The Best of 2024 (so far) with Sarah Guo and Elad Gil

0:07Hi, listeners. Welcome back to KnowPriors. We're halfway through 2024, so we're doing a mid-year best of episode where we go back to some of our favorite moments from episodes so far and catch you up on everything that's been going on in AI, from the state of the art in research to hyperscalers and upstarts. We'll list all the episodes featured so you can go back and re-listen to the whole conversation. To kick it off, we're going to hear a little bit from Emily Glassberg-Sands, who's the head of information at Stripe. We talked a lot about how AI can help small businesses make a big impact in the economy.

0:39Here, she talks about the intersection of fintech and AI. When you think forward on the directions that the overall financial services industry is going, and let's put Stripe aside for a second, because I think Stripe is obviously a core company to sort of the internet economy, and it touches so many different pieces of fintech and things like that. But where do you think outside of Stripe, the biggest white space for fintechs employing AI is? like from a startup perspective or even an incumbent perspective? Like where do you think this sort of technology will have the biggest impact? It's a great question.

1:14And I don't know exactly what others will do. I think having a really robust understanding of identity, who businesses are, what they're selling has always been important. And, you know, I think often in industry, we think it's important for marketing or sales or sort of go to market motions. But it's also super important in fintech. Yeah, it's important for credit lending decisions, but it's also important for supportability decisions and understanding where, you know, the business does or does not meet the requirements of a given card network or a given bin sponsor. And so I think that that identity piece, like who is this merchant?

2:04Are they who they say they are? But also, what are they, what's their business? What are they selling? And how does that map to this pretty complicated regulatory environment is a really interesting and hard problem that lots of folks are solving in their own ways, but is likely an opportunity. I think there's Almost certainly an opportunity to, you know, whether Stripe does it or somebody else does it, to make sort of financial integrations way more seamless. Stripe has a whole suite of no-code products, so you can use, you know, payment links or no-code invoicing. But how does one actually build a really robust, specific to the user integration without needing, you know, a substantial number of payments engineers or any complicated developer work?

3:05LLMs are proving that they can be very good at writing code. We have a couple of cases actually where we're already seeing it work. But as the decisions get more and more complicated, I think there's still a lot of work to do to build the right integration and to build it well in an automated way. And then I think, as I mentioned before, some of this layer on top of the payments data, it's like, okay, you could build solutions that make payments work better, but payments actually allows you to really deeply understand and improve the business is pretty fascinating. And you'd have to think about like, is it a startup that does that or is it an incumbent that does that?

3:48And what's the business model there? But if I think about the case of Stripe,

3:59Stripe has the opportunity to be beneficent, right? Incentives are super aligned. The more Stripe can help its users, businesses grow, the more Stripe grows and the more the economy grows. And so whether it's Stripe or someone else using financial data to help businesses be more successful, to grow the pie, to grow the GDP, I think is really powerful. Up next, we have a clip from our conversation with friend and formidable founder, Dylan Field, whose company is using AI to change the design process and bridge the gap between design and development. We talk about how bringing AI into the creative process changes the creative job.

4:39Basically, you're moving from a human-to-human collaboration company to a human-to-AI collaboration company over time in some sense, because, you know, what you're describing seems like a really interesting way to have co-pilots augment humanity or augment creativity. Are there other ways that you've thought about the substantiation of that sort of creativity augmentation or how AI really interacts with human creative potential? Well, and these are just examples of things that I have seen or thought about that I think could be cool in the creative space because you asked about. But I think in the design context, one thing that really matters a lot is the iterative loop and being able to keep going back and forth to an agent and give more instructions over time.

