The Exchanges

Q But you dispute, uh, you dispute, or at least you, you question the, Uh, sort of recent exponential aspect of this you, you feel is mostly, uh, sort of a question of perception and emotional, or do you think there's something?

A I mean, there's definitely like, there's definitely positive feedback loops with the additional money and people coming into it. So obviously things go harder and faster and, you know, getting to the point that we can use things like synthetic data effectively also has some nice positive feedback loops. Um, so there's definitely some, Exponentials at play. The basic economics says that that's the exponential that is at the start of a sigmoid curve, you know, and they're impossible to tell the difference for a while. So we're starting to see that with training already. The amount of people and money being put into AI, it's going to keep going up, but again, the low hanging fruits being done now. So yeah, I think we'll continue to see the The tailing off of, of, of growth. But regardless, you know, none of that means ASI You know, like, it really is a continuation of trends we've been seeing for quite a while.

AI assessment note: “there's definitely some, Exponentials at play. The basic economics says that that's the exponential”

Q The plan is to eventually, uh, commercialize or, or at least make Solvid broadly available. You mentioned thousand beta testers, or, or is that going to be the secret sauce, uh, for, for Answer AI that's going to enable you to create those 5000 or X many thousand?

A Yeah, I think, I think it'll be like AWS. I think we'll, I think we will continue to make it more and more available to people. The w the way we did those first thousand, Was, we barely even mentioned SolveIt, the product. We just talked about creating a, which we did do, we created a course called How to Solve It with Code, based on Polya's classic math book, um, How to Solve It. And we said, you know, we'll open up registrations, but if it gets to a thousand people, we'll close it off. And that was hit within 24 hours. So there was a lot of interest in, like, Understanding, like, how to solve problems with code, and we, we took people through the process in the platform, because there's no other platform that's designed for this kind of iterative approach like SolveVit is. Now we've had hundreds of people of those come back to us and say basically this has actually changed my life. I can now do things I could never do before, You know, I got this job. I started this startup. I, you know, solve this long running academic problem. So it's been a bit mind blowing, but it does feel like it's not something we can just say like, here's the product, go use it. It kind of needs the training as well. So maybe we're thinking at the moment the next one will be like, maybe we'll cap it at 10,000 rather than 1000 and go through a similar process, but we want to make it each time we do it.…

AI assessment note: “I think we will continue to make it more and more available to people.”

Q A few months into that big, you know, deep seek moment that some call the spotting moment for AI and all the things. Is that still something that from, uh, you know, an AI community perspective feels as important today as it did then? Or was that like overblown in the press?

A Well, I, I never understood it at the time. We'd been using deep seek for ages and they just, Like, I don't know how these things suddenly pop into the public perception. Had been asking all of our vendors, you know, to add DeepSeq and Quen models, and they're all like, oh no, we wouldn't do that, they're Chinese, blah blah blah. And then, for whatever reason, one particular one gets picked up by the public at large, and suddenly everybody's like, oh look at us, now we've got, ah, one on our, Serving infrastructure. It's like the very same people that just two weeks earlier were telling me that they would put a Chinese model on their serving infrastructure. So I found it all really weird. You know, DeepSeek's been great for a long time. They've kept on improving. They'll keep on improving. For me, there was no technology DeepSeek moment. One was just another expected journey along their path, and, um, was interesting to me to see how Again, just some things can break into the public awareness, I guess.

AI assessment note: “For me, there was no technology DeepSeek moment.”

Q You guys made the very clear decision to not be an AGI lab. So if Test Time Compute is going to, um, start producing decelerating returns at some point, you know, what, what do you make of the whole AGI, ASI kind of thing? Is that something that you think about a lot?

A Yeah, I don't really have a strong opinion about it. I don't think we have any more evidence that ASI might be close now than we did 15 years ago, 15 years before that. At any of those times, you could say, like, oh, ASI might be close. I think a lot of people have the impression it might suddenly be really close because when you change the user interface to a computer from a kind of interface design to be computer friendly to an interface designed to be human friendly, i.e. natural language, our brains think we're dealing with a different kind of thing, but all that's changed is the user interface to that thing, you know, it's doing, it's doing the same things it was always doing before, but now we have natural language import. Natural language output. You can kind of ingest and calculate with natural language text, audio. These are really useful things, um, but I think the impact on our brain of thinking like, oh, this is closer to superintelligence is Maybe our brains are getting a little bit tricked.

