Why AI’s Next Breakthroughs Could Come from Outside the Big Labs

1. 0:00 – 0:50

   Intro

   1. ALAaron Levie0:00

      If you regulate AI too early, you actually don't solve anything. You still just kinda have the same risk ultimately.

   2. MCMartin Casado0:06

      You willed the thing into being-

   3. ALAaron Levie0:07

      Yeah

   4. MCMartin Casado0:07

      ... but you haven't figured out how to control it.

   5. SSSteven Sinofsky0:09

      The problem we have now is this rift between the labs and the security community that keeps coming to two conclusions: sloppy, and you're not complete in what you're telling us happened.

   6. MCMartin Casado0:18

      An employee is like 10% chance of species extinction.

   7. ALAaron Levie0:22

      Agent swarms completely flip that. These are just roaming drones. But like times 10,000-

   8. MCMartin Casado0:27

      Right? It's just more

   9. ALAaron Levie0:27

      ... and they will easily mistake a good task for a bad one.

   10. SSSteven Sinofsky0:30

       So now we need a whole layer internally that just is tracking way more about what authentications are being done, what APIs are being done.

   11. ALAaron Levie0:37

       Yeah, yeah.

   12. SSSteven Sinofsky0:37

       The US, about 15 years ago, stopped leading in tech antitrust. The problem is that Europe is gonna lead with that because they have nothing to lose. This could change the nature of software fundamentally. The center of innovation has just moved. This is the signal

2. 0:50 – 3:50

   Pacing the Frontier: Reacting to the AI Safety Discourse

   1. SSSteven Sinofsky0:50

      that-

   2. ETErik Torenberg0:52

      Guys, welcome back to the podcast.

   3. ALAaron Levie0:54

      Happy to be here.

   4. MCMartin Casado0:55

      Thank you.

   5. SSSteven Sinofsky0:55

      Great to be here.

   6. ETErik Torenberg0:56

      We've lo-

   7. ALAaron Levie0:56

      Didn't think we'd ever do this again. This is ama- I can't believe it. This is great.

   8. MCMartin Casado0:59

      [laughing]

   9. SSSteven Sinofsky1:00

      I mean, Martin's just building these-

   10. ETErik Torenberg1:02

       I know [laughing]

   11. ALAaron Levie1:02

       ... like $100 billion companies, too busy for this podcast, so.

   12. MCMartin Casado1:05

       Or at least taking credit for them as V- as VCs do.

   13. ETErik Torenberg1:07

       Exactly. Um, we have a lot to discuss today, but fir- first, look, A- Aaron, why don't we start with you? Pacing the frontier.

   14. ALAaron Levie1:13

       Yeah.

   15. ETErik Torenberg1:13

       How have you, uh, reacted and reflected on, on what's happened there and just the discourse that, that's followed? How do you make sense of it?

   16. ALAaron Levie1:18

       Oh, boy. I think we should start with Martin on this one. The, uh, he, he-- you were fighting lots of good ground wars. Um, I, I... Maybe I'll say one thing that we probably all agree with, and then we can figure out where we maybe, maybe, uh, fr- kind of fracture off. Um, I think we would agree that, uh, any AI lab right now at the frontier should be building, you know, in the safest way possible with, with, you know, the, the highest degree of governance and, and security and, you know, kind of whatever your definition of alignment is. Like, this is an incredibly important area of research. It's an incredibly important area for the diffusion of AI. Like, you're not gonna have AI diffusion without extremely high-quality products that can be, you know, trusted by enterprises and that aren't kind of constantly hacking systems. So when, when at least I read the Dario post, um, I actually didn't agree with m- didn't disagree with almost anything, um, because it was all about how do you have better, you know, uh, security of these systems, sandboxing, better testing. There's gonna be some debates around the embedded nature of e- of, of the testers and, like, do you agree with who those are, and does the industry all align on that? But I think actually all of the major points were, were probably salient and ap- and appropriate. Then the only question is, does this get sort of used or leveraged to do things that maybe we don't agree with? Which would be, like, a slowdown of, of AI dramatically because of regulatory controls that would, would sort of, you know, not make it eas- easy to compete, you know, with, with the frontier labs. Um, or do politicians kind of end up sort of taking the, the message and run with it and maybe even worse outcomes happen? Like, you, you... You know, it's used to ban data centers, you know, far faster and, and whatnot. And so, so I think the, the actual, like, substance of the topic is actually incredibly important and, and I, I think very important for, for AI advancement in general. And then the question is what do you do about it? Especially what do you do about it from a regulatory standpoint? And that's probably where the industry's gonna land on very different points in the continuum. But Martin was putting up a good fight on, on, like, let's make sure that we don't use this for regulatory kinda capture, um, which I th- I, I also agree with. But, but I think the, the ideas in the pacing conversation are important. Um, you know, again, like, it's a little bit of a funny concept 'cause maybe it's not even pacing as much as just, like, good hygiene and, and then good w-

