Expert AI as a Healthcare Superpower

1. 0:00 – 0:20

   Intro

   1. VPVijay Pande0:00

      [upbeat music]

2. 0:20 – 3:28

   Why AI Has Reached a Critical Mass Moment

   1. VPVijay Pande0:20

      Hey, I'm Vijay Pande. I'm the founding general partner of the a16z Bio & Health Fund.

   2. MAMarc Andreessen0:25

      Um, and I'm Marc Andreessen, uh, co-founder of, uh, of a16z.

   3. VPVijay Pande0:29

      Marc, thank you so much for joining us.

   4. MAMarc Andreessen0:31

      Yeah, it's great to be here.

   5. VPVijay Pande0:32

      Yeah. So, you know, you famously wrote about software eating the world.

   6. MAMarc Andreessen0:35

      Mm-hmm.

   7. VPVijay Pande0:35

      And that was, um, basically, what, 10-plus years ago?

   8. MAMarc Andreessen0:38

      Yeah.

   9. VPVijay Pande0:38

      And actually, that very much seems to have come to fruition if you look at all these other industries that software really wasn't a part of. Uh, software's actually become a dominant part. But, uh, actually, this year's been kind of an amazing year for another type of software, for AI.

   10. MAMarc Andreessen0:52

       Mm-hmm.

   11. VPVijay Pande0:52

       And, and I'm curious to sort of talk about the arc of what we think is gonna happen in the future based on what we've seen in the past and, and really how this new technology is gonna change everything, much like we've seen software change the last 10 years. I'm curious what you think for just, like, this year.

   12. MAMarc Andreessen1:06

       Yeah.

   13. VPVijay Pande1:06

       It's been kind of an amazing year. Y- we always seem like not much happens in any given year, but 2022 seems to have been an amazing year for AI.

   14. MAMarc Andreessen1:13

       Right. Well, so Vladimir Lenin, um-

   15. VPVijay Pande1:16

       Yes

   16. MAMarc Andreessen1:16

       ... once said-

   17. VPVijay Pande1:17

       Yes

   18. MAMarc Andreessen1:17

       ... um, "There are decades in which nothing happens."

   19. VPVijay Pande1:19

       Mm-hmm.

   20. MAMarc Andreessen1:19

       Um, "And then there are weeks, uh, in which decades happen."

   21. VPVijay Pande1:21

       Yes.

   22. MAMarc Andreessen1:22

       Um, and, uh, let's not hope that happens-

   23. VPVijay Pande1:24

       Yeah

   24. MAMarc Andreessen1:24

       ... politically anymore.

   25. VPVijay Pande1:25

       Yes.

   26. MAMarc Andreessen1:25

       Um, but, uh, it does happen, you know, in science and technology. It does happen. There are sort of moments where things kind of hit critical mass.

   27. VPVijay Pande1:30

       Yeah.

   28. MAMarc Andreessen1:31

       Um, and, uh, you know, this, the, the sort of AI machine learning revolution seem, seem, seems like that's what's happening right now. You know, it's been interesting to watch. You know, it, it's, it's sort of like it, it feels to me at least it was like there was, like, a breakthrough moment in 2012, right?

   29. VPVijay Pande1:43

       Yeah.

   30. MAMarc Andreessen1:43

       That had to do with images.
3. 3:28 – 5:45

   Training Models, Kids, and the AGI Hand Wave

   1. VPVijay Pande3:28

      Yeah, well, and it seems almost less like writing software, almost like training something.

   2. MAMarc Andreessen3:32

      Yeah. That's right.

   3. VPVijay Pande3:33

      Like, when I think about machine learning and image recognition you talked about, it's felt like it was almost like training a dog.

   4. MAMarc Andreessen3:37

      Yeah.

   5. VPVijay Pande3:37

      Right? Like, reinforcement learning is, like, we'll give treats-

   6. MAMarc Andreessen3:40

      Yeah

   7. VPVijay Pande3:41

      ... uh, as it gets better. But there's something different now. It feels like, I don't know, we've gone from, like, training a dog to recognize a bird versus, like, a hot dog or hot dog or not hot dog or so on to actually something where it feels closer to, like, training a person.