5:25If you just kind of like go to first principles here, there's so much that you're not able to communicate via a prompt. Like if you think about great design, it often captures something about the culture, the ethos of the moment. It captures something about the temporal aspect of the sequence of interaction someone's having or the context they will have mentally. Something about affordances, what people are used to in terms of the language of design, which is sometimes similar and dependent on the platform. But oftentimes there's something about emotional state too. uh there's you know videos that the designers probably watched or or in-person research interviews they've conducted and so i think like fitting all that plus the product requirements um plus visual style into a prompt that's hard even if you could just get unblocked by an ai helping you brainstorm and thinking through problems you know that's your first sort of draft and from there you can keep iterating from there you can keep evolving things i think that could be very, very interesting as a, as a first step.

6:32What's your response to people who worry that, um, AI like in every role are going to, you know, eliminate the need for designers. For all the reasons that is mentioned around, you know, emotions, uh, user context, uh, knowing how flows go, um, having that history of interactions and whatnot. I think it's unlikely that that's like the world we're seeing in the short term. I think no one knows what's happening in the long term. You know, if we have, you know, superhuman intelligence, like I don't know what it means for any of us on this call podcast or anyone listening. If we don't try to ask about what that case looks like and instead ask about, okay, if we assume that there's continued improvement, what does it mean for design?

7:24I think design's actually in a really good place. probably before you see potential replacement of any part of the design role, you instead see augmentation and you see access. You see efficiency so that designers can get more done. And I think probably a lot of engineers put more of their time towards design than they put towards what we consider coding tasks today. And the abstraction level of coding changes. There's probably still a human loop for engineering, but I think that it's not clear to me that humans are going to write like every line of code in a year, three years, five years. So obviously already we have Copilot, but I think that you could go even further than that.

8:03And a lot of companies are trying to do that. I mean, I can't make multi-year bets in the current environment, but my expectation would be that we, maybe it's because I'm an optimist, but I think we're just going to get better and more software and better designed software versus fewer designers or engineers. Yeah, I definitely think that as a metric, like number of pieces of software that will be created will go up tremendously. And it's interesting, like there's some visions out there of the future where people interpret the capabilities of AI to mean that you won't like have any interface at all.

8:44I think it's really cool to see this explore, like the rabbit we've talked about, Sarah. I haven't used it yet. I think you did. Is that right? But I think it's a really cool vision. And I think that there will be so much more software in a year, two years or five years from now than there is today. Like both could be true that there's demand for that. And there's just way more software. Next, we talk to Brett Adcock, the CEO of Figure AI. Figure is creating a fleet of human-aid robots to take on the dull and dangerous jobs that humans shouldn't be doing. In this clip, we talk to Brett about how he runs a team with Velocity to drive hardware, software and AI into reality.

9:23Big question, but can you describe like if you want to run a hardware project, a hardware and software project like this with this complexity at velocity, like how do you manage product development? From like a thesis perspective, I strongly believe in like an iterative design approach. We really don't believe on spending a lot of time like just doing research and analyzing. We spend a lot of time on just testing, building the testing here. And, um, that helps us really shake out all the problems. It helps us learn helps us recursively add it into a continuum of product that's coming down, uh, coming out.

9:57And, um, so first that's our strategy. We, um, we want to be continuously updating the hardware and software forever. It'll, I don't think it will ever be good enough for us. Um, so we have a whole process built around building a robot from a, like a basically hardware and software design that we run here. We first set out with understanding who are the customers, like what does a robot need to do? From there, we, uh, we basically set requirements like, okay, we need the robot to lift this much pounds. It needs to run this long and needs to charge here. The safety requirements are that it can't battery can't burn down the building.

10:41And there's like a bunch of stuff we have to, um, the environment on IP rating has to be done on the actuators. There's just a bunch of requirements that come from there. from there, we look at those requirements and we do engineering design. And we have basically like three big phases. We have a conceptual and preliminary and critical design review that we do here throughout the year. The whole company is involved. So we have these like design gates that we work through. Similar practice that I instituted. Exactly similar. Well, similar practice. I instituted Archer from an engineering design perspective or philosophy.