   17. MCMartin Casado3:30

       Engineering.

   18. ALAaron Levie3:31

       And, and engineering. And, and so with, with good engineering, obviously there is a slight slowdown, but it's a slowdown that obviously allows acceleration of your diffusion because you, you wouldn't be able to have any of the AI be diffused if, if nobody would, you know, trust using it. So-

   19. MCMartin Casado3:46

       The post is, is very reasonable, but the atmospherics

3. 3:50 – 6:49

   Species Extinction Talk: Do Labs Actually Believe Their Own Rhetoric?

   1. MCMartin Casado3:50

      are not, right?

   2. ALAaron Levie3:50

      Right.

   3. MCMartin Casado3:51

      Like, so an employee is like, "This is gonna kill, whatever, 10% chance of species extinction." And you know what Dario says? "I agree with him more than I disagree with him," right?

   4. ALAaron Levie4:03

      Disagree with him.

   5. MCMartin Casado4:03

      Like-

   6. ALAaron Levie4:03

      On, on TV in an interview

   7. MCMartin Casado4:04

      ... on, on, on TV the same day that he landed these things.

   8. ALAaron Levie4:08

      Yeah.

   9. MCMartin Casado4:08

      And so in some way you can't have these conversations in isolation.

   10. ALAaron Levie4:11

       Mm-hmm.

   11. MCMartin Casado4:11

       Which is, of course, if he's gonna agree in species extinction-

   12. ALAaron Levie4:15

       Mm-hmm

   13. MCMartin Casado4:16

       ... this post that he has looks like this milk toast capitulation that's totally not adequate for the task at hand. And so I think the atmospherics are totally broken. And a lot of my comments were on the atmospherics.

   14. ALAaron Levie4:28

       Mm-hmm.

   15. MCMartin Casado4:28

       And then I have this quibble, but it really bothers me because I'm a pedant, which is I think pacing is the wrong way to describe this.

   16. ALAaron Levie4:36

       Right.

   17. MCMartin Casado4:36

       For, for one-

   18. ALAaron Levie4:37

       Yes

   19. MCMartin Casado4:37

       ... it is orthogonal to security.

   20. ALAaron Levie4:39

       Mm-hmm.

   21. MCMartin Casado4:40

       Right? So, like, you can very slowly build a nuclear weapon-

   22. ALAaron Levie4:44

       Right

   23. MCMartin Casado4:44

       ... and that doesn't make anybody feel better that it's slow-

   24. ALAaron Levie4:46

       Right

   25. MCMartin Casado4:46

       ... versus fast. So that's one. The second one, it just feels like a capitulation to the pause folks without actually, you know, addressing it. So you're saying, "Well, we're not gonna pause, we're gonna pace to make them happy, but we'll also somehow make the regulators happy." And I think it makes them both unhappy.

   26. ALAaron Levie5:02

       Mm-hmm.

   27. MCMartin Casado5:03

       'Cause the pause people are like, "Well, that's not a pause, that's just pacing." And then the regulators are, "You're still doing the thing."

   28. ALAaron Levie5:09

       Right.

   29. MCMartin Casado5:09

       And so I just feel like they're trying to... My, my sense of what's happening is the labs are actually trying to do the right thing.

   30. ALAaron Levie5:15

       Yes.
4. 6:49 – 11:32

   The Nationalization Question & Whose Job Safety Really Is

   1. MCMartin Casado6:49

      ... the answer is to nationalize it and actually put controls that we know that work.

   2. ALAaron Levie6:54

      Mm.

   3. MCMartin Casado6:54

      Right? Now, if they don't actually believe that, and in the private conversations I have, the most sensible people don't, it's a small fraction that do, this is an HR problem.

