OpenAI CSO Jason Kwon - power, policy and the race to scale

Jason Kwon is the Chief Strategy Officer at OpenAI. In this episode of World of DaaS, Jason and Auren discuss: Compute bottlenecks and power infrastructure challengesUser expectations vs AI capabilitiesGeopolitical competition in AI developmentBringing top AI talent to the United StatesLooking for more tech, data and venture capital intel? Head to worldofdaas.com for our podcast, newsletter and events, and follow us on X @worldofdaas. You can find Auren Hoffman on X at @auren and Jason Kwon o...

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:02Auren Hoffman?

  Hello, fellow data nerds. Welcome to World of DAS. I'm your host, Orin Hoffman, CEO of Incubate and GP of Flex Capital. Discover more episodes, get weekly data as a service news, original content, articles on data and more at worldofdaas.com. That's worldofdas, D-A-A-S, dot com. Hello, fellow data nerds. My guest today is Jason Kwon. Jason is the Chief Strategy Officer at OpenAI. Jason, welcome to World of DAS.

• 0:31Speaker 01

  Great to be here, Orin.

• 0:32Auren Hoffman?

  I'm really excited. If you think of like, what's super important for AI, there's this debate, like, it could be compute, it could be data, it could be algorithms, it could be regulatory things, like, where do you fall on what's the most important thing for AI now?

• 0:46Speaker 01

  I think this kind of depends on what level that you're asking this question at. Are you asking at the industry or country level? Or are you asking at the level of the individual organization? At the country or industry level, I think it's probably compute. And that's because just the breadth and diversity of experiments that you can run on research, which is what powers everything, you need compute for that. And the more compute you have, the more experience.

• 1:08Auren Hoffman?

  And there's just a bottleneck on compute?

• 1:10Speaker 01

  Yeah, I think to some degree there is, but the market is trying to respond to that demand signal. And even like the scale up of inference that you want in order to benefit from the advanced capabilities in the models, because the reasoning paradigm has definitely shown that there's a scaling law that also applies, the amount of test time compute that you apply. Compute is going to matter both for advanced capabilities and also making the most of the capabilities. And this is at an industry level.

• 1:33Auren Hoffman?

  And by compute, would you couple that with like, things like power and stuff? Or is it like, one big bundle?

• 1:39Speaker 01

  Yeah, so I would think of the compute equation as the capital and power that you need to build the physical infrastructure to make use of the AI infrastructure, the models and software. There's another level at which you could ask this question, which is at the level of an organization. There, I think the answer is a bit different. And this is where you can get a lot of debates, is because people might be thinking about this in terms of different levels. So which level of generality are you talking about? So at the organization level, I don't think it's necessarily compute. And that's because if you just reduce compute to capital, it's just money and then you buy the physical infra. We don't necessarily assume in lots of other industries, or even in technology industries, that if you're just the most capitalized company. Or organizations that you're automatically going to win.

• 2:24Speaker 01

  It's how you make use of that resource and apply it to various bets. And so

• 2:28Auren Hoffman?

  Often it's the opposite, you know, the most capital like something and someone comes up with something very innovative to respond to that.

• 2:35Speaker 01

  Completely, right? And so it's just the scarcity that sometimes creates the innovation, right? And so bet selection, I think, you know, in terms of research bets, that matters a lot at the organization level. And then having some kind of organizational capacity or or structure that enables you to make those bets well and sustain them. And have like the right taste and selection criteria for them. That is really important. And I think you can think about that when you look and zoom out at various labs. What types of bets are they making? How are they set up competitively in relation to each other? And what's their cultural aspects tied to those things?

• 3:12Speaker 01

  I think that's an interesting analysis, but and that's why at at an industry level, it's different because it just that analysis just benefits from the aggregation of all the individual organizations.

• 3:21Auren Hoffman?

  Where do you think a country's power might become the bottleneck? Or where do you think like some of these other things might come to the bottleneck to make these things happen?

• 3:28Speaker 01

  There's certainly regulatory bottlenecks. You go from country to country, it's hard to get permitting to build new things. And depending on which country that you're in, there is certainly power. That's a bottleneck, which is related a little bit to the regulatory bottleneck too, because if you have more friction on that process, it's harder to build up the capacity faster. And certain countries are more renewable energy, pro nuclear, and other countries are a little bit more conservative about these things. And that also impacts the capacity on a power side. And so one interesting anecdote here is when we did a bunch of surveys around the energy question, sort of around when we were thinking about Stargate.

• 4:05Speaker 01

  It's really most of the world was building out energy capacity without taking into account the growth of AI, which is, you would assume is is actually a pretty rational thing to do. A couple of years ago, because people weren't planning on this. And so the amount of base load capacity people were planning on building over the next decade or so, it was just key to what they

• 4:24Auren Hoffman?

  Based on like their population growth and a few other things, electric cars are coming on.

• 4:28Speaker 01

  Yeah, it's just base rate of whatever economic growth they're expecting in relation to demographics and. You know, whatever they're already seeing in their economy. And so it was already not incorporating what power demands AI was going to bring. So, and then you take into account that was probably 18 months ago. And certainly a lot of people updated since then, but that still takes time to work through the system to get to adjusting the planning cycles around, you know, new power generation capacity building. A good data point that drives us home is the amount of total capacity in the U.S. is about 1,200 gigawatts or something like that. And if you have some projections, I think there are some that say we'll need 50 gigawatts of additional capacity by 2030.