11:11And yeah, we work through it in a very methodical way, like all the way through that serially. And how does integration and testing work in a way that's different from a software company since you've also done that? I imagine really differently. Yeah, we try to test and we try to prototype and test as fast as we can to see if we're right. Same with software. It just happens on a longer timeline. Okay. Well, software, you'll come in one day and I'll say, okay, we talked to the client. We believe the client. We talked to the client. We believe we have all these things on the product backlog list we want to do.

11:43you'll somehow have some heuristics where you'll score those and you'll basically comb the backlog and you'll say, I'm going to go, we're going to add these like six things to the sprint. They'll do story points and you'll basically, you'll, you'll assign those out and you'll basically manage that whole process. And then you'll launch it and you'll get feedback, right? You'll try to either A-B test things, you'll watch the analytics and you'll say, did that work? Did that work? You really want to do that. And you want to have that kind of a scientific method around it. Say like, okay, was that, did that actually help, you know, fix this problem?

12:12uh, same here. We have the client, we have requirements that we set, like they need to do this. We are designing things. Like we are designing hardware from scratch. Like, and, um, so we take our, we're designing an actuator. We're going to take our CAD system and we're going to, from scratch, design it. We're going to make, uh, assumptions on and trade studies on like what the different trade-offs are of how we can do it up front. So we don't spend a lot of time designing something that just didn't work. Uh, so we're going to be pretty methodical about it. Like much more methodical than you are software because the timelines are, you know, order of magnitude plus longer.

12:42Up next is a snippet from a conversation we had with the OpenAI research team building Sora. Here, we talked to this team about their generative video model and whether or not video is on the path to AGI. Do y 'all have a favorite thing that you've seen artists or others use it for or a favorite video or something that you found really inspiring? I know that when it launched, a lot of people were really stricken by just how beautiful some of the images were, how striking, how you'd see the shadow of a cat in a pool of water, things like that. But I was just curious what you've seen sort of emerge as people, more and more people started using it.

13:18Yeah, it's been really amazing to see what the artists do with the model because we have our own ideas of some things to try, but then people who for their profession are making creative content are like so creatively brilliant and do such amazing things. So Shy kids have this really cool video that they made this short story uh uh airhead with um this character that has a balloon and they really like made this story and there it was really cool to see a way that sora can unlock and make this story easier for them to tell and i think there it's even less about like a particular clip or video that sora made and more about this story that these artists want to tell and are able to share and that Sora can help enable that.

14:05So that is really amazing to see. You mentioned the Tokyo scene. Others? My personal favorite sample that we've created is the Bling Zoo. So I posted this on my Twitter the day we launched Sora. And it's essentially a multi-shot scene of a zoo in New York, which is also a jewelry store. And so you see like saber-toothed tigers kind of like decked with bling it was very surreal yeah yeah and so i love those kinds of samples because as someone who you know loves to generate creative content but doesn't really have the skills to do it it's like so easy to go play with this model and to just fire off a bunch of ideas and uh get something that's pretty compelling like the time it took to actually generate that in terms of iterating on prompts was you know really like less than an hour so i get something i really loved um so i had so much fun just playing with the model to get something like that out of it and it's great to see if the artists are also enjoying using the models and getting great content from that.

14:59What do you think is a timeline to broader use of these sorts of models for short films or other things? Because if you look at, for example, the evolution of Pixar, they really started making these Pixar shorts and then a subset of them turned into these longer format movies. And a lot of it had to do with how well could they actually world model even little things like the movement of hair or things like that. And so it's been interesting to watch the evolution of that prior generation of technology, which I now think is 30 years old or something like that. Do you have a prediction on when we'll start to see actual content, either from Sora or from other models that will be professionally produced and sort of part of the broader media genre?