   8. MAMarc Andreessen3:52

      Yeah.

   9. VPVijay Pande3:52

      Or I don't know how you feel. Like, when we talk about learning and training and data, like, what, what are we training? What is... Where, where do you think we are in that arc of getting to, like, eventually HAL 9000 and so on?

   10. MAMarc Andreessen4:02

       Well, so I, I, uh-

   11. VPVijay Pande4:03

       Yeah

   12. MAMarc Andreessen4:03

       ... while, while this has been happening, you know, I have-

   13. VPVijay Pande4:05

       Yeah

   14. MAMarc Andreessen4:05

       ... you have kids like I have a, you have a, you know, young, young child.

   15. VPVijay Pande4:07

       Yeah.

   16. MAMarc Andreessen4:07

       So I have a seven-year-old now. So a- as this stuff has all been popping-

   17. VPVijay Pande4:10

       Yeah

   18. MAMarc Andreessen4:10

       ... I've been simultaneously training, training now the, the seven-year-old.

   19. VPVijay Pande4:13

       Yes. Yes.

   20. MAMarc Andreessen4:13

       Yeah. [laughs]

   21. VPVijay Pande4:14

       Yes.

   22. MAMarc Andreessen4:15

       You know, any- anybody who's had, who's had kids will recognize what, you know, kind of what I'm, what I'm about to say. But, um, you know, it is really interesting watching little kids. The, the way I think about, or at least, at least, at least the little kid I have, uh, who's great, um, it's like, you know, it's like everything... For the first, for the first few years, it's like every single thing he did was, like, a little applied physics experiment.

   23. VPVijay Pande4:32

       Yeah.

   24. MAMarc Andreessen4:32

       Which is like, "Let's see what happens if I drop this."

   25. VPVijay Pande4:34

       Yeah.

   26. MAMarc Andreessen4:34

       "Let's see what happens if I eat this."

   27. VPVijay Pande4:36

       Yes. [laughs]

   28. MAMarc Andreessen4:36

       "Let's see what happens if I do this to Daddy."

   29. VPVijay Pande4:37

       Yes.

   30. MAMarc Andreessen4:38

       Right? And see what the response is.
4. 5:45 – 11:28

   GPT's Sleight of Hand and Human Creativity

   1. VPVijay Pande5:45

      But, but, uh, where we are now is kind of amazing. Like, can, people can take, like, GPT-3, you can give it SAT exams.

   2. MAMarc Andreessen5:52

      Mm-hmm.

   3. VPVijay Pande5:52

      It can do okay. It can, like, uh-

   4. MAMarc Andreessen5:54

      Actually, it can do quite well.

   5. VPVijay Pande5:55

      Yeah, yeah, it can do reasonable-

   6. MAMarc Andreessen5:56

      It scores at, like, the-

   7. VPVijay Pande5:56

      Yeah.

   8. MAMarc Andreessen5:56

      So the one I saw, it scored at, like, what, 1200?

   9. VPVijay Pande5:58

      Yeah, yeah, something like that. Yeah, so it's not bad, right? Uh, it can do homework.

   10. MAMarc Andreessen6:02

       That, that would get you in a lot of colleges, right?

   11. VPVijay Pande6:03

       Yeah. I actually gave it, like, asked it questions like, uh, to explain the derivation for the Schwarzschild radius, you know, the black hole radius, to write a code for f- let's say eight-by-eight tic-tac-toe. Like, random things that it should never be able to do because it's not just memorizing it, it's generalizing.

   12. MAMarc Andreessen6:20

       Right.

   13. VPVijay Pande6:20

       Uh, and it's getting that. But then also it actually seems to have some sort of weir- weird hiccups.

   14. MAMarc Andreessen6:25

       Yeah.

   15. VPVijay Pande6:25

       In that, uh, actually one thing that it really does not seem to get is humor.

   16. MAMarc Andreessen6:28

       Right.