• 5:13Speaker 01

  That's like around 5% of our total nationwide capacity and that's significant. Then you take into account how much new gigawatts capacity that power capacity that gets built. It's a lot less than 50 per year. I don't know what the latest estimates are, but just getting a gigawatt cited and built in our experiences, that's a process. And it's going to take probably 18 months or so to complete that.

• 5:37Auren Hoffman?

  I assume with like, if you're training AI models, you could almost guarantee you always want this power 24/7. Whereas for most power, it's like even a factory, you're going up and down or, you know, or some other type of thing. Whereas I assume with the training model, you could just say, I don't know if that's true. Like you could say, I want this all the time and I will pay for it all the time. And I can guarantee you this thing, or is it quite variable? Like there, there's intense training where you need it and then it kind of like levels off or something.

• 6:10Speaker 01

  I think there's probably a difference between training and inference. You probably want something closer to constant uptime for inference, just because you want to be able to serve. Training, you may be able to segment more because, you know, people have to prepare the training run. There's compute allocation and things like that, but. I think the other factor that kind of pressures on this is people are always looking for more compute capacity. And our experience has always been that whenever we have decided not to acquire some compute capacity, we later on regret it. And I'm sure we're not the only lab. And that just kind of indicates that even if you could sort of do this load balancing that you're talking about, it's still at the end of the day, you want more compute. So.

• 6:50Auren Hoffman?

  But if we have a point where compute's not the bottleneck and power is the bottleneck, or, you know, we can't like way more efficient chips, or who knows what will happen over the next X number of years. Is there also a point where you could be, okay, we're just going to do this inference when there's extra capacity on the grid? Like, there's points right now, even in California, where the energy price goes negative, because there's just not enough people wanting to take from the load, even today, that's happening. So could you have some sort of like start and stop type of thing? Or is it just too hard? Like, we got to run this thing for like a week at a time before where, and, you know, it's hard to plan around that?

• 7:31Speaker 01

  I think it would be probably difficult from the training side. Something that would potentially be possible is if you're running a very large inference job, that's going to take a lot of compute. And I think this is going to be increasingly a type of workload that we see over some time span because of that inference scaling law that we were just talking about. So I think most people's typical day to day experience with AI like ChatGPT or Claude or Gemini is you ask it a question, it responds instantly or within a couple of minutes. But if it is true that you're going to get better answers and better results on high value problems by running compute inference compute, you could imagine a world where you're running it for days and people are willing to pay that much.

• 8:14Auren Hoffman?

  And then you could also price it in a way of like, hey, if you want to wait a little bit longer, we could optimize the when it's the better power or when when these chips are more available or whatever it may be.

• 8:24Speaker 01

  Yeah. So I think that there you might see some interesting economic patterns emerge. Don't know yet, though.

• 8:30Auren Hoffman?

  And if you're only running it for hours rather than days, there might be times of the day where it might actually be cheaper to to run that thing or something like that.

• 8:38Speaker 01

  Yeah.

• 8:39Auren Hoffman?

  A lot of these things are built on like the Nvidia roadmap. There are some folks in the space like Google that have more custom hardware. Like, how do you see that as a, there's kind of this codependency with OpenAI and some of the other labs? Like, how do you see that playing out?

• 8:56Speaker 01

  I think the simple equation is just how much capability do you get per watt? At some level of reliability? And whatever inputs into that makes sense from a GPU perspective, whether it's Nvidia or custom, that's always what you're kind of working with. And then on specifically on the Nvidia question, it's just, they've been a market leader for a really long time, and it's a well-deserved position. And so I think it's not just us, but a lot of other labs are buying as much of it as they can, because the quality is so good in relation to that equation.

• 9:35Auren Hoffman?

  The key equation is per watt.

• 9:37Speaker 01

  Yeah, how much like GPU capacity or capability are you getting?

• 9:40Auren Hoffman?

  And how does the cost break down? Let's say, like, over the life of a chip, how much of it is power versus how much of it is the actual chip?

• 9:48Speaker 01

  There, I'm probably not the best expert to comment on the exact actual composition. I'd have to go talk to somebody and be like, hey, how does this exactly break down between energy?

• 9:57Auren Hoffman?

  Probably ask ChatGPT the question.

• 10:01Speaker 01

  We'll talk about this, you know, at some point, probably later in the conversation too. But the nice thing about ChatGPT now is you can plug it into the org, and it's kind of like walking down the hall and talking to one of the experts too.

• 10:10Auren Hoffman?

  ChatGPT recently released GPT-5. Where in that realm are you most excited about?

• 10:17Speaker 01

  I think that there are two aspects to this. One has to do with accessibility, and the other has to do with really deep use cases. So on the accessibility question, one of the things that's, I think, not obvious to a lot of people is that most of the user base of ChatGPT has been not on state-of-the-art models for a while, using 4.0 or variants of it. And they're not, they haven't been really interacting with O1 or O3.

• 10:40Auren Hoffman?

  And that is because it's just takes too long or the cost is too high?

• 10:45Speaker 01

  Model picker, I think, and the default generally being 4.0.

• 10:49Auren Hoffman?