AI assessment note: “I don't think we have any more evidence that ASI might be close now”

Q And in terms of what those big labs produce, you just mentioned all three in deep research a minute ago. So the, the whole, like, um, Test time compute approach and all the things it seems, uh, from your perspective as, you know, deep experts, like a very promising avenue. Like, do you see that, uh, continue and, um, you know, continue to improve model performance?

A Yeah, it's weird that people took so long to care about that. It's another of these things that lesser known papers have been kind of indicating about for a long time is, you know, adding a few tokens to give some breathing room or thinking room or whatever is Important. I guess the weird thing is people seem to treat it as being like, oh, it's like part of some kind of exponential curve, but actually it's like, oh, that was like a really obvious thing that lots of people knew had to be done, was take advantage of inference time compute, not just train time compute. It's not some infinite money supply. It's a thing where you get most of the juice out of it in the first year or two, so we're still in that. Phase at the moment, and we'll start to hit the curve off point pretty soon. Just like we did for, for training. But you know, the nice thing is now people are starting to understand on the training side, for instance, that, okay, COGS, you know, cost of goods sold actually does matter. And people are starting to invest in efficiency again, which is nice to me because that's kind of always been my thing is I care a lot about efficiency because I care about maximizing the number of people that can use a technology to make their lives better. And when it's expensive, that reduces the number of people a lot. So it's nice to see both people starting to care about costs And compani…

AI assessment note: “we're still in that Phase at the moment, and we'll start to hit the curve off point pretty soon.”

Q And the fact that OpenAI announced intentions to open source, you know, at least one more model, what do you make of that?

A Really take any account of vaporware announcements. Well, we'll see if and when it happens, like vaporware announcements can only ever be made as a way to try and stay relevant. So I'd rather Look at actual things they're doing. So, for example, O-three is an interesting model. I hardly ever use it. One of our team who's been building Triton kernels has been found for some of the things he's doing. It's the only one that's really able to help. Also, um, their deep research tool is still the best. So, you know, it's nice. We've got lots of good Products coming out from open source world and closed source world and in this multi-party environment, I think it's great for everybody actually.

AI assessment note: “vaporware announcements can only ever be made as a way to try and stay relevant”

Q And then another eight from Angels. And then, uh, so the, the, there, there is, is there already a business model? Is it something that's coming up? In other words, are you, are you charging for some of the products or not yet? To rephrase the question, is, is the business model, uh, active, uh, today?

A Slightly. You know, we've done some little tests, we've made a small amount of money. The main thing that we've been working on is more the platform for us to rapidly experiment with AI applications. And actually that's the thing we've made money from as well as we've let a thousand people use that platform as a kind of a pre-release beta test. Um, Yeah, it's an interesting position to be in. It's kind of a bit AWS like, you know, it's, Platform is built for us. Obviously it's very useful for other people as well. So that question of how do we think about it or invest in it? We're kind of like, okay, we're primarily going to think about it and invest in it as an internal tool. And we're going to focus on us as customers, but that's quite a good way to create tools that other people like anyway, you know, or at least people who are a bit like you.

AI assessment note: “Slightly. You know, we've done some little tests, we've made a small amount of money.”

Q I hadn't realized, I thought that was just for code, but that's for any, that's for text as well?

A No, it's, it can't be just for code, because if we're going to make 5000 products, like GE did, that use AI, they're going to be as broad as GE's products were that used electricity. So yeah, I'm, I'm using it to help me manage my business. I'm using it to help me to System administration of our servers. Yeah, I, you know, use it for everything. You know, I kind of live in it. I'm either writing stuff to make it better, or I'm using it to use AI. So, yeah, definitely I feel like we are more, I can tell we're more advanced users of AI than anybody else, because people can keep complaining about all the problems with AI, and I'm always like, I don't have any of those problems, you know. And because we just use this really different approach where the the human is considered a vitally important part of the process to

AI assessment note: “No, it's, it can't be just for code”

Q The Kaggle journey then led to the creation of fast AI, right? I mean, that in some ways sort of feels a continuum of the submission of democratization and, and teaching. Is that, is that fair?