   4. ALAaron Levie7:03

      Mm.

   5. MCMartin Casado7:04

      Right? So to me, an HR problem is a company problem. Like, if they are worried that they can't recruit people-

   6. ALAaron Levie7:10

      Right

   7. MCMartin Casado7:10

      ... you know, or they can't retain people, which it seems to me a lot of this is just that concern. It's almost more of this kind of researcher currency. If that's the case-

   8. ALAaron Levie7:18

      Mm-hmm

   9. MCMartin Casado7:18

      ... I think that is the wrong reason to cause a national level lockdown on a very promising technology.

   10. ALAaron Levie7:25

       Yeah.

   11. MCMartin Casado7:25

       And so listen, I think the post, again, I think the post is actually very sensible. I think there are real concerns around security. We've had many compute epochs that have real security concerns. I don't think you can ex- you can reconcile discussions on X risk with the proposal that was put out.

   12. ALAaron Levie7:41

       Right.

   13. MCMartin Casado7:41

       You just can't reconcile those two things, and that has always been my primary point.

   14. ALAaron Levie7:44

       Right. Right. Okay, unleash. [laughs]

   15. MCMartin Casado7:47

       [laughs] I've just been holding it back.

   16. ALAaron Levie7:48

       [laughs]

   17. SSSteven Sinofsky7:49

       I'm holding it for you.

   18. ALAaron Levie7:50

       So, okay. The, the first thing is you can't have... You... Is there a schedule that they've published that says when all this stuff, whatever bad stuff is gonna happen, is gonna... They haven't. So you can't pace it because nobody knows when it was supposed to finish in the first place.

   19. MCMartin Casado8:05

       Right. It also, it also just seems disingenuous too.

   20. SSSteven Sinofsky8:06

       It, it's complete nonsense.

   21. ALAaron Levie8:07

       100%, yeah. Totally.

   22. SSSteven Sinofsky8:08

       Like, you can't claim... It, this is like when the press reports on Apple's latest iPhone is late.

   23. ALAaron Levie8:13

       [laughs] From what?

   24. SSSteven Sinofsky8:14

       The phone that nobody knows exists, that they haven't told anybody about.

   25. ALAaron Levie8:17

       Yeah, my Apple Car was very late. [laughs]

   26. SSSteven Sinofsky8:19

       Yeah. Like, I, I don't understand. Like, in order for something to be slower, you need to know the rate at which it was moving in the first place. So it's all just utter nonsense.

   27. ALAaron Levie8:28

       [laughs]

   28. SSSteven Sinofsky8:28

       And, and you can't escape that.

   29. ALAaron Levie8:30

       Yeah.

   30. SSSteven Sinofsky8:31

       And then the-
5. 11:32 – 12:08

   David Sacks vs Government: "You're Asking Us to Regulate What?"

   1. MCMartin Casado11:32

      Well, my favorite thing is like, is, was it David Sacks was like... I don't, I don't... I think it was David Sacks, but w- someone from the government said, "Uh, you're asking us to regulate you. No." [laughs] So like basically the answer was no.

   2. SSSteven Sinofsky11:46

      Well, but the thing that-

   3. ALAaron Levie11:47

      That was probably just Trump, I think. But, uh-

   4. SSSteven Sinofsky11:49

      But the thing that they know now-

   5. ALAaron Levie11:50

      Yeah

   6. SSSteven Sinofsky11:50

      ... that, that you just know from experience is once the wheels start on regulating-

   7. ALAaron Levie11:55

      You can't slow that one down.

   8. MCMartin Casado11:56

      No.

   9. SSSteven Sinofsky11:56

      You, you... And now it's become an election issue for every party in every jurisdiction up and down the whole government stack.

   10. ALAaron Levie12:01

       Yeah, yeah.

   11. SSSteven Sinofsky12:02

       So there is, there's now this- This whole basket goods, basket of, of regulatory approaches

   12. ALAaron Levie12:07

       Well, the, the

6. 12:08 – 28:39

   2028 as the AI Election

   1. ALAaron Levie12:08

      next election will 100% be a referendum on, on AI. So, like it has to happen that, that 2028 is like the AI election, and you could basically run on... The problem is, is like it's not obvious what- who would run on the pro-AI story because it would be, it's gonna be too nebulous-

   2. SSSteven Sinofsky12:25

      No

   3. ALAaron Levie12:25

      ... to tell that story. So then it's just, it's just basically varying degrees of how much do you regulate it or at least try and like avoid the topic.