  Also, sometimes you're like, do I really want to wait longer for this answer? Probably won't good enough. It was hard to know. Yeah.

• 10:55Speaker 01

  Yeah. And so especially if you are a casual user of ChatGPT, it's sort of like, what is this model picker? What is this naming convention, which we've been teased about a lot? The fact that you have this interface now that sort of helps kind of automatically route. And then some of that routing goes to these reasoning models. That is, I think, the first time a lot of people have exposure to this. And I think that accessibility of really this latest paradigm that was uncovered last year is really making its way to most people this year. And I think that's one of the things. And so it'll just be interesting to see what emerges out of 700 million people interacting with that capability. The other thing is a lot of common usages of ChatGPT have to do with coaching. Like, help me with this, walk me through this particular task, you know, et cetera.

• 11:40Speaker 01

  And it's good at that. But now you can have ChatGPT go and do some very deep research, as the name of the product would suggest, but also get into some pretty extended dialogues or conversations on, you know, whether it's health questions, legal questions, finance questions, technical questions. It can go pretty deep. And then you can also have it work on a problem, walk away and then come back and it'll have a bunch of answers for you. And another example of this is on our health task benchmarks, it scored the highest out of all our models. So it would not shock me if it does better on all these other professional tasks as well that we don't necessarily have evals for. So I think what that also reveals is you have this breadth of accessibility.

• 12:24Speaker 01

  But then if you go down into specific domains, you can, I think, now go even deeper than before because of this combination of it's a better base model. It's got reasoning embedded into it. It's got model routing automatically to kind of make the cost reasoning, quality and trade off optimal for you.

• 12:42Auren Hoffman?

  And some of the API costs have gone down pretty dramatically too.

• 12:46Speaker 01

  Yeah, that's right. I think all of these things will produce lots of new interesting use cases, just like if you go from GPT-3 to GPT-4, there was a bunch of new software applications and new startups that were formed around these tools that couldn't have existed, I think, during the GPT-3 timeframe because, you know, it just wasn't deep enough, wasn't broad enough, and it costs too much for the things that you're getting out of the model.

• 13:13Auren Hoffman?

  There are certain cases where ChatGPT still just hallucinates so much, and it's so frustrating for users. I understand broadly why it happens, but why is it so hard to fix?

• 13:27Speaker 01

  One of the reasons is user intent itself is ambiguous. Sometimes I could ask it, describe for me the principle of free speech. You could do a pretty good job there, but what was my intent? Did I want like a full-on legal analysis? Did I want something casual? Did I want a joke? Was I being sarcastic?

• 13:44Auren Hoffman?

  But even with the intent, you probably don't want it necessarily to tell you like a factual error or something.

• 13:50Speaker 01

  This is true, but it still triggers the question of should I be engaging system one or system two type thinking? And, you know, it's a little bit like when you talk to a person, like there are more errors when they do system one thinking, and this deliberative kind of reasoning does reduce the error rate. And then there's also, the model has been trained to go look up information so that the inferences are not ungrounded anymore when there is the ability to look up information and reason from that. The training on that is also dependent on understanding user intent signals. And this is why I think that you see variation in responses in the terms of, you know, user experience, where some people say, this model is great, and other people say, it's not working for me. I think

• 14:31Speaker 01

  it probably has to do with how they are asking questions or prompting it, and what was their pattern of usage before versus now in relation to the model capabilities. And so if you were kind of using it as a coach before, and it did very well with ambiguous questions, in this latest iteration, it might do exceptionally much better at requests that are very well-defined, but when they're very, very vague, it might actually do something different. It's this kind of this junk between expectations and how the model's actually processing, you know, what you want to do. And I think all these things will continue to improve because... You know, this is the first time we've kind of shipped something like this in this format. And this is probably as, you know, I think Sam likes to say whenever we do new things, that this is going to be, this is as bad as it's going to be. It's only going

• 15:15Speaker 01

  to get better from here. We're just going to learn more in terms of user intent signals and all of that. And the training will continue.

• 15:24Auren Hoffman?

  World of DAS isn't just a podcast. World of DAS membership is a private invite-only community for founders and executives of important data and AI businesses. It's a place for off-the-record, high-trust conversations focused on the biggest questions data leaders are facing. The community is about peer learning, curated intros, and access to private events. Everyone in the group is building or leading something meaningful in data. If that sounds like you, apply to join at worldofdaas.com. That's world of D A A S dot com.

• 16:05Auren Hoffman?

  One of the things that is so weird about this whole AI moment is that, as a user, my expectations for what it should do keep going up very, very fast. In fact, my expectations go up as fast as it actually is improving. Whereas like normally when I interact with a product, whether it's a car, or my phone, or whatever it is, even a piece of software, my expectations of that product grow very, very, very slowly over time. And so it's really hard for the AI to even ever hit my expectations because my expectations keep

• 16:46Auren Hoffman?

  growing. Is that just like a common weird like human thing?

• 16:51Speaker 01

  I think so, actually. And it might be accentuated in this case because there are enough interactions that you have that seem like interacting with human intelligence, right?

• 17:02Auren Hoffman?

  Right.

