How AI Changes the Economics of Innovation

1. 0:00 – 0:56

   Intro

   1. MCMartin Casado0:00

      Right now, if I give 20 people a billion dollars, they can actually use it usefully. We've kind of moved the industry from like this engineering-bound problem to a capital problem that's fundamentally very different.

   2. SSSteven Sinofsky0:11

      Math is very much a leading edge indicator of what the market might be interested in and why.

   3. MCMartin Casado0:17

      Some people will walk in and say, "The foundations to AGI and to reasoning is gonna be math," but like that doesn't tell you anything about reality. For me, it's still in the domain of like it's really good at playing a game.

   4. SSSteven Sinofsky0:26

      The startups don't aim straight at the incumbents-

   5. MCMartin Casado0:29

      Yeah

   6. SSSteven Sinofsky0:29

      ... and the incumbents just don't pay attention. Microsoft is worried way more about what Amazon and Google are doing than anyone in a startup space. Everybody who's from a big company in Silicon Valley, you always think, "Oh my God, we're just gonna crush all of these little companies," and then you realize they never get crushed.

   7. MCMartin Casado0:44

      And I think this is why we're seeing such meteoric growth of the Cursors, the Anthropics, and the OpenAIs.

   8. SSSteven Sinofsky0:50

      Although capital is scarce and it's hard to get and all of these other things, once you get it-

2. 0:56 – 12:05

   Making Sense of AI & Math: The Riemann Hypothesis Moment

   1. ETErik Torenberg0:56

      First off, thanks for both of you making, making time to come on the podcast.

   2. MCMartin Casado0:58

      It's fun to be here. That's great.

   3. ETErik Torenberg1:00

      Um, Jared Sumner tweeted a few days ago s- something along the lines of, uh, how he told Claude to try to solve the Riemann Hypothesis and to try harder.

   4. MCMartin Casado1:09

      Yeah.

   5. ETErik Torenberg1:09

      And, uh, I, I don't know if there was actually any, any progress made, but it's part of the larger conversation around, hey, it seems like there's some, uh, accomplishments that are, that are, that are being made. H- How do we make sense of this in terms of w- what is actually happening and what does it mean for, for math?

   6. MCMartin Casado1:24

      I'm gonna let, uh, Steve go first.

   7. SSSteven Sinofsky1:26

      Oh, well, I'm no mathematician at all, but I am ... I mean, I think it's, it's just, it's an important moment because it, it sort of divides the world into, into two groups. Like the groups that are just very, very excited that like, "Oh my God, these, these things are being solved." It doesn't matter if you understand them. Actually, nobody under- The number, the universe of people who understand what these things are is very small. And then there are the people who are just like, "Oh, it's fake. It's gonna put people out of jobs, that no one's gonna know the future of where these fields go." And the most interesting thing about it is the group that's most excited are mostly the mathematicians.

   8. MCMartin Casado2:03

      [laughs]

   9. SSSteven Sinofsky2:04

      A- And, and they're the ones ... And so that actually confuses everybody because if you're of the school, the people who are like, "It's gonna put people out of work, and it's gonna, we're gonna all get dumber, and it's, you know, the dawn of idiocracy because computers are doing all of our work," you're confused that the people who are impacted most by, by what this level of AI did are the most excited.

   10. MCMartin Casado2:25

       Yeah, yeah, yeah.

   11. SSSteven Sinofsky2:26

       And, and I think that's just ... I, I, I think that that is itself shining a light on this moment that we're in right now.

   12. MCMartin Casado2:33

       You know, you're, you're talking to two systems guys.

   13. SSSteven Sinofsky2:35

       [laughs]