   4. SSSteven Sinofsky12:33

      Yeah.

   5. ALAaron Levie12:33

      But it is, it is too bad that we've, we as a country are in a spot where like the pro-AI case is just like, it sounds too-- It's like, it's just like takes too many words, you know, it's way too nuanced.

   6. SSSteven Sinofsky12:45

      Well, no, it's defensive.

   7. ALAaron Levie12:46

      Uh, yeah, it's defensive.

   8. SSSteven Sinofsky12:46

      It's defensive-

   9. ALAaron Levie12:47

      Yes

   10. SSSteven Sinofsky12:47

       ... and we don't, and we, we own none of the vocabulary.

   11. ALAaron Levie12:49

       Right. Right.

   12. SSSteven Sinofsky12:50

       So like the whole debate is, is, is pause, it's, it's swarms, it's rogue. Every word-

   13. ALAaron Levie12:57

       Mm-hmm

   14. SSSteven Sinofsky12:57

       ... has been chosen by the people who don't wanna do AI.

   15. ALAaron Levie13:00

       Yeah.

   16. SSSteven Sinofsky13:00

       And so it means the first thing you have to do is invent new words and say that their words are wrong, which takes so many words that-

   17. ALAaron Levie13:06

       Yeah, so we gotta be like union jobs and, and, you know, cancer and like, you know, there, we need, you know, there needs to be another word cloud that emerges.

   18. MCMartin Casado13:15

       I mean, what I, what I don't, what I don't understand is why the labs have not taken a position on X risk. Like short of that-

   19. ALAaron Levie13:21

       Yeah

   20. MCMartin Casado13:21

       ... I don't think this goes in any direction other than heavy-handed regulation.

   21. ALAaron Levie13:25

       Yeah. Well, the-

   22. MCMartin Casado13:25

       It would, it would just be negligent of the government to be like, "There's a 10% chance of species extinction." The CEO of the top company says he agrees with it. Like h-how can a government not do heavy-handed regulation?

   23. ALAaron Levie13:37

       Well, but what, who, but what would anybody real- knowing the, knowing this, this ecosystem though, I don't know that you would be able to pin anybody down on that other than something-

   24. MCMartin Casado13:46

       No, I would say Dario said it on-

   25. ALAaron Levie13:47

       No, but I'm saying you're not gonna pin anybody down on a lower number.

   26. SSSteven Sinofsky13:49

       Well, well, in fact, he, he, he does the worst thing about it, which is he agrees with people who claim that they believe that there's an X percentage of extinction happening.

   27. ALAaron Levie13:59

       Yeah.

   28. SSSteven Sinofsky13:59

       But he w- he specifically goes out of his way to say, "I'm not gonna put a percentage on it."

   29. ALAaron Levie14:03

       Mm-hmm.

   30. SSSteven Sinofsky14:04

       Which I just think is the weirdest-
7. 28:39 – 32:30

   Why Cybersecurity Has Historically Rejected Practical Solutions

   1. MCMartin Casado28:39

      Can, can I, can I just interject an annoying aside? So, um, which is, which is-

   2. SSSteven Sinofsky28:43

      Which is more annoying than what I just did. [laughs]

   3. MCMartin Casado28:44

      Which is... No, no, this is good. Which is very in line with this, which is, um, I've been in the security com- like the actual cybersecurity community for a long time, and it is, you know, it's always been one of these things where like they just don't like practical solutions, so even if you build like, you know, a secure system, like well what if somebody, you know, shows up-

   4. SSSteven Sinofsky28:59

      Oh, yeah, yeah

   5. MCMartin Casado29:00

      ... and like, you know, like, you know, can Russell Crowe can like break the encryption or something like that. And you-

   6. SSSteven Sinofsky29:04

      No, it's, it's "Mission Impossible." Tom Cruise just crawling through the ceiling.

   7. MCMartin Casado29:06

      Yeah, no, yeah. There's, there's always like these kinda like whatever. So my favorite thing that happened recently was like Noam Brown-

   8. ALAaron Levie29:11

      Yeah, yeah.