• 17:02Speaker 01

  And then you start to expect that kind of same reliability that you're accustomed to. And then when it kind of goofs, you're disappointed. I think the pace of innovation too, has this kind of conditioning effect where in a few years, going from GPT-3 to where we are now, you know, back during the GPT-3 days, you know, as much as we were just talking about hallucinations, it was You know, there are many more back then. It didn't have the ability to look up information or process it, didn't have the ability to reason. But even back then, it felt kind of interesting and new because you could kind of have this naturalistic conversation, the chatbot, that wouldn't veer off course like the prior versions of chatbots we had seen before. And then seeing stuff continue to progress really rapidly.

• 17:45Speaker 01

  So you kind of expect things to change and improve really quickly. And so then I think your human mind just kind of extrapolates to broad progress. And then when you see kind of pockets of capabilities that are not progressing at the same pace, I think you end up disappointed. The flip side of this, I think, is this is actually why I'm one of the people that believe that this technology actually has the potential. Now, it's not predetermined, but I think, because I think we have agency in this outcome, but that it could create a lot of new jobs and new types of work. Right? So, you know, we're just talking earlier about you could get vastly different results based on how you are interacting with the system. That's a skill in and of itself that's going to, you know, develop.

• 18:28Speaker 01

  Some people have gotten pretty good understanding where are the limits of the models in terms of reliability? Or where it's likely to hallucinate versus not, and how to kind of minimize that outcome versus not. And other people understand how to apply it to various, you know, disciplines like STEM fields or, you know, healthcare and all that, and get really, really good results or use it to perform advanced analysis. But then understand how to look for the particular areas where they need to do their refinement, because they understand a little bit about where things start to become a little less reliable in terms of the results it's giving you. And so those are all skill elements that come from this tool. If it was perfect, you know, you wouldn't have that. But because it is imperfect, it is another tool that has, I think, proficiency associated with it.

• 19:13Speaker 01

  And that itself, I think, is actually quite promising and could be like a very high skilled type of work that evolves.

• 19:22Auren Hoffman?

  One thing I use ChatGPT for all the time is like random tech support things like this isn't working, or this has a bug, or this software I'm having trouble with, or my phone, I need, you know, whatever it might be. And it's quite good at most of those things. But like the one area of tech support, I feel like it's the worst, is when I ask it about itself. I often have to go to the actual ChatGPT tech support, which may be an AI interacting with me, I don't know, but it's like the separate interface where it's like I'm often sending them an email, they're sending me an email back on Zendesk or whatever they use there. Like, why can't that be incorporated into the product?

• 19:59Speaker 01

  This is, I think, another instance where the intuitive extrapolation is actually leads you astray a little bit, because what's really going on in the hood is the surface area of ChatGPT is changing very rapidly. And the information, the content around that is not that widely dispersed online, partly because it's the rate of change is so high. And then the distribution of information, because it's relatively new, isn't that why?

• 20:25Auren Hoffman?

  And are you worried about like training on your own internal stuff? Because then someone could like somehow attack it and get some weird information back or something?

• 20:33Speaker 01

  I think it's less that and more just the product features go out and we're constantly iterating. You know, the documentation help might trail a little bit behind. And then the discussion around that online is also going to be trailing too. And so when you're training these models, everybody's heard about the knowledge cutoff, that's not going to be part of the training mix because of the knowledge cutoff. And then getting the model to a place where it can handle arbitrary questions about itself and reliably go to the help site, that itself is also a training challenge, right? Because you have to get users to kind of interact to understand what they're asking.

• 21:06Auren Hoffman?

  Because it's changing so fast, it just makes it such like so much of a harder problem. Whereas the problem I'm experiencing on my iPhone, probably somebody else experienced three years ago or something like that.

• 21:18Speaker 01

  Yeah.

• 21:19Auren Hoffman?

  That actually makes a lot of sense to me. People are talking about this kind of competition with China, and the Chinese models are almost certainly like training on every copyright copyrighted thing available. At least that's the rumor. I'm not in there to go see that. And obviously, there's, there's different things about what US companies can do and what they can't do and what they're allowed to do and what can you go inside the firewall, outside, etc. Where are we going to have to evolve to stay competitive?

• 21:48Speaker 01

  I think if they are training on more than we have access to, that probably does give some kind of edge, at least for pre-training, where the volume of data still matters a lot, because it enables you to get the most out of the compute that you're applying. And the quality of the base model that comes out from pre-training still matters, even though there's a lot of work that now happens on the post-training side with reinforcement learning applied to have the model perform better in terms of responses to human instructions and commands and particular types of knowledge tasks. I think the other thing that's kind of interesting here is that if you play out this dynamic,

• 22:29Speaker 01

  where one part of the world trains less because of copyright and one part of the world trains without any guardrails, because they don't really share that perspective, the natural outcome of this competitive equilibrium is that the party that can train on the most will train on the most. There's nothing stopping them. And so if AI keeps taking off, there's actually a competitive dynamic or an advantage that occurs in perhaps content, in that it's going to be less based on when you can copyright a thing and more based on how deep, wide, and quickly you can generate new content with the assistance of AI. It could take many forms, right? It could be you have a UGC platform that just is able to generate content at a pace that we have not seen before, and that can quickly grow because of that.