   14. MCMartin Casado2:35

       Two product guys. And so you're gonna get ... Like we're, we're at like, uh, to have the same caveat. I feel that there's some things like we're actually both very expert on. This is not one of them. So I'm gonna kind of, from the peanut gallery, I, I've got two comments. So one of them is like, okay, so I, I view like economic utility to be a very important, um, uh, measure when you're talking about AI, right? So I was trying to think like there's a lot of hours been trying to solve some math thing, right? But like if you sum up the entire postdoc salaries [laughs] of all the people that have been working over the years on these problems, it's probably not very much. And so part of me is saying like it's great that there's these capabilities. I'm not sure that the fact they've been longstanding is that much of an indication because there hasn't been a huge economic incentive, um, in order to solve. Now, that doesn't mean that it's not hard or whatever. It's just like I just don't think we have like that, like that validation of this unlocks some deluge of like economic value. And the second one is it's kind of not surprising to me that AI is very good at solving a almost purely axiomatic domain that, you know, requires knowing a whole bunch of different things and, and, and putting, you know, putting the solutions together from very disparate spaces 'cause, uh, often really when I read ... So I've, I've been reading all these like everybody else has been obsessively and they're like, "Oh, like it came up with a solution." It's like, yeah, the solution was pretty straightforward. It's just like bar from a bit of math that I didn't know. And so I've ... I think if there's like a meta learning here, the meta learning is, is there is a set of problems that probably, you know, require you to be too broad for most humans or most education, and it's gonna solve those. It's clearly very good at solving axiomatic systems, but it, it, I don't think it provides a strong indication of is this solving things that the market hasn't been able to solve? 'Cause there really hasn't been a market around these. And so I think that's the next question for us to answer. So very exciting. Seems kind of reasonable and understandable. Not sure what the longer term implications are.

   15. SSSteven Sinofsky4:40

       I d- I do think that there's something interesting that, that math is very much a, a leading edge indicator of what the market might be interested in and why. I mean, if you ... Like I remember when I was in school, like there was some big thing that someone at AT&T invented a new, um, algorithmic, a new program for doing, um, l- linear algebra, like a new way to solve linear, which is super important right now in the AI world. But his big thing was, well, now we can just calculate like the United Airlines flight map in like three hours less time than we could the, the last week.

   16. MCMartin Casado5:15

       Right. Yeah, but, but let's dig into it. So it's just not clear to me that the problems being solved are those that are roadblocks to like existingly economically useful tasks.

   17. SSSteven Sinofsky5:24

       Right.

   18. MCMartin Casado5:25

       And if they were, it's not clear to me that they wouldn't have been solved. Like postdoc that's been ruminating on a problem and getting paid 30K a year for [laughs] five years-

   19. SSSteven Sinofsky5:34

       Yeah, yeah

   20. MCMartin Casado5:34

       ... like it's very different than like the market has decided that this is like the one thing to unlock. And, and maybe they're there. Maybe these problems that are being solved are like the key problems to unlocking some big economically productive use case. I just haven't seen that yet.

   21. SSSteven Sinofsky5:48

       Oh, I-

   22. MCMartin Casado5:49

       So that for me is like the next thing I'm kind of looking for.

   23. SSSteven Sinofsky5:51

       I, I don't even know what 12-dimensional space is or what that means, and so like I, I'm comp-

   24. MCMartin Casado5:56

       [laughs]

   25. SSSteven Sinofsky5:56

       I, I'm completely with you on it. Like I don't even know what problems are in 12-dimensional space. Like are you very skinny? Are you very tiny? I'm really confused by that.

   26. MCMartin Casado6:04

       And, and, and may- and maybe I'm wrong here, but for me it's still in the domain of like it's really good at playing a game.

   27. SSSteven Sinofsky6:08

       Yeah, yeah.

   28. MCMartin Casado6:09

       Like this is the best StarCraft player ever, which is cool and it's very powerful, but like I have a hard time connecting that With A, like maybe the reason we didn't have them before is because there just was an economic need, and B, like how does that actually map? And so listen, I, there's a huge range of these things. We get pitches all the time. Some people will walk in and say, you know, the, the, the foundations to AGI and to reasoning is gonna be math, and once you do that, you'll be able to answer every question because the universe is based on some, you know, fundamental mathematical principles, and once you understand that, you understand everything. And then, you know, there's other people candidly that walk in the door and they're just like, "Listen, that's great, um, but like that doesn't tell you anything about reality." And so, you know, I think that there's more work to do, and this isn't just about like getting better at math.

   29. SSSteven Sinofsky6:54

       Yeah. I do think what's interesting is that the, part of the reason that the mathematicians are very excited about it though is because they, they work a certain way.

   30. MCMartin Casado7:02

       Yeah.
3. 12:05 – 19:43

   Will AI Math Ever Map Onto Physical Reality?