   9. MCMartin Casado29:11

      On some podcast. We've all said stupid stuff on podcasts. I've already said stupid stuff.

   10. ALAaron Levie29:13

       Oh, you didn't like this one?

   11. MCMartin Casado29:15

       No, no, it was great. No.

   12. ALAaron Levie29:16

       I thought it was fun.

   13. MCMartin Casado29:17

       No, it was fantastic.

   14. ALAaron Levie29:17

       Okay. [laughs]

   15. MCMartin Casado29:18

       No, no, no, no.

   16. ALAaron Levie29:18

       Okay, okay.

   17. MCMartin Casado29:19

       I'm setting, I'm setting it up.

   18. ALAaron Levie29:19

       Okay, okay, okay.

   19. MCMartin Casado29:20

       So, so Noam Brown was like, "Listen, you know, uh, uh, you don't know what a superintelligence could do. It could maybe use like the heat of a, uh, CPU to exfiltrate itself to another computer." Which this brought me back to my-

   20. ALAaron Levie29:32

       Okay, okay, okay.

   21. MCMartin Casado29:32

       And I'm like, "I'm very, now I'm very comfortable." [laughs]

   22. ALAaron Levie29:34

       Okay, okay, okay.

   23. MCMartin Casado29:35

       Like great, now I can have endless pointless discussions on this. But what is interesting is you basically have the X-risk people saying something they thought was plausible.

   24. ALAaron Levie29:42

       Mm.

   25. MCMartin Casado29:42

       And then you have like the security people having this kind of endless discussion. And so I think at some level now these communities are being bridged.

   26. ALAaron Levie29:50

       Yeah.

   27. MCMartin Casado29:50

       Which is like, you know, I actually think that was-

   28. ALAaron Levie29:52

       Huh

   29. MCMartin Casado29:52

       ... a very reasonable thing for Noam Brown to say. I think you can pick holes in it, but we all say weird stuff on podcasts. I actually think exfil risk is real. I mean, I worked in secure, um, computing environments that were highly classified that you would... The covert channels were unbelievable.

   30. SSSteven Sinofsky30:06

       Oh, yeah, yeah.
8. 32:30 – 48:38

   The Craziest Covert Channel Ever Seen

   1. MCMartin Casado32:30

      Can I tell you the craziest covert channel I've ever seen? So it turns... Remember the, remember the OCR? I know this is, like, one of my-

   2. SSSteven Sinofsky32:34

      Yeah.

   3. MCMartin Casado32:35

      So I-

   4. SSSteven Sinofsky32:35

      I'm glad that someone's older than me. Not really, but acting it. [laughs]

   5. MCMartin Casado32:38

      So remember the old CRT? So it turns out if, like, let's say it's night and you're, you're using a, a CRT terminal i- in your room. The lightest thing in the room is actually the pixel that the raster beam is on.

   6. SSSteven Sinofsky32:48

      Yeah.

   7. MCMartin Casado32:48

      So most people think it's, like, the glow of the monitor, but it's actually that given pixel.

   8. SSSteven Sinofsky32:51

      Mm-hmm.

   9. MCMartin Casado32:52

      So somebody figured out that, like, let's say you're in, like, you know, a hotel room and you're on your computer. If you have something that can sample the color of the window, you can reconstruct the screen.

   10. SSSteven Sinofsky33:02

       Oh, that's crazy.

   11. MCMartin Casado33:03

       Right? You just do it at the same hertz that the raster beam is moving. Then somebody else figured out if you can subvert three pixels, you can use those, you know, 'cause it just looks like bad pixels, you can use those to basically send a message. So, like, you could literally, like, sit out in-

   12. SSSteven Sinofsky33:15

       Right

   13. MCMartin Casado33:15

       ... whatever, like a parking lot.

   14. SSSteven Sinofsky33:16

       Do your own SOS.

   15. MCMartin Casado33:17

       You sample the message and you can... It is a relatively high bandwidth one-way communication.

   16. SSSteven Sinofsky33:21

       Right.

   17. MCMartin Casado33:22

       And so, like, what Noam Brown was saying, like, maybe heat is not the way to do it, but that level of sophistication-

   18. SSSteven Sinofsky33:26

       Right

   19. MCMartin Casado33:26

       ... is actually real.

   20. SSSteven Sinofsky33:27

       I, my-

   21. MCMartin Casado33:27

       Like, that's a, that's a thing.