A So I was the second person At Kaggle, and was the chief scientist and president, so I kind of, I rewrote all the software platforms from scratch, and then Anthony and I raised money for it, and together we founded Kaggle Inc., you know, the American company, just the one that ended up becoming Kaggle Kaggle. Uh, we created this kind of, uh, motto, um, making data science a sport. And, you know, what we wanted to do was to highlight the world's best data scientists, make them rich and make them successful and make, give them accolades and make people want to be like them. And it worked really well. Uh, you know, we got, we, we worked hard on the PR side. We got top competitors into the press and into TV. And yeah, really then, yeah, fast AI was kind of an extension of that to be like, okay, we want everybody. To be able to use these technologies and improve their lives with them and improve other people's lives with them.

AI assessment note: “yeah, fast AI was kind of an extension of that”

Q all the universe of projects that you could be picking, uh, you, one of the guiding criteria is to build stuff that enables you to go faster. Is that, is that the key principle? Is that only one of the principles, I guess, The broader question is that in a universe where if you're going to build 5000 applications or more, you could be doing anything. How do you choose?

A It's not far away from it. It wouldn't be quite how I describe it, but it's not bad. So we have this idea we call the substrate, which is to say, okay, in a normal company like GE, you want to expand, you want to create 5000 products. You do that by hiring more people, creating departments, setting up offices in different geographies. So that's a very recognizable thing that you build up, you know, a company. Um, you know, the multinational was a bit of a different kind of company to perhaps what, you know, kind of industrial conglomerate, but so they kind of innovated a bit on that idea of like, what, what is it? That's doing this work and producing these products. So we're not going to have that if we've got 12 to 14 people. We're not going to have offices and departments. We don't, you know, we don't have any roles. We don't have any professional managers. So the substrate for us will be kind of the AI and the automation and the connectivity to the people, you know, iterating with that. So things that contribute to the substrate are the things that We particularly prioritize, but not in an abstract way, you know. So for example, you know, it's all about solving problems as they kind of come up, you know, it's very much a lean startup approach. Um, so for example, not having professional managers is a bit unusual and, you know, we've naturally faced this problem of people som…

AI assessment note: “things that contribute to the substrate are the things that We particularly prioritize”

Q But for the rest of the world, like you guys are deep, deep experts, uh, but for the rest of the industry, where do you think we are?

A I think that's all that matters though, right? Like it's, it's always the early adopters that tell you where everybody's going to be in two years time. So it's also interesting that all, you know, all the best open sources coming out of China, Now, so they seem to have a much more, kind of, pro-social approach to this, whereas, oddly, the US companies that used to lead the way have all drawn up the, the drawbridges, and that's gonna cause China to keep moving faster, because when you're in that more, both collaborative and competitive environment, you just Go way ahead, as we've seen in every other area of open source software over the last 30 years. Yeah, I think it's going to keep, keep going in that direction. I will say the one Slight outlier is Google who've really come out in front.

AI assessment note: “it's always the early adopters that tell you where everybody's going to be”

Q Do you think that's a world where, where, uh, that's a temporary issue, you know, a little bit related to the earlier part of the conversation about, uh, AGI, SAI, or whatever, um, is there a world where, uh, I don't know, the underlying models improve so much that, uh, this ends up actually working?

A I don't really make predictions, you know. I've just always liked to make things based on, kind of, where we are now, what works now, and what's You know, in a kind of a very direct extrapolation to where we might be in like a year. That's always worked pretty well so far in my career. Um, like it helps, like, understand the details of the technology well, so I kind of, these things like DeepSea Car One or whatever don't seem, they don't come out of the blue, they don't seem like wild jumps or whatever, you know, I can, you can kind of, they're all part of the same path. Well, obviously AI models will be able to do Larger and larger pieces, more and more stuff, make less and less mistakes, because they're getting better. Does that mean we'll get to a point where humans have no useful input to provide at all? That's the same as saying, well, we get ASI, and if that happens, then the world's a very different place. There's no point in me doing anything. I have no particular reason to believe that will happen, and for every point until we get there, if we ever get there, by definition, there's going to be humans interacting with AI, and so that's what I care about, is how do we do that In an optimal way. And I feel the optimal way is to have the humans and the AI in the same environments, sharing, sharing that where they can both then bring the bits that they're best at.

AI assessment note: “I don't really make predictions, you know.”