   1. MCMartin Casado12:05

      doing." So listen, I, listen, it's hard not to get philosophical when you're talking about AI, so I'm gonna get philosophical, and you can tell me to shut up- Please ... but I, I ca- I just can't. Like you, you kind of do. So, so the, the math, this math thing seems to me a, a, a little different because like it, it kind of begs the following question which is will math ... ever be representative physical phenomenon, right? Like, has anybody ever, like, taken a bunch of equations and actually predicted something, like, physical? And I don't know the answer to that. Like, so I worked in these large simulation codes, and these large simulation codes are actually, um, trying to compute physical phenomenon like the explosion of a star or, like, you know, what would happen to, like, whatever, an airplane in, like, a, an, uh, an air simulator or a wind, uh, wind simulator. Um, but all of tho- And even though they're just calc- there's calculating these, like, large, you know, differential equations, they were all based on empirical results.

   2. SSSteven Sinofsky12:59

      Yeah.

   3. MCMartin Casado12:59

      Like, literally the equations of state for the-

   4. SSSteven Sinofsky13:01

      Well, they were a model. They were-

   5. MCMartin Casado13:02

      Just-

   6. SSSteven Sinofsky13:03

      They were like, "We can measure temperature in these places."

   7. MCMartin Casado13:05

      That's exactly right.

   8. SSSteven Sinofsky13:05

      Like, like, so-

   9. MCMartin Casado13:06

      So it was all, it was all based on empirical equations of state.

   10. SSSteven Sinofsky13:08

       Yeah, yeah.

   11. MCMartin Casado13:08

       And so I've always wondered, like, like, is simulation computationally irreducible, and so you actually have to actually run the simulation? In that case, it's not clear to me to what extent AI helps. Like, I know people are trying to solve this problem with AI, but, like, I don't know if these math an- these math answers have any impact on that type of stuff, right? And so maybe there's some separate algorithmics domain to, to your point where they do or, you know, maybe, like, like modeling or logistics. But when it comes to, like, you know, will, you know, will this star explode? Will this building stand up? Like the actual simulation, I think these things, things are pretty disjoint. And, and then I read a lot of these discourses on the, the math solutions, and there's kind of these claims where if it can solve all math, you can predict anything, and I just think that that's-

   12. SSSteven Sinofsky13:58

       Yeah

   13. MCMartin Casado13:58

       ... a huge, huge logical leap which is not clear to me that is, is, is, is, is, is, is obviously true.

   14. SSSteven Sinofsky14:04

       Yeah, I think-

   15. MCMartin Casado14:05

       Or, or there's any indication it's true at all. Like ...

   16. SSSteven Sinofsky14:07

       So the way, one way to, that I think I might, um, talk about that, like, you know, again, like, this is so out of my league on the actual math. But-

   17. MCMartin Casado14:15

       [laughs] This is what happens when two systems people-

   18. SSSteven Sinofsky14:17

       But, but I'm gonna-

   19. MCMartin Casado14:18

       ... talk about, talk about math. [laughs]

   20. SSSteven Sinofsky14:18

       But, but I'm, I'm, I'm in- I'm inherently a tool, I'm inherently a tools person.

   21. MCMartin Casado14:22

       Yeah, yeah, yeah.

   22. SSSteven Sinofsky14:22

       And so I kinda get this part of it, which is that what's happened is, is that, that AI might not be the next tool to solve math problems-

   23. MCMartin Casado14:33

       Yeah

   24. SSSteven Sinofsky14:33

       ... at, at some scale that matters, but it might lead to the development of a new kind, a new level of model. And so I brought, like, props to sh- show this off. So of course, this is the original-

   25. MCMartin Casado14:42

       Yeah

   26. SSSteven Sinofsky14:43

       ... math tool.

   27. MCMartin Casado14:43

       Yeah, yeah, yeah.

   28. SSSteven Sinofsky14:44

       And so before something like this, this is a re- uh, you know, one of these real ones from, like, Beijing market.

   29. MCMartin Casado14:49

       Oh, it's the actual-

   30. SSSteven Sinofsky14:50

       Well, well, I, I, you know, it's the ones they tell tourists it's from. But, but I'm very proud of that 'cause I negotiated it down to, like, seven cents. But, um-
4. 19:43 – 38:16

   The Cold War, IBM 1953 & the Cultural Roots of Computing

   1. MCMartin Casado19:43

      and-

   2. SSSteven Sinofsky19:43

      Is that because of the Cold War? Was it because...