• 23:12Speaker 01

  It could mean something like there's a form of entertainment where there's a core piece of content that's created, and then you're able to generate multiple variations with the user engagement, fan engagement, what have you, with the assistance of AI. And that itself is potentially a very powerful economic opportunity if you're in the content industry. So then I think it becomes very important who is leading the AI competition, because that will impact whether copyright IP has value or it doesn't.

• 23:42Speaker 01

  In the US, training on copyrighted content is considered fair use, but not if you completely repeat it. But I don't think the PRC labs think about it the same way. So if the ecosystem around AI is centered more on the US lab approach, you will have this kind of competitive equilibrium that still has guardrails around the technology when it comes to original content. But if the ecosystem shifts the other way, then you're going to have basically a tech stack and market structure around this that's probably oriented towards driving down the value of copyright IP. So I think this is the non-obvious part of this kind of geopolitical competition when it comes to the copyright question.

• 24:23Auren Hoffman?

  And obviously, there's all these internal docs that one could potentially train on. If you get access to it, I presume like Google already has access to, or Microsoft has already has access to so many of these people's internal docs and internal emails and internal communications and stuff like that. And they may or may not have the right to train on it. But if you off it in, you might have like some right to do it. How do you think all that plays out?

• 24:49Speaker 01

  Companies that have had a chance to build up large data asset pools, let's go for lack of a better word, they by default would seem advantaged and they probably are in some sense. And I think there are going to be a bunch of interesting competition questions that come about over the next several years.

• 25:08Auren Hoffman?

  Now you have to deal with all the internal lawyers that are debating it and screaming at each other too.

• 25:13Speaker 01

  Yeah, it's like a whole new layer of complexity and perhaps compliance that people don't necessarily welcome, but it could appear. If you remember about, I don't know, half a decade ago, there are all these questions about interoperability. Can you port the data from one system to another? And that is a way to trigger more competition, partly because it reduces the power of network effects and switching costs. And you might see some kind of push related to that, like similar to that around the data question.

• 25:44Auren Hoffman?

  I should be able to bring my data with me if I choose to in some sort of way.

• 25:48Speaker 01

  Or it could be that if you have a large asset pool and you are using it in a particular way for AI training, maybe you're required to share some of it. That could be an outcome. I don't know. It just, I think, depends on how some of the competition theory evolves over the next decade or so. I think there's also a very interesting question here too about there probably pools of data where you know, companies have them, but it's customer data and it seems like an engine, but it may not be because there's a massive trade off there when it comes to questions about customer trust, user trust and whatnot. I think that it'll be interesting to see how that is conceptualized, right?

• 26:29Speaker 01

  So I think that before, again, going back to like half a decade or a decade ago, those concerns kind of crystallized around concepts like privacy. I'm not sure that that's the right mechanism for something like this where it's not that your data or information is going to be revealed or that it's going to be used in a way, you know, to add target you or something like that. Potentially being trained on it has more, I think, to do with this question of, you know, what is your relationship with a company in terms of trust? This might just actually be solved by the market because if you're a company that decides to use that data in a particular way, there's probably a cost that you're going to pay when it comes to customer relationships. And maybe you decide to incur that cost.

• 27:12Speaker 01

  And in other cases, people running companies may value that so much that they would prefer not to pay that cost. And I think it may not really require much more than market forces. We'll see.

• 27:22Auren Hoffman?

  One thing that surprises me is that a lot of these AI tools are not native to the apps. So if I'm in Google Sheets, I want to do something in Google Sheets or something like that. I'm already or I'm in Excel. If I want to use like the most powerful Gemini, it's not in there and I have to like, I have to go somewhere else. I have to like copy my sheet in. I have to like say, hey, how do I do this thing? I'm trying to do this weird V lookup this thing X. And then it can then tell me, then I have to like go back or I have to re-describe it. Like, why is it so hard to get into the user experience that they're already used to?

• 28:02Speaker 01

  So it's a little bit of speculation on my part because we are not necessarily dealing with those legacy products. But I would think that's probably a combination of there's probably shipping an org chart then sheets that you're talking about, or some analog to that in some other company. Yeah, there's a PM that owns that and they like to their product to work a certain way. And then that's not necessarily the same PM that does AI integrations or AI innovation. So you get something that is a little bit kludgy. The other could be you have users who have been habituated to use it a particular way and you need to build some kind of transition and people are working out how

• 28:42Speaker 01

  to do that smoothly. And I think that this is why you see in various AI applications, this whole notion that there's such a thing as generative AI search. You would think that really at the end of the day, people want to just get to the answer. And I think that that is actually what people want a lot of the times, but a lot of people are used to the search paradigm. So you kind of have to build this bridge. And I think that that is what you see, for example, with applications like perplexity. It's a little bit of the motivation for why we have a search feature inside of ChatGPT two. It's just, we recognize that for certain, certain types of queries or interactions, people do want that old experience because they're used to it. But I think that over

• 29:23Speaker 01

  time, what you'll see probably is the evolution of all this stuff and like the 16 year olds that are playing with this technology that are not really using search, they're using AI, those behavioral patterns will change over and then you'll probably see software interfaces evolve too in accordance with that.

• 29:40Auren Hoffman?