   17. VPVijay Pande6:29

       ... you know. Um, so [laughs] I'm kinda curious where you think it's gonna go because before we get to AGI, there are things that, uh, an average human can do pretty well that GPT-3 can't. But then there's also what experts can do.

   18. MAMarc Andreessen6:41

       Yeah.

   19. VPVijay Pande6:41

       And, uh, what I'm very curious about is actually we may get to some of the expert stuff first-

   20. MAMarc Andreessen6:46

       Yeah

   21. VPVijay Pande6:46

       ... uh, before it can do even something like humor.

   22. MAMarc Andreessen6:49

       Yeah.

   23. VPVijay Pande6:49

       The irony is that something like humor that we take for granted might actually be really hard-

   24. MAMarc Andreessen6:54

       Right

   25. VPVijay Pande6:54

       ... and other areas might be easier.

   26. MAMarc Andreessen6:55

       Right.

   27. VPVijay Pande6:56

       Yeah.

   28. MAMarc Andreessen6:56

       Well, the ultimate example are the things it can't do, like it can't like pack your suitcase.

   29. VPVijay Pande6:59

       Yeah, yeah.

   30. MAMarc Andreessen7:00

       Like, there's no robot that will pack your suitcase.
5. 11:28 – 14:36

   The Turing Test, Self-Awareness, and the PhD Bar

   1. MAMarc Andreessen11:28

      Ha- having said that, um, you know, again... Oh, oh, and then that, that takes me to the other, the other thing on AI, which is I, I do... You know, one of the ways that people are testing AI is-

   2. VPVijay Pande11:36

      Mm

   3. MAMarc Andreessen11:36

      ... with the so-called Turing test, right?

   4. VPVijay Pande11:37

      Yes.

   5. MAMarc Andreessen11:38

      And the, and the, the-

   6. VPVijay Pande11:38

      Mm

   7. MAMarc Andreessen11:38

      ... simplified form of the Turing test is you're chatting with-

   8. VPVijay Pande11:41

      Mm

   9. MAMarc Andreessen11:41

      ... somebody that may be a human or may be a bot-

   10. VPVijay Pande11:43

       Right

   11. MAMarc Andreessen11:43

       ... and you chat for 20 minutes, and at the end-

   12. VPVijay Pande11:45

       Yeah

   13. MAMarc Andreessen11:45

       ... can you guess better than random as to whether-

   14. VPVijay Pande11:46

       Yes

   15. MAMarc Andreessen11:46

       ... it's a human or a bot?

   16. VPVijay Pande11:47

       Yeah.

   17. MAMarc Andreessen11:48

       You know, the, the... My take on that is the Turing... You know, Alan Turing was a genius, but the Turing test is malformed.

   18. VPVijay Pande11:53

       Yes.

   19. MAMarc Andreessen11:53

       Humans are too easy to trick.

   20. VPVijay Pande11:54

       Yeah, yeah, yeah. Yeah.

   21. MAMarc Andreessen11:55

       But, but that's too low of a bar-

   22. VPVijay Pande11:56

       Mm-hmm

   23. MAMarc Andreessen11:56

       ... because tricking a person is not that hard-

   24. VPVijay Pande11:59

       Mm

   25. MAMarc Andreessen11:59

       ... and does not prove anything other than that you've tricked the person like this.

   26. VPVijay Pande12:01

       Yeah.

   27. MAMarc Andreessen12:01

       Like I, I think... And, and this is relevant 'cause I think the-

   28. VPVijay Pande12:04

       Mm

   29. MAMarc Andreessen12:04

       ... you know, things like GPT are about to pass the Turing test-

   30. VPVijay Pande12:06

       Yes, yes
6. 14:36 – 20:02

   Expertise, Taste, and Market Tests

   1. VPVijay Pande14:36

      Well, so in the PhD, let's talk about like at least in that old school mentality of a PhD of some advanced learning where you become an expert in something.

   2. MAMarc Andreessen14:43

      Right.

   3. VPVijay Pande14:44

      You know, I think that's the thing where, uh-

   4. MAMarc Andreessen14:45

      Expert. What do you mean by expert?