15:35That's a good question. I don't have a prediction on the exact timeline, but one thing related to this I'm really interested in is what things other than traditional films people might use this for. I do think that maybe over the next couple of years we'll see people starting to make more and more films, but I think people will also find completely new ways to use these models that are just different from the current media that we're used to. Because it's a very different paradigm when you can tell these models kind of what you want them to see, and they can respond in a way, and maybe there are just like new modes of interacting with content that like really creative artists will come up with.

16:14So I'm actually like most excited for what totally new things people will be doing that's just different from what we currently have. It's really interesting because one of the things you mentioned earlier, this is also a way to do world modeling. And I think Aditya, you've been at OpenAI for something like five years. And so you've seen a lot of the evolution of models in the company and what you've worked on. And I remember going to the office really early on, and it was initially things like robotic arms and it was self-playing games and things or self-play for games and things like that. But as you think about the capabilities of this world simulation model, do you think it'll become a physics engine for simulation where people are, you know, actually simulating like wind tunnels?

16:50Is it a basis for robotics and uses there? Is it something else? I'm just sort of curious where some of these other future forward applications that could emerge. Yeah, I totally think that carrying out simulations in the video model is something that we're going to be able to do in the future at some point. Bill actually has a lot of thoughts about this sort of thing. So maybe you can. Yeah, I mean, I think you hit the nail on the head with applications like robotics. You know, there's so much you learn from video, which you don't necessarily get from other modalities, which companies like OpenAI have invested a lot in in the past, like language.

17:23You know, like the minutiae of like how arms and joints move through space. You know, again, getting back to that scene in Tokyo, how those legs are moving and how they're making contact with the ground in a physically accurate way. So you learn so much about the physical world just from training on raw video that we really believe that it's going to be essential for things like physical embodiment moving forward. Up next, we talk to Scott Wu, the co-founder of Cognition, the company behind Devon, which is building an AI engineer. Here, we talk about the design for Devon and what it means to work with AI engineers.

17:57What do you think is going to be important from a human software engineer or just like human technology person? five years from now. I realize that's a really long timescale in AI, but it's certainly not like encyclopedic knowledge anymore. Right. Yeah. Yeah. And I mean, I think there's, there's, there's a meme that, you know, the hottest new programming language is English. Right. And I mean, I think there's a lot of truth to that. But with that said, I think that, you know, the software engineering fundamentals are obviously still super, super valuable. Right. People, you know for example, like I think, you know, the internet today is something that we all kind of are able to use and kind of take for granted, but people who work with these networks, it's certainly very helpful for them to understand the details of TCP, right?

18:40And I think similarly, I think, you know, I think we'll be able to communicate our ideas in English and work with all these things, but, you know, understanding the internals of how computers work and understanding logic gates and, you know, a lot of these core pieces, like these core foundations, I think will still be very useful. And so whether that's algorithms or technologies or logical reasoning or things like that, I think the role of a software engineer five or 10 years from now, it looks something like a mix between a technical architect and a product manager today, where a lot of what you do is you take problems that you're facing or that your business is facing or whatever.

19:25And you're really thinking about and breaking down what exactly the solution should be. How do you think about it in an even farther timeframe? Because when I, if it was five years ago, I would have told either my kids or people who have kids, you know, you should study computer science and math. 20 years from now, I'm not as certain. So I'm sort of curious how you think about the future of this field, if much or all the work, including a lot of the planning is actually done by machines at some point. Yeah. I mean, I love that. So I have to say it's a worthwhile experience, even if it doesn't end up being practically useful.

20:02But no, I mean, I think a lot of these fundamentals will stay useful for a long time. There's obviously a lot of questions that come up about, you know, super intelligence and singularity and all of this. And, you know, it's very hard to predict. I think everyone in AI, it's, you know, we've all made our own predictions and, you know, tried to make our guesses. But I think it's hard to be very high confidence. But with that said, I do think that we're going to see AI's concrete impacts on work and economy and people's lives, I think, a lot sooner than that. I think the way that we think about the problem is that even with the tools that are available today and the technologies that exist today, there's so much that's possible to really impact people's lives.