  The chat interface is very good for many things. It's kind of good. Like if I'm talking to a human and I'm trying to get information, it's kind of that type of thing, whether it's, whether it's text or I'm typing it in or on voice or so. But there's other UIs that are also really good. Like the spreadsheet is an amazing UI that has, you know, been around for, I don't know, 40 plus years now and has really kind of dominated the way we think about things. And like our whole brain is wired in a way to think about spreadsheets in an interesting way today. And it'd be great if like there are other ways to benefit from AI beside that kind of chat interface way of thinking about it.

• 30:25Speaker 01

  I think what would be interesting, right, is you're able to respond with this is so to just to kind of riff off what you're saying. You have a spreadsheet and you're able to just respond, not with the chat, but with another spreadsheet. Maybe it'll reply back and say, what do you want to do with these two data sets? Do you want to join them to use SQL language, inner outer join, what have you? You want to remove duplicate of rows, all of that. And I think that those could be interesting avenues for, you know, people working on this to take this, which is, okay. We certainly have the ability to chat through the data now with AI. But is there a way that you can use different formats of inputs and prompting? And you see a little bit of this in that you can upload files now and interact with them.

• 31:10Speaker 01

  But what are some new sort of form factors of interaction in the prompting itself? That might take advantage of the way in which the L L M's operate. And I think, you know, we're still in early days. I think right now we're still pushing the capabilities of the models. But there are all these entrepreneurs that are fiddling with the applications. And that has been going on only for a couple of years. And so we're still early days.

• 31:34Auren Hoffman?

  One thing is in the Google search paradigm, there are all these people whose job it is is to game the Google search results to be in their interest, right? S E O, all this other stuff. And today there's just so many people already whose job it is to game like the AI outputs to be in their favor. And I imagine it's already creating this like very interesting cat and mouse game like Google's been used to for the last 25 years. From your side, like, how is that playing out?

• 32:08Speaker 01

  I don't know that we track this really that closely. I think mostly we're focused on, you know, training the best models, pushing the capabilities. And then having in the search experience it trying to return relevant results that are then reasoned through by our reasoning engine and then continuously refine there based on user signals. If you kind of answer your question from thinking from first principles a little bit, in the long run, and this might be speculative, but I think if reasoning and agent capabilities continue to advance, this gamability of search results, the way that people do it now at least, might just be really, really difficult. It might be a new type of gaming.

• 32:52Speaker 01

  But if users want to not have results gamed, and if that's what model providers also want, it should be easier to actually avoid this because you can now actually tell the system, find me something objective, avoid sites that look like SEO, you know, analyze the results you're getting back to understand if there might be a financial bias or motivation, and get me back 10 results from these types of sources, publications, independent bloggers, what have you, and synthesize it. Don't give me just one, actually go through and then synthesize the results. So I'm getting something that isn't that new. And again, there's skill involved here in terms of how to do this,

• 33:34Speaker 01

  but then also there's a desire that you don't want the gaming to occur. And so that's a capability that's now at people's fingertips that didn't necessarily exist in the search paradigm where you're kind of restricted by keywords and having to filter stuff out. And that is just, I think that's a new thing that people will have to contend with. They're really trying to gain results and, you know, if there's a way to do it, it won't be based on the old way.

• 33:55Auren Hoffman?

  Like I have a human assistant and I'll often ask her to get me a hotel or get me a restaurant reservation or buy me a charger for my phone or something like that. And she does the research and I kind of trust her with being a, in some ways, my agent. I assume at some point I'll be, you know, some of those queries, I'll be able to ask a AI to do that for me and I want to trust it with the results. And I imagine there'll be a lot of folks in the world who'll be incredibly incentive to try to influence those results to get me to go to a different hotel or a different restaurant or a different charger. So I imagine these games will go on forever.

• 34:37Speaker 01

  I think the thing that's interesting, and this kind of goes back to the thing that we're talking about with being satisfied and dissatisfied and people continue to adapt. We're kind of answering the question in a different context, but I think people always figure out a new type of game to play. And it might not be in this case, though, that it's a system of gaming that is necessarily misaligned with what the user wants. It could be the situation where the marketers that succeed here are actually, they're doing their best job of actually matching the products that the user are actually intending to look for. And what was gaming at one point is actually closer to now looks more like optimization.

• 35:22Speaker 01

  That's like, that would be like a very interesting way to develop. We'll see.

• 35:25Auren Hoffman?

  One thing that is still seems so hard, for whatever reason, in the non-AI world is good content recommendations. Showing up on Amazon and even though it has a history of like 30 years of what I bought, it still doesn't tell me good things to buy. I show up on Spotify and still doesn't give me good music to listen to. I show up on Netflix and the recommendations are just terrible. Why is that? Like in that new AI world, maybe they haven't taken advantage of it, or maybe they don't want to put in a million tokens to, you know, in memory. But why is this so hard?

• 36:04Speaker 01

  It's hard for me to comment on, you know, people's dissatisfaction with Spotify and Netflix algorithms. But in our own sort of work on personalization, because that happens inside of ChatGPT too. You know, one of the things that we found is just that what people want in terms of signals is not necessarily one to one with actually what they want. It's a little bit of an art to figure out what are the right ways to, if for this is inside a product that let the user give you reliable signals for what they actually did want. And, you know, we have this thumbs up, thumbs down, but that's very lossy, I think. Because why do they thumbs up? Because the model gassed you up? Was it a good answer? Were you just feeling generous that day? Right? That I think is one part of it.

