No Priors: Artificial Intelligence | Technology | Startups
00:21:49 · Elad Gil
[00:21:49] S0: Yeah. I just think there's a lot of second order effects that are happening or going to happen. Part of them are driven by this belief system and potential burnout over time. Part of it is going to be One thing that I've noticed that is happening at some of the labs is that as compute becomes really the scarce resource, it turns out that there's say a few dozen researchers that drive 80% of the results at any given place, which is a really interesting human power law. If you actually look at it in any field, there's at most a few dozen people who drive the field. You look at breast cancer research, you look at certain subfields of mathematics, you look at subfields of physics, You look at the entrepreneurial ecosystem and founders. There's a handful of people, dozens of people who drive most progress. And that also happens in AI research. And increasingly, compute is differentially provided to those people. Right? And so I know some labs have slowed down on their hiring of researchers unless they're above a very, very high bar because the cost isn't the researcher, it's the compute associated with the person. That's where the real bottleneck is. I think there's this broader concept of return on invested tokens, like an ROI kind of metric, which is if you have a certain token budget, who do you give it to and why? This is kind of like engineering back in the day. Right? The internal tools teams at companies were always starved for resources because many, at least tech companies, would rather use the same engineers to build product than to build internal tools that would make other functions more productive. That's why I think the death of SaaS is a little bit overstated because why would you use tokens on a bunch of SaaS stuff that you're not actually paying that much for per year relative to the outcome of those same tokens being invested against a core product or against some massive margin lift or some other thing? And so I think increasingly we've shifted from a world where people said, Hey, everybody use AI and do whatever you want, to hey, we have to measure, spend, and move more things to open source. And then I think the next wave is what are the projects and people that should actually get outsized pieces of a token budget, and what is that return on investment? It's this sort of next shift that's coming? It'll take some time, though.
00:23:57 · Elad Gil, Sarah Guo
I mean, many people are still at like, hey, everybody try AI or whatever, you know, big enterprises. [00:24:02] S1: What do you think is the appropriate compute, like token budget for a business or a human being three to five years from now? Should I look at it like rent?
00:24:14 · Elad Gil
[00:24:14] S0: No. I mean, it depends on what the budget is for what. I mean, people forget too. Minecraft was like, what was it, five people, 10 people, and it was bought for billions of dollars by Microsoft. People keep talking about someday there'll be a multi billion dollar single person company. That was basically Minecraft, roughly. It already happened fifteen years ago or whenever that was. So there are always people who can take outsized advantages of technology, and AI has accelerated that radically. And so at some point, it's like, why give tokens to people who can't do that on a relative basis unless you just run out of those people? And you may run out of them. This is back to like, well, all the engineers get laid off. Probably not anytime soon. But you could argue that at some companies, even before AI, there's a bunch of engineers that weren't that productive that could be let go, especially some of the big tech companies. And I think a lot of those folks will be very coveted by GE or PG and E or Hershey's. So even if there is some displacement of engineers at some point in the future, I don't know when that is or if it happens. But if it does happen, there's lots and lots of homes for them Because there's tons of enterprises that never had the capability set or ability to recruit these people, and they want the capabilities they bring. Even if they're mediocre in the context of a Google or Meta or whatever, they may be exceptional in the context of a certain subset of old school enterprises. And so I do think there's gonna be this permeation through the enterprise landscape of engineering talent in an unexpected way. This probably is many years away. I'm just saying I think that's probably a likely outcome.
00:25:45 · Sarah Guo, Elad Gil
[00:25:45] S1: I think relatedly, if, you know, if you are a researcher 800 and you have not been allocated an outsized number of tokens to work with at one of the major labs. [00:25:57] S0: Yeah. [00:25:58] S1: I think the opportunity to go spend your energy on, you know, something where you have, like, compared advantage and understanding and should benefit from all of this. The supply chain bottlenecks, domains that should accelerate like bio, diffusion into other valuable fields. Like, to me, that that seems a lot more exciting than being concerned about the downfall of mathematics and going on, you know, vacation until the world ends.
00:25:58 · Sarah Guo, Elad Gil
[00:25:58] S1: I think the opportunity to go spend your energy on, you know, something where you have, like, compared advantage and understanding and should benefit from all of this. The supply chain bottlenecks, domains that should accelerate like bio, diffusion into other valuable fields. Like, to me, that that seems a lot more exciting than being concerned about the downfall of mathematics and going on, you know, vacation until the world ends. [00:26:25] S0: Oh, yeah. Lots of places to go do stuff. And I do think that's where a subset of the research community will end up over time. Right? And that's an interesting question is how many researchers do you need if you're a top AI lab? And how does that number relate to the number that you have now? And is it you have the right number? Is it you need five times as many people? Is it you need half as many, but they only know you have a certain bar because it's compute constrained? And you wanna map it against the best ideas, and the best ideas come from a subset of people on average. Not always, but on average. And so it's a really interesting question of how does all this stuff fall out, and then where does the n plus one person go? And there's lots and lots and lots of places for the n plus one person to go. They're still exceptional. They're still top of the bell curve. Again, I don't wanna have that misinterpreted as the person is not being amazing. It's just at some point, people will do some cutoff on their power law.
00:27:18 · Sarah Guo, Elad Gil
[00:27:18] S1: I think you will appreciate this. Maybe you've heard it because it's an old ex Googler joke. But when Google is like, I don't know, 50,000 people or something, the question was, how many people does it take to run Google? And if you ask somebody within search and ads, they're like, oh, like 20% of people in search and ads. And if you ask somebody outside of, you know, search and ads, they'd say, like, 50,000 people or whatever Google is. So I think this is probably some varied perspectives on the concentration of contribution. I don't know, having never worked at Google. [00:27:53] S0: Yeah. I mean, we definitely know that yeah. I mean, I worked at Google. I thought it was a wonderful place. [00:28:00] S1: In search? [00:28:01] S0: I worked on mobile search a bit, and I worked on worked on ads a bit. I mean, I worked on a bunch of mobile stuff, and then I worked on a bunch of, like, AI ads related stuff.