   5. VPVijay Pande14:46

      Uh, let's say the ability to be in the top .1% of-

   6. MAMarc Andreessen14:50

      Of what?

   7. VPVijay Pande14:50

      ... humanity of, let's say, designing a drug or building-

   8. MAMarc Andreessen14:54

      Oh, actually doing something

   9. VPVijay Pande14:54

      ... doing something. Yeah, yeah, yeah, yeah, yeah.

   10. MAMarc Andreessen14:55

       Oh, interesting. Is that what they, is that what they teach?

   11. VPVijay Pande14:57

       Yeah. Well that's, that's-

   12. MAMarc Andreessen14:58

       Is that what they teach in universities? I have no idea

   13. VPVijay Pande14:59

       ... that is my goal, certainly. [laughs]

   14. MAMarc Andreessen15:00

       I w- I wasn't aware, I wasn't aware that was part of the curriculum.

   15. VPVijay Pande15:02

       I think it is.

   16. MAMarc Andreessen15:03

       Is it?

   17. VPVijay Pande15:03

       Sometimes. [laughs]

   18. MAMarc Andreessen15:05

       [laughs]

   19. VPVijay Pande15:05

       Or at least that's what you have to do eventually when you get out.

   20. MAMarc Andreessen15:08

       Right. Okay.

   21. VPVijay Pande15:08

       Yeah, you know? And that you have to apply it, and I think-

   22. MAMarc Andreessen15:10

       Right

   23. VPVijay Pande15:10

       ... it's where... I think it is... One of the things about being an expert in my mind is that th- something that is the difference between bad, good, and great-

   24. MAMarc Andreessen15:20

       Right

   25. VPVijay Pande15:20

       ... can be really close.

   26. MAMarc Andreessen15:22

       Right.

   27. VPVijay Pande15:22

       Like, um, I could probably write a piece of music, but no one would think it's all that great.

   28. MAMarc Andreessen15:26

       Right.

   29. VPVijay Pande15:26

       You know? And then you could have someone who's a good musician, but not a great one, then you have like a genius.

   30. MAMarc Andreessen15:31

       Right.
7. 20:02 – 23:27

   Do We Understand Consciousness Well Enough for AGI?

   1. VPVijay Pande20:02

      Or, or so why, why would you say, why, why wouldn't it get there?

   2. MAMarc Andreessen20:04

      'Cause like do we even understand how it works in people? Like we-

   3. VPVijay Pande20:07

      Yeah, but we, maybe we don't have to.

   4. MAMarc Andreessen20:08

      Well, maybe we don't have to.

   5. VPVijay Pande20:10

      Yeah, yeah.

   6. MAMarc Andreessen20:10

      Th- so this is where I describe-

   7. VPVijay Pande20:11

      Yeah, yeah

   8. MAMarc Andreessen20:11

      ... this, this is like the AGI question.

   9. VPVijay Pande20:12

      Yeah.

   10. MAMarc Andreessen20:12

       This is where I, uh, call the hand wave.

   11. VPVijay Pande20:14

       Yeah, yeah.

   12. MAMarc Andreessen20:14

       Or, or it's, or the second-

   13. VPVijay Pande20:15

       Yeah.

   14. MAMarc Andreessen20:15

       The embedded assumption that it's a when is that-

   15. VPVijay Pande20:17

       Yeah

   16. MAMarc Andreessen20:17

       ... it will be an emergent process-

   17. VPVijay Pande20:18

       Yeah, yeah

   18. MAMarc Andreessen20:18

       ... that will sort of unlock as a consequence of greater and greater levels of scale.

   19. VPVijay Pande20:21

       Yeah, yeah.

   20. MAMarc Andreessen20:22

       Maybe.

   21. VPVijay Pande20:22

       Yeah, yeah.

   22. MAMarc Andreessen20:24

       One way of looking at that is yes, that is-

   23. VPVijay Pande20:25

       Yeah

   24. MAMarc Andreessen20:25

       ... just what's gonna happen. That's how-

   25. VPVijay Pande20:26

       Yeah

   26. MAMarc Andreessen20:26

       ... human consciousness emerged.

