you made the made the big bet. [00:52:03] S1: Yeah. So so that's been kind of incredible. [00:52:06] S2: Coding agents have have changed a lot of what it means to be an engineer. How, if at all, has it changed how you think about hiring?