   22. SSSteven Sinofsky33:29

       My, I... When I was at the missile factory, like, I had to lock my keyboard up. I, that's really an impolite word, but that's what we call, call some. I had to lock my keyboard up at night because they didn't want the custodians who didn't have clearance walking by and, and just noticing which keys were dirtier or cleaner.

   23. MCMartin Casado33:44

       Oh, yeah.

   24. SSSteven Sinofsky33:44

       And I once left it out, and there's, like, a, a note from security, you know, telling me to report to security and pick up my keyboard. Like, the, the guard-

   25. MCMartin Casado33:53

       Right

   26. SSSteven Sinofsky33:53

       ... who walked the hallways just took the keyboard that night off my machine, and that's, like, baseline. I was not cleared. I was just that sensitive, you know? I was, like, nothing. I was an intern.

   27. MCMartin Casado34:01

       The big problem with this conversation is now this is all in the training data of every AI model-

   28. SSSteven Sinofsky34:06

       Yeah, yeah

   29. MCMartin Casado34:06

       ... in the future, so w- w-

   30. SSSteven Sinofsky34:07

       But that's also-
9. 48:38 – 53:01

   LLMs as Decision Engines vs Chatbots

   1. ALAaron Levie48:38

      it so remarkable, yeah.

   2. SSSteven Sinofsky48:39

      Well, okay. So, so the way I think about it is, um-

   3. MCMartin Casado48:42

      So, okay, so LLMs were kind of text in, text out, right? They generate text. And, and, and they came from chat, right? It was to communicate with a human, and we've spent the last few years ti- trying to take this thing that spits out text and cram it into a, a traditional program.

   4. SSSteven Sinofsky48:54

      Mm.

   5. MCMartin Casado48:55

      Right? And but traditional programs don't really speak text, right? And so then you end up doing this janky thing-

   6. SSSteven Sinofsky49:00

      [chuckles]

   7. MCMartin Casado49:00

      ... where you're like, in the prompt you're like-

   8. SSSteven Sinofsky49:02

      Here's the schema and-

   9. MCMartin Casado49:02

      Here's the schema.

   10. SSSteven Sinofsky49:03

       Yeah, yeah.

   11. MCMartin Casado49:04

       But like the thing is generating text-

   12. SSSteven Sinofsky49:05

       Yeah

   13. MCMartin Casado49:05

       ... and it kind of ignores it, and it's just been super janky. And so what, what Jeff basically said is that, "Listen, um, you know, generating the text on the out- is, is a very expensive thing, but it's also kind of, you know, it's, it's more complicated than you need. So why don't we, we, we'll, we'll, we'll read text and we'll have all of that kind of knowledge to read the text, but then rather than generating text, which is very expensive, we will just, if you give us a set of options, we'll choose the best option." We can do that incredibly, um, we can do it incredibly fast, incredibly cheaply, uh, but also we can do it with much more accuracy 'cause we can train just for this. And so for all of the use cases that are not talking to, to like a, a chatbot, but are actually trying to put it in traditional software, this is a great fit. And so this has probably been the fastest adoption of an AI model since ChatGPT. It's just been remarkable because we're all primed for this.

   14. SSSteven Sinofsky49:51

       I just wanna pile on, on this one 'cause I, I can't tell you how much I love seeing this exact form of, of, of innovation and because what it's does is it does the thing that's bugged me from the very beginning, which is there's been no user study ever that shows like interacting with a computer using, um, full natural language is efficient.

   15. MCMartin Casado50:11

       Hmm.

   16. SSSteven Sinofsky50:12

       It's, it's like literally always the least efficient way, and it's very simple, and it, it's just like ask yourself, how many people are asked, are really, really good at asking questions? And immediately that's like less than half the people can ask a good question in a meeting.

   17. MCMartin Casado50:25

       Yeah.

   18. SSSteven Sinofsky50:25

       And then how often do you look at the answer and get really frustrated before it's finished, but you have to pay all this money to watch the seven paragraphs come out and then apologize-

   19. MCMartin Casado50:33

       [chuckles]

   20. SSSteven Sinofsky50:33

       ... that it's only a little? And so that's one, like having a different model. And then the other, of course, is my favorite, which is the output of it is designed for probabilistic programming.

   21. MCMartin Casado50:43

       Yeah. Yeah.