   27. VPVijay Pande20:27

       Yes, yes.

   28. MAMarc Andreessen20:27

       Like that's obviously what's gonna happen.

   29. VPVijay Pande20:28

       Yeah.

   30. MAMarc Andreessen20:29

       The other approach to that is it's just a massive, it's a hand wave. It's a hand wave-
8. 23:27 – 26:51

   Better Than Human: Self-Driving Cars and Medical Diagnosis

   1. VPVijay Pande23:27

      But okay, so but the, I would pull back from this. So AGI is, is the fun thing. There is a sort of step back, which is to pick a domain, and, and you know the domains I, I think a lot about, like life sciences, um, designing drugs, doing healthcare, like seeing if you can do a di- pick a diagnosis, can you suggest a drug. In those areas, now we're talking about a much more limited domain.

   2. MAMarc Andreessen23:47

      Yeah.

   3. VPVijay Pande23:48

      So we're not talking about, we don't, we don't need to go all the way to consciousness for that necessarily.

   4. MAMarc Andreessen23:52

      Right.

   5. VPVijay Pande23:52

      You can, you can have something that's more limited. In that limited domain- Right now, it seems like generative AI isn't quite far enough yet to be able to, like, j- uh, yeah, I don't, I don't see the examples quite yet.

   6. MAMarc Andreessen24:03

      Are you sure?

   7. VPVijay Pande24:04

      Yeah. Well, that, it's... We'll, we'll see. I mean, uh, so-

   8. MAMarc Andreessen24:06

      Yeah

   9. VPVijay Pande24:06

      ... yeah, what's the, what's the counter? And I, I know you, uh, especially thinking about healthcare a lot.

   10. MAMarc Andreessen24:09

       Yeah.

   11. VPVijay Pande24:10

       Yeah.

   12. MAMarc Andreessen24:10

       Well, so the first thing is whenever you're score- Well, let's talk about, like, medical diagnosis-

   13. VPVijay Pande24:13

       Yeah

   14. MAMarc Andreessen24:13

       ... which is kind of just low-hanging fruit-

   15. VPVijay Pande24:14

       Yeah, yeah, yeah

   16. MAMarc Andreessen24:15

       ... uh, question 'cause everybody-

   17. VPVijay Pande24:15

       Yeah

   18. MAMarc Andreessen24:15

       ... everybody experiences it. So to start up front, you have to ask the question up front, which is, like, is the goal... What, what's the threshold? Is the threshold perfect-

   19. VPVijay Pande24:22

       Yeah

   20. MAMarc Andreessen24:22

       ... or is the threshold better than human?

   21. VPVijay Pande24:23

       Yeah. That's a great point. Yeah.

   22. MAMarc Andreessen24:24

       Right?

   23. VPVijay Pande24:24

       Yeah. Yeah.

   24. MAMarc Andreessen24:25

       Um, and, and by the way, this, you know, this is a topic that comes up all the time with self-driving cars.

   25. VPVijay Pande24:28

       Yes.

   26. MAMarc Andreessen24:28

       Right?

   27. VPVijay Pande24:28

       Yes.

   28. MAMarc Andreessen24:28

       Which is, is it perfect? It will never make a mistake, or is it just gonna be better, better than human? And the way that self-driving cars score this is accidents per thousand miles driven.

   29. VPVijay Pande24:35

       Yeah.

   30. MAMarc Andreessen24:35

       And self-driving cars are already lower than, than human drivers.
9. 26:51 – 30:22

   Group Decisions and Wisdom of Crowds

   1. MAMarc Andreessen26:51

      Well, and then there's-

   2. VPVijay Pande26:51

      Yeah

   3. MAMarc Andreessen26:51

      ... and then there's all the other sort of things. So then there's, like, drug interaction. You know, is any one doctor tracking all the interactions of your drugs?

   4. VPVijay Pande26:57

      Yeah.

   5. MAMarc Andreessen26:57

      Then there's this other issue, which is, okay, they give the prescription. Does, is there actually compliance for taking the prescription?