   3. MCMartin Casado19:45

      Well, obviously, the Cold War was a big cultural part of it-

   4. SSSteven Sinofsky19:48

      Yeah

   5. MCMartin Casado19:48

      ... uh, uh, for sure, but it was just a general, the, the future. I, like-

   6. SSSteven Sinofsky19:52

      Yeah

   7. MCMartin Casado19:52

      ... I found this, uh, incredibly cool brochure from-

   8. SSSteven Sinofsky19:56

      Oh, yeah

   9. MCMartin Casado19:56

      ... uh, IBM from, it's from 1953. So 1953, the-

   10. SSSteven Sinofsky19:59

       Oh, wait, like, do you just have this stuff in your house? I just stumbled across it. Like, this-

   11. MCMartin Casado20:03

       It's amazing

   12. SSSteven Sinofsky20:03

       ... this one I just got. I can't even believe this exists. But this is, like, this is a brochure about the future of computing.

   13. MCMartin Casado20:09

       Wait, I wanna see it. You haven't shown it to me.

   14. SSSteven Sinofsky20:10

       But, but, but, like, first you gotta look, it's, it's got, like, nuclear, uh-

   15. MCMartin Casado20:14

       Oh, that's fantastic

   16. SSSteven Sinofsky20:15

       ... like, the whole thing, the future of computing is, like, a guy with, like, atoms racing around his head. Oh, wait, we're zooming in and doing the Carol Merrill thing. So but the, the fascinating thing is it's from 1953, so your ENIAC and, and that point, like, that's it. That's the computer at the time. This is pre-'704-

   17. MCMartin Casado20:32

       Yeah

   18. SSSteven Sinofsky20:32

       ... pre-

   19. MCMartin Casado20:32

       Phenomenal

   20. SSSteven Sinofsky20:33

       ... 370. And so it's a brochure from IBM explaining what a computer might be, not even is. And it's like, "It took millions of years to invent and recognize the usefulness of the wheel." That's the opening sentence of, of, of the...

   21. MCMartin Casado20:48

       [laughs]

   22. SSSteven Sinofsky20:48

       And, and but, like, you, people were eating this stuff up.

   23. MCMartin Casado20:50

       Yeah.

   24. SSSteven Sinofsky20:50

       But here's the part that I wanna get to. It talks about computers, and it's the two families of computers.

   25. MCMartin Casado20:55

       Yeah.

   26. SSSteven Sinofsky20:55

       And so, of course, you get the slide rule-

   27. MCMartin Casado20:57

       Yeah

   28. SSSteven Sinofsky20:57

       ... and that's explaining the history. And what this is really leading up to is we could do this for text, too.

   29. MCMartin Casado21:04

       Yeah.

   30. SSSteven Sinofsky21:04

       And so the idea w- and I mean, like, imagine who is reading this in 1953 that it has to explain hex and decimal and binary-
5. 38:16 – 46:28

   Rethinking Fundamental Assumptions About Software

   1. ETErik Torenberg38:16

      And to, if we do need to rethink some fundamental assumptions, the, the, what may that look like?

   2. MCMartin Casado38:23

      Well, I just think that like, um, people like, you know, Steve and myself have built these deep intuitions on how systems function and how they hit the industry based on 40, 50 years of like watching this stuff, and I just don't know, like things like will value go to the model or to the app, how much capital can you apply to this stuff, what classes of problems can you solve versus not solve, um, what guarantees that can you provide, uh, how does this impact productivity? There's a lot of things that we've got intuitions on, and for me the big question is do we have to like reshape those assumptions or not, and to what extent do we have to? 'Cause the laws of physics feel a little bit different. I'll just give you one example. I mean, I've said this many times, just think it's so important. 20 years ago, and if you're a startup of 10 people and I gave you a billion dollars, what would you do with it?

   3. SSSteven Sinofsky39:15

      Y- you would end up spending a, a ton of money on building out, buying your own computers and things, if that's where you're going, if that's what the point-

   4. MCMartin Casado39:21

      Well, you hire people, you buy computers-

   5. SSSteven Sinofsky39:23

      Yeah, yeah

   6. MCMartin Casado39:23

      ... you blow up. Like you wouldn't know what to do with a billion dollars 20 years ago.