   22. SSSteven Sinofsky50:43

       And so instead of saying like, "Is this a customer service question?" Then route to customer service, otherwise route to general help desk or whatever. It's like, "Well, this is 80% customer service."

   23. MCMartin Casado50:54

       Mm-hmm.

   24. SSSteven Sinofsky50:54

       And that's exactly simulation.

   25. MCMartin Casado50:57

       Mm.

   26. SSSteven Sinofsky50:57

       And it turns out there's like 50 years of computer science research in literally like probabilistic if statements.

   27. MCMartin Casado51:02

       Mm.

   28. SSSteven Sinofsky51:03

       And so suddenly the coolest place to be in computer science is gonna be in probabilistic programming-

   29. MCMartin Casado51:08

       Mm

   30. SSSteven Sinofsky51:08

       ... which was like all of computer science in the 1960s and '70s. So it was basically how do-- Because all of computers started with doing math, and it was all simulation.
10. 53:01 – 54:49

    Why the Labs Haven't Built This: The Being vs Tool Mindset

    1. SSSteven Sinofsky53:01

       in the if statement.

    2. MCMartin Casado53:01

       But the consequential thing is like finally we have a way to integrate these language models into a traditional-

    3. SSSteven Sinofsky53:07

       New software. Right

    4. MCMartin Casado53:08

       ... software. And, and I think it's kind of funny 'cause it, like you asked the question like why haven't the labs done this, right?

    5. SSSteven Sinofsky53:12

       Right.

    6. MCMartin Casado53:13

       And it, [chuckles] it's kind of like a r- like not an indictment, but a reflection on how they think. Like they're trying to create beings, and beings speak.

    7. SSSteven Sinofsky53:19

       Mm.

    8. MCMartin Casado53:19

       If you're trying to create God, God speaks in natural languages or whatever, where this is really about something that's for traditional software, which is kinda not the direction that they've been taking. But one of the reasons the uptick has been so dramatic is because a lot of us software people have been trying to integrate these models into software-

    9. SSSteven Sinofsky53:36

       Mm

    10. MCMartin Casado53:36

        ... it just hasn't. So even before you get to the probabilistic, like if I want a language model to drive an if statement, like it's really hard today-

    11. SSSteven Sinofsky53:43

        Yeah, yeah

    12. MCMartin Casado53:43

        ... with this model, it makes it much, much, much easier. And then of course this could change the nature of software fundamentally to make it more stochastic over time.

    13. SSSteven Sinofsky53:50

        Well, I think, but I, I, absolutely, and I think that what's so cool is that, that it is happening outside the models because that's what I think is just gonna happen, which is the center of innovation has just moved.

    14. MCMartin Casado54:02

        Yeah.

    15. SSSteven Sinofsky54:02

        And it's just, and now people need to... Like it turns out that the plat-

    16. MCMartin Casado54:05

        Oh, you mean outside of the lab, the big-

    17. SSSteven Sinofsky54:07

        Right

    18. MCMartin Casado54:07

        ... the front ends, yeah.

    19. SSSteven Sinofsky54:07

        The platform providers-

    20. MCMartin Casado54:08

        Yeah, yeah, 100%

    21. SSSteven Sinofsky54:09

        ... you know, basically reach a critical, a point of critical mass where the innovation stops happening at the platform layer.

    22. MCMartin Casado54:15

        Mm.

    23. SSSteven Sinofsky54:15

        And, and then, you know, Apple, there's this famous expression in the Apple community called Sherlocking, where Apple looks around and the, the things from the outside world become features and people complain and it's a real... But that's sort of how the innovation works because once you're a platform, you're overwhelmed, no matter how many people you add, you're overwhelmed with just keeping the thing running and compatibility and stuff like that. And so I think that this is the signal that now people have figured out that there's innovation to be done-

    24. MCMartin Casado54:43

        That's awesome

    25. SSSteven Sinofsky54:44

        ... to the model-

    26. MCMartin Casado54:44

        Yeah

    27. SSSteven Sinofsky54:44

        ... but outside the model.

    28. MCMartin Casado54:46

        Yep. Yep.

    29. SSSteven Sinofsky54:46

        Cool.

    30. ETErik Torenberg54:47

        Guys, thanks for coming. This was a great episode.

Episode duration: 55:03