• 36:47Speaker 01

  And then the second part of it is even when the model gives you what you want, what you say you wanted, and this is especially true in consumer tech, right? What you say wanted is not actually necessarily the same thing as what you actually wanted. People preference falsify sometimes. Those things I think are. You know, probably more like permanent features of trying to understand what people want personalized things and how to personalize things to them. And that's probably why it's an enduring challenge.

• 37:12Auren Hoffman?

  As we get into this global competition for AI, it kind of harps back to back in the day in like the 40s of Operation Paperclip, like bringing these thousand German rocket scientists to the US after World War II to keep them out of the Soviet hands. Should we do something similar for these like amazing AI engineers? You're sitting in Hungary or something and you may have a lot of opportunities for, should we like get them to the US with any scenario or?

• 37:43Speaker 01

  In a short answer, I think, yeah, we should be figuring out a way to have as much AI talent come to the United States as possible if we want to be a leader in this technology because talent is one of the key ingredients in terms of continuing innovation.

• 37:59Auren Hoffman?

  Do you agree there's a relatively small number of just like ultra talented people in the AI field?

• 38:06Speaker 01

  I think that's the condition today, but I think as more talent comes in, as the industry matures and more people go into the field, you know, that'll change too. Get market correction and all that. But I do think that we should be trying to find ways to collect more talent here, assuming of course, the objective is for the United States to continue to lead in innovation here. And I think the other part of this though, too, is how do you bring the talent over here so that it stays here, settles here?

• 38:33Auren Hoffman?

  Right, they don't like come for a few years, then go back to where they gave or go to the highest bidder or something, right?

• 38:39Speaker 01

  And becomes part of the fabric of society, right? And I think that's important too, because it wasn't just, you know, Paperclip brought a bunch of scientists over. Those scientists came over and, you know, they became American citizens and.

• 38:50Auren Hoffman?

  Yeah, they raised their family here, their kids are here, they stayed here. This was the best place for them to go.

• 38:56Speaker 01

  The talent is very important and we should definitely work to bring the talent here, but there's a values question too. And how do you make sure that the two are evolved together, complement each other? Because I think that's what also creates durable, you know, sort of talent accumulation.

• 39:11Auren Hoffman?

  Where else should we be thinking about if we're really trying to make sure the US is like by far the dominant country for AI? Like what else should we be thinking about from like a US policy perspective?

• 39:25Speaker 01

  I really love this question and you know, people ask me this sometimes and I always say the same thing, which is we could have a research advantage and we could have a technology innovation advantage. But these advantages are measured in terms of time. Like you might be a year ahead, two years ahead, 18 months ahead, whatever. And that time element only matters so much if you are actually broadly adopting the technology too. You could advance the science, but if people don't use it, but other people are behind, but even if they're behind, their population uses it, embraces it, gets all the leverage out of it. You know, what have you done with your lead really? I think that from a policy perspective, and I think the current administration is really trying hard to do this is how do you

• 40:10Speaker 01

  come up with ways to increase the pace of adoption and increase the efficiency of adoption? And increase the human capital around the adoption so that people understand how to use the technology well. And I think then you get the most out of not only the technology, but actually the the research lead that you have.

• 40:30Auren Hoffman?

  It's not like AI is hard to use. And it's not like there was some government thing to to teach you how to use a smartphone. Like everyone just figured it out pretty fast. And I assume with AI it's similar, like it's just, there's so many benefits so quickly that once you start it and once you see other people using it, like you just jump in. I know a bunch of engineers, even six months ago, they weren't going to use AI to to do their coding. And they've all switched over the last six months. And a lot of marketers are similar and a lot of artists I know are similar. Like so many people will start seeing other people using it and once it's there, they'll start moving. So it's not like we need government being like, hey, use this tool. It's amazing or something, right?

• 41:12Speaker 01

  Outside of government, maybe not so much, right? Because you know, private sectors. The kind of, you know, have the market incentive push to adopt technology as soon as somebody else in their field does, right? At the governmental level, you know, you still have, I think, things like all procurement processes that are a little bit slower, budgets that are written for a particular type of technology or tool that don't necessarily fit the new thing coming.

• 41:33Auren Hoffman?

  That was even true, like when the iPhone came out, it took probably more than five years before anyone in the government could buy it. Like they were still using the older phones and stuff like that. Like that just happened, right?

• 41:45Speaker 01

  I think that figuring out how to just shrink the gap between when the technology hits the market to how does a particular, you know, government agency or constituency get the technology in, diffuse it through their organization and along with the, the ability for people to make the most out of it in relation to their actual day-to-day workflows. Because it's one thing to kind of know how to use ChatGPT as a coach to help you learn things or do some research online and whatnot. It's another thing to say, okay, I'm going to deeply integrate it into my day-to-day work and for example, get rid of some of this paperwork that I do. And it also means that probably there's a turnover of other legacy systems that you

• 42:28Speaker 01

  need to have that make it even more high leverage. And so an example of this is we have connectors inside of ChatGPT, plugs into things like Gmail, Slack, Google Docs, all of that. But if you were an organization, you know, running on Lotus 123, for example, it's going to limit the leverage that you get from ChatGPT. And so there's like a transformational aspect to this.

• 42:49Auren Hoffman?

  A couple of personal questions. What is something you personally wish AI could help you with, but it's just not there yet?