20:52And we're still very, very early in this whole AI revolution. I mean, even ChatGPT was about a year and a half ago at this point. And there's a lot more to do and a lot more to build, you know, both on the research side and on the product side. Finally, we have Alex Wang, the founder of Scale AI, the data foundry for AI. Here, we talk about what's next for Scale as models approach and go beyond human abilities.

21:20With great power comes great responsibility. if these AI systems are what we think they are in terms of societal impact, like trust in those systems is a crucial question. Like, how do you guys think about this as part of your work at scale? A lot of what we think about is how do we utilize, how does the data foundry enhance the entire AI lifecycle? Right? And that lifecycle goes from, you know, A, ensuring that there's data abundance as well as data quality going into the systems, but also being able to measure the AI systems, which builds confidence in AI and also enables further development and further adoption of the technology.

21:57And this is the fundamental loop that I think every AI company goes through. They get a bunch of data or they generate a bunch of data, they train their models, they evaluate those systems, and they sort of go again in the loop. And so evaluation and measurement of the AI systems is a critical component of the lifecycle, but also a critical component I think of society being able to build trust in these systems. You know, how are governments going to know that these AI systems are safe and secure and fit for, you know, broader adoption within their countries? How are enterprises going to know that when they deploy an AI agent or an AI system that it's actually going to be good for the consumers and that it's not going to create greater risk for them?

22:39How are labs going to be able to consistently measure what are the intelligences of my of the AI systems that we build and how are we going to you know how do they make sure they continue to develop responsibly as a result can you give our listeners a little bit of intuition for like what makes evals hard one of the hard things that you know because we're building systems that we're trying to approximate and and build human intelligence grading one of these AI systems is is not something that's very easy to do automatically and it's it's sort of like, you know, you have to kind of build IQ tests for these models, which in and of itself is a very fraught philosophical question.

23:16It's like, how do you measure the intelligence of a system? And there's very practical problems as well. So most of the benchmarks that we as a community look at for - The academic benchmarks. Yeah, the academic benchmarks that are what the industry used to measure the performance of these algorithms are fraught with issues. Many of the models are overfit on these benchmarks. They're sort of in the training data sets of these models. And so... You guys just did some interesting research here. Yes. Published some. Yep. So one of the things we did is we published DSM-1K, which was a held out eval.

23:46So we basically produced a new evaluation of the math capabilities of models that there's no way would ever exist in the training data set to really see how much of the... How were the performance of the models... What were the reported performance of the model capability versus the actual capability? And what you notice is some of the models perform really well, but some of them perform much worse than the reported performance. And so this whole question of how we decide we're actually going to measure these models is a really tough one. And our answer is we have to leverage the same human experts and kind of the best and brightest minds to do expert evaluations on top of these models to understand, you know, where are they powerful, where are they weak, and what are the sort of risks associated with these models.

24:30So, you know, one of the things that we're very, you know, we're going to, we're very passionate about is there needs to be sort of public visibility and transparency into the performance of these models. So there need to be leaderboards, there need to be evaluations that are public that demonstrate in a very rigorous scientific way what the performance of these models are. And then we need to build the platforms and capabilities for governments, enterprises, labs to be able to do constant evaluation on top of these models to ensure that we're always developing the technology in a safe way and we're always deploying it in a safe way.

25:03So this is something that we think is, you know, just in the same way that our role as an infrastructure provider is to support the data needs for the entire ecosystem. system, we think that building this layer of confidence in the systems through accurate measurement is going to be fundamental to the further adoption and further development of technology. Thank you all so much for listening. We've really enjoyed talking to people reshaping our world with AI. To listen to any of the full episodes, please find the links in the description for this podcast. And we'll be back with new interviews next week.

25:37Find us on Twitter at NoPriorsPod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts for every episode at no-priors.com.