   7. SSSteven Sinofsky39:25

      Oh, oh, I see what you're saying. Yeah, yeah, yeah.

   8. MCMartin Casado39:26

      Yeah, 10 years ago, I give you a billion-

   9. SSSteven Sinofsky39:27

      Yeah

   10. MCMartin Casado39:28

       ... like you hire-

   11. SSSteven Sinofsky39:28

       I, I just-

   12. MCMartin Casado39:28

       You hi- you hire engineers and you'd be fucked.

   13. SSSteven Sinofsky39:30

       Right, right, right.

   14. MCMartin Casado39:31

       Like [laughs] what do you do? Like-

   15. SSSteven Sinofsky39:32

       Right, right

   16. MCMartin Casado39:32

       ... write code, you've got to, you know, you've got, you know, that's-

   17. SSSteven Sinofsky39:36

       The billion, the, the important part of that is it's a billion.

   18. MCMartin Casado39:39

       It's a-

   19. SSSteven Sinofsky39:39

       It's not that you got money, it's that it's a, it's a huge-

   20. MCMartin Casado39:41

       It's a billion.

   21. SSSteven Sinofsky39:41

       It's a, it's a ton of money.

   22. MCMartin Casado39:42

       If I, if I give you a billion dollars-

   23. SSSteven Sinofsky39:43

       Right

   24. MCMartin Casado39:44

       ... two years ago-

   25. SSSteven Sinofsky39:45

       'Cause 10 million you'd buy a bunch of stuff from Hewlett-Packard and then the money would be gone and out.

   26. MCMartin Casado39:48

       Yeah, yeah. For sure, for sure.

   27. SSSteven Sinofsky39:48

       Right, right.

   28. MCMartin Casado39:49

       This one is a billion dollar-

   29. SSSteven Sinofsky39:50

       Yeah.

   30. MCMartin Casado39:50

       I mean, like in software you hire people and then it's-
6. 46:28 – 55:05

   Incumbents vs Startups: Why the Innovator's Dilemma Still Wins

   1. ETErik Torenberg46:28

      I wanna talk about any other fundamental assumptions that might be interesting to revisit. I- is, is it sort of how about incumbents versus startups? You know, the, we've talked a lot about innovator's dilemma. Does, does that, uh, you know, uh, the, now that these startups are, or these incumbents are, you know, have the capital advantage, are they able to, to do more? But at the same time, we're seeing startups that you would think incumbents would just destroy the-

   2. MCMartin Casado46:46

      This is the crazy thing. If you would've told me six months ago, you'd have asked this question, say like what, like what advantages do incumbents have? They have the same advantage all incumbents always have. They have the capital, and they have the cash flow, and they have like whatever.

   3. SSSteven Sinofsky46:59

      Distribution and, you know.

   4. MCMartin Casado47:00

      They have distribution. And like what's crazy is AI, A, solves the distribution problem. It just solves the demand problem. And B, these companies are able to raise so much money that they're actually on competitive footing-

   5. SSSteven Sinofsky47:11

      Mm

   6. MCMartin Casado47:11

      ... with like the Microsofts and the Metas and, and the Microsofts. And, uh, so I think we're in a very new territory when it comes to these new challenges versus the incumbents, specifically for these two reasons. You know, I think that the, the, the, the, um, the distribution point is, is, is often misunderstood how impactful it is. In the past, if you had a company and you wanted to get people to use your stuff, it was hard. You'd hire marketing. You have no idea how much like to invest and where, and like you didn't know what you were getting on return on investment. But the demand is so unlimited for tokens and for GPUs. Like literally you can just decide how much money you're putting into it in order to drive top of funnel and growth. And so the things have typically been very, very hard for startups, you know, are much easier now. And I think this is why we're seeing, um, such meteoric growth of the Cursors, the Anthropics, and the OpenAIs. That results in capital access, and that has put them on, on, on even footing, so very interesting space.

   7. SSSteven Sinofsky48:09

      Yeah, I, I, and I think it's to your point about how hard it... Look, m- my whole life was managing thousands of people of, of engineers to build things that couldn't be built anywhere else. Like it was the moat to build an operating system, it was infinite.