• 42:57Speaker 01

  Ambient learning is probably the thing that comes to mind for me and it's because it comes out of this mild frustration dynamic that we were talking about before. So I ask it questions, I have it do things for me, but I wish there was a way for me to just work and there's some kind of observability of my work by ChatGPT. And then it prompts me every so often, you seem to be doing this type of work and this is how I think you do it. And based on observing you, this is what it seems like a, a good result is and what a bad result is. And then I'm able to give a feedback. And so it's, it's a little bit like if I were to bring a mentee, then they come to me and say, okay, it looks like I'm learning these things from you. Am I learning the right lessons and I can kind of steer them?

• 43:42Speaker 01

  And then it gets to a point where it's able to do certain things. And then it's just pops up one day and says, okay, I think based on watching you work. This is like one thing that you do, and I think I know how to do it. Now I can just do it for you if you want me to from now on. I think once we hit that, that's like another step change in sort of productivity that you get and empowerment that you get from the tool.

• 44:02Auren Hoffman?

  Yeah, that's super cool. All right, two questions we ask all of our guests. What is a conspiracy theory that you believe?

• 44:09Speaker 01

  So cheating a little bit, because I think you asked me this question once a long time ago, but it was in like a group setting. I didn't get a chance to go, so I already have the answer in my head. It's more humorous than serious, but so I'm a basketball fan. There are these playoffs from way back in the early 2000s between the Kings and the Lakers. It was like game six and it was some crazy statistic where in like the third or fourth quarters, the free throw disparity was something like three to one. And the Lakers squeaked out with like one point victory or something like that. And then a few years later, it was revealed that the referee calling the game was gambling on the games or something like that. And, you know, it's just, it makes you scratch your head a little bit. That's one that I kind of think about sometimes.

• 44:46Auren Hoffman?

  For sure, there's somebody like often tipping the scales to a better, like a team like the Lakers. The NBA front office is tipping the scales to the Lakers or those types of teams. They just want the Lakers in the finals and not the Kings. Like this is going to be better for every team because they're going to get better. They're just going to get more people watching, et cetera.

• 45:09Speaker 01

  Although that Kings team, I gotta say, was a lot of fun to watch. And, uh, yeah, it was just, sometimes you wonder what might have happened.

• 45:17Auren Hoffman?

  All right, last question we ask all of our guests. What conventional wisdom or advice do you think is generally bad advice?

• 45:23Speaker 01

  I think it's probably this advice where it's something along the lines of do what you love or follow your passions. And I think it's because it's just that, you know, the assumption somewhere in this statement, at least the way that it lands with me, is that intrinsic motivation comes from some kind of embedded part of your nature. It's just kind of who you are, and so you must love what you love. And therefore you need to go with it, right? I think of it more like. It should be a mental game of creating intrinsic motivation from deciding you just want to do something. And that is, for me, I think, how I think of the essence of agency. And so what I tell people instead is, love agency and figure out how to have more of it in your life. And I also think that that's also part of this other thing that some people

• 46:08Speaker 01

  say, which is just, just love to learn and grow. Because engaging in that way in your life, actually, it's just an extension of, if I want to learn more, that enables me to have more agency over my life. And that's actually something I'm consciously doing. It's not something that I'm pursuing.

• 46:24Auren Hoffman?

  What do you mean by agency?

• 46:26Speaker 01

  A selfish determination. I just, I just get to decide, these are the things that, you know, I love. It's not something that I, it happened to me. And it's a, you know, it can be like, all the way to like reinventing your identity, to taking something that you don't like, and actually figuring out how to like it, so you can get it done, right? And so, I mean, other people might call it discipline. I don't know. That's what I think about.

• 46:48Auren Hoffman?

  It is interesting, because like, sometimes people say, I was born to do this. And I'm sure there are some people who are born to do something like the prince of a country or something, and they were born, they're eventually going to become the king. But most of us, like, I don't think we're born to do anything particularly. Like, we get to choose where we want to go in life.

• 47:08Speaker 01

  I think that's right. And I think that that's, that's the beautiful part of life. You get to choose. And I think that that is, that's really kind of what I'm getting at, is you just get to decide who you are, and what you love, and what you're going to follow. And it doesn't have to be something that you feel like, oh, like, this has been decided for me, and therefore I must do it. I think that that's a narrative sometimes we tell ourselves. And that could be good, because that can be a source of motivation. But it's also good to realize that that is a narrative that you tell yourself. And you can decide to just rewrite it one day if you want.

• 47:37Auren Hoffman?

  All right. That was actually very beautiful. It's a great way to end. Thank you, Jason Kwon, for joining us at World of DaaS. I follow you at Jason Kwon on X. I definitely encourage our listeners to engage with you there. This has been super interesting and a ton of fun. So I really appreciate it.

• 47:53Speaker 01

  Thanks, Oren. Great to be here.

• 47:57Auren Hoffman?

  Thanks for listening. If you haven't already done so, please subscribe to World of DAS wherever you consume your podcasts, YouTube, Apple, Spotify, and more. And please help us get discovered by leaving a review. And check out worldofdaas.com. That's worldofdas, D-A-A-S, dot com. And of course, connect with me on Twitter, at Auren. That's A-U-R-E-N. Would absolutely love to hear from you.