   8. MCMartin Casado48:22

      Yeah.

   9. SSSteven Sinofsky48:22

      And-

   10. MCMartin Casado48:24

       Yeah.

   11. SSSteven Sinofsky48:24

       And I think-

   12. MCMartin Casado48:24

       You have to have one Cutler. Is that the-

   13. SSSteven Sinofsky48:26

       Well, it's, it, but it's, it, it really, it read, read-

   14. MCMartin Casado48:30

       That, that was mostly you.

   15. SSSteven Sinofsky48:30

       Yeah, I know. No, I get it. No, but, but-

   16. MCMartin Casado48:32

       He's brilliant

   17. SSSteven Sinofsky48:33

       ... but read the, um, read the exile, Steve Jobs in Exile book because it-

   18. MCMartin Casado48:37

       Can't wait

   19. SSSteven Sinofsky48:38

       ... you, you can, you really get a sense for like building op... In fact, you know, of course, NeXT was famously just, it took the code from Mach at, at Carnegie Mellon and started from there. We couldn't have done it from, from scratch completely. But this, this whole idea Of, of just, um, how, how important it is to, to think through the, the domain specific and the, and how you disrupt people, because y- there, there's an old joke at Harvard Business School when Clay was, was still with us, which was they really-- It's weird that they teach disruption as a theory in the business school, when really it should just be a fact in the physics department.

   20. MCMartin Casado49:12

       Oh, I like that.

   21. SSSteven Sinofsky49:13

       And, and I love that. I was there in '98 when he was writing the book and the paper and everything.

   22. MCMartin Casado49:17

       That's nice.

   23. SSSteven Sinofsky49:17

       That's when I was teaching.

   24. MCMartin Casado49:18

       That's actually great. Yeah.

   25. SSSteven Sinofsky49:19

       And, and, and I really, I used to be a... Of course, there's a, a lore with everybody who's from a big company in Silicon Valley or when you arrive, like I did. And the theory is always like, you always think, "Oh my God, we're just gonna crush all of these little companies." You always think that when you're at the big company, and then you realize they, they never get crushed.

   26. MCMartin Casado49:37

       Yeah.

   27. SSSteven Sinofsky49:37

       Like, and that, Ben always makes this point. Like, they just, and Mark does in his, um-

   28. MCMartin Casado49:41

       Yeah

   29. SSSteven Sinofsky49:41

       ... hi- his movie did-

   30. MCMartin Casado49:42

       It was, it was AWS actually put out of business.
7. 55:05 – 1:02:15

   The Limits of Current AI Architecture & What Comes Next

   1. ETErik Torenberg55:05

      The, um, last thing, when Vishal came in, uh, and we had him on the podcast, he was sort of, um, he thought LLMs were a great achievement, but he was bearish on their ability to invent new discoveries, particularly like scientific breakthroughs or, or things like that. And I'm curious if you think the, the sort of math progress, um, is consistent with that or, or, or what, what is your latest thinking on sort of the limitations of the, of the current sort of, uh, you know, model architecture versus like will we need more, um-

   2. MCMartin Casado55:37

      So here's my, here's kind of my new view, which is I think we know exactly how these things work. You put a bunch of data in them, they're stuck to that data. They o- can only do in-distribution stuff, and they can move along that manifold, um, in a perfectly Bayesian way. So we, we, okay, so we can say these words, and then, then the question is, is, okay, but what are the implications of that? Like, what problems can it solve, right? I think it's just so hard for a human being to reason, to reason about a digital artifact, in this case the model, that was built with five billion dollars. So like in the history of humanity, we've never created a single digital artifact that had that many flops and that much data in it. So on one hand, we know exactly how it works from a mechanic standpoint. On the other hand, that is so much data and that is so much compute, maybe all of that stuff's already in there, and it can solve anything that you want. And so, you know, the conversation has moved from the how do these things work, we know. Can it do out of distribution stuff? No. Um, uh, does, you know, is there transfer learning? Probably not. Like if I RL one thing, it doesn't teach something else. Like are these, is the singularity here? Probably not. I think everybody kind of, most, many people kind of agree on like we're not in fast takeoff. You know, we're stuck to it being in distribution. We haven't closed it. We all agree about that. But what I don't think anybody knows is, okay, but you're still putting ten billions of dollars in that thing. What's it capable of now? And if you, if you consider this meta-economic machinery, which means the ability from Anthropic to raise lots of money and then pour all of that money into this thing to create this super powerful thing, I don't think any of us can predict what that means and where that goes. And so it's a different conversation, but the question is the same. It's like, will that be able to cure cancer? Maybe. But if you put twenty billion dollars into something, maybe it can cure cancer ef- effectively. And that's where I think the discourse has evolved and where it is now. And I, I honestly have decided that I cannot predict what an artifact worth that was, you know, that like you use twenty billion dollars to create is capable of.

   3. SSSteven Sinofsky57:42

      Uh, I, look, I, I think it's just so important. It's important for people who are deep in watching everything that's new to admit that they can't predict, and I think that that's great because it turns out, like I wrote fifty-eight memos on what the internet was gonna be, and I was wrong w- a lot of them by far. But I, I do think on, on the, and, and I, I think-

   4. MCMartin Casado58:03

      But, but, but even this one is a little different. This is like, I, I, I take twenty billion dollars and I put it into a model-

   5. SSSteven Sinofsky58:08

      Right

   6. MCMartin Casado58:08

      ... and, and then you and I look at that model, and we can do whatever we want. I don't think we can comprehend the, like what that even means.

   7. SSSteven Sinofsky58:15

      Yeah.

   8. MCMartin Casado58:15

      It's so many flops and so much data. Like I don't know what that's capable of.

   9. SSSteven Sinofsky58:19

      I, I, and I think we are, I think that that's really true, and I, and I think... But I will say on, on biomedicine in particular, look, the other half of my household is a research doctor who uses AI. We have a Spark at home, and she's loaded-

   10. MCMartin Casado58:33

       A Spark? Like a-

   11. SSSteven Sinofsky58:34

       ... like a, a ton-

   12. MCMartin Casado58:35

       Like a Sun Spark?

   13. SSSteven Sinofsky58:36

       No, no, a, a, a, a NVIDIA Spark.

   14. MCMartin Casado58:39

       Oh. [laughs]

   15. SSSteven Sinofsky58:39

       Yeah. No, no. No, not, not with a C or the K. Oh, yeah. Wow. We were in old times there for, I was like, "Not a Scott McNealy Spark." No. No. Um, and-

   16. MCMartin Casado58:48

       Not McNealy, like a relic

   17. SSSteven Sinofsky58:49

       ... and no, and, and, um, and like it's all AI. Like she does brain stuff and, and surgical brain stuff, all AI, and it's so interesting to see. 'Cause what it, what it really can do is it, it just, it, it sees the patterns that you can't, that only experience could tell somebody. But if there's ten thousand papers on a topic that's part of her model, then like it's just finding the patterns-

   18. MCMartin Casado59:17

       Yeah

   19. SSSteven Sinofsky59:17

       ... that you just, that no one has.

   20. MCMartin Casado59:19

       Yeah.

   21. SSSteven Sinofsky59:19

       And that's a pretty basic AI capability at this point.

   22. MCMartin Casado59:23

       Yeah.

   23. SSSteven Sinofsky59:23

       But it's actually opening up solutions or problems or research directions and things like that. I will say just for the, like this is not a magic to discover drugs because the hard part of drugs has always been candidates.

   24. MCMartin Casado59:37

       Yeah, sure.

   25. SSSteven Sinofsky59:37

       Not candidates. It's always been efficacy and safety. The candidates have, since the '80s, have been able to develop more than we could test. It's human patients, and it's very, very, very hard.

   26. MCMartin Casado59:48

       Can I, can I-

   27. SSSteven Sinofsky59:48

       Please

   28. MCMartin Casado59:49

       ... just tell you something that I got wrong on this?

   29. SSSteven Sinofsky59:50

       Yeah.

   30. MCMartin Casado59:50

       So, um, I, I love the question that you asked, which is how has our thinking evolved on like, you know, whether these things, you know, like their capabilities and generality, which is, um, I was responding to this Bost- Bostrom notion of recursive self-improvement, fast takeoff.

Episode duration: 1:02:29