   6. VPVijay Pande27:02

      Yes, yes.

   7. MAMarc Andreessen27:02

      Does the doctor actually know whether you're taking the prescription?

   8. VPVijay Pande27:04

      Yes.

   9. MAMarc Andreessen27:04

      Are they able to evaluate-

   10. VPVijay Pande27:04

       Oh, compliance is one of the biggest disasters. Yeah.

   11. MAMarc Andreessen27:06

       Right.

   12. VPVijay Pande27:06

       Yeah.

   13. MAMarc Andreessen27:06

       But, but that means, like, the ability for a median doctor to even evaluate the success of a treatment, they actually may not be able to do it 'cause they may not have the data on compliance.

   14. VPVijay Pande27:12

       Yeah, yeah, yeah.

   15. MAMarc Andreessen27:13

       And so, like, you look at the existing s- I don't know. For me, you look at the existing system by which this all happens. It's very similar to looking at the existing system by which people actually drive cars-

   16. VPVijay Pande27:20

       Yes, yes

   17. MAMarc Andreessen27:20

       ... which is like, oh, my God, this is not good.

   18. VPVijay Pande27:22

       [laughs] This is not good.

   19. MAMarc Andreessen27:23

       Like, this is really not good.

   20. VPVijay Pande27:24

       Yeah.

   21. MAMarc Andreessen27:24

       Uh, and, and we kind of fool ourselves into believing that it's good 'cause it kind of feels good.

   22. VPVijay Pande27:27

       Yeah.

   23. MAMarc Andreessen27:27

       And we don't really wanna look behind the curtain, but we look behind the curtain, and it's pretty horrifying.

   24. VPVijay Pande27:31

       Yeah. Yeah. [laughs]

   25. MAMarc Andreessen27:32

       And so from that standpoint-

   26. VPVijay Pande27:33

       Yeah

   27. MAMarc Andreessen27:33

       ... if, if you follow that logic, then it says, okay, if the machine could do a better job-

   28. VPVijay Pande27:36

       Yeah

   29. MAMarc Andreessen27:36

       ... you know, if, if the machine was twice as good-

   30. VPVijay Pande27:39

       Yes
10. 30:22 – 36:48

    Augmented Intelligence: Jobs, Doctors, and Teachers

    1. VPVijay Pande30:22

       What do you think society needs to, to do to change? 'Cause there's like all these things we were talking about, and this seems bigger than like just, just the revolution of software over the last 20 years, or internet from last 20 years. 'Cause we're talking about how it changes government, how it changes regulatory, how it changes education.

    2. MAMarc Andreessen30:37

       Right

    3. VPVijay Pande30:37

       I mean, I don't even know where you, where you wanna start with that, but I think that's w- something where it may take us 10 years just s- culturally-

    4. MAMarc Andreessen30:44

       Right

    5. VPVijay Pande30:44

       ... to be able to get ready for this thing that may arrive in 10 years or may already be here.

    6. MAMarc Andreessen30:48

       Right. Right.

    7. VPVijay Pande30:48

       Yeah, I don't know where you wanna start [laughs]

    8. MAMarc Andreessen30:50

       Yeah, so where I would start-

    9. VPVijay Pande30:51

       Yeah

    10. MAMarc Andreessen30:51

        ... is we've already fallen into, I th- we've, we-

    11. VPVijay Pande30:53

        Yeah

    12. MAMarc Andreessen30:53

        ... have deliberately kind of fallen into a trap already, which is we've only been using a single kind of example.

    13. VPVijay Pande30:57

        Yeah. Yeah

    14. MAMarc Andreessen30:57

        W- w- and we've used it both in our discussions on medicine-

    15. VPVijay Pande30:59

        Yes

    16. MAMarc Andreessen30:59

        ... and also in education, which is basically a, something is done today.

    17. VPVijay Pande31:02

        Yes

    18. MAMarc Andreessen31:02

        People are doing something today-

    19. VPVijay Pande31:03

        Yes

    20. MAMarc Andreessen31:03

        ... and then maybe the machine can do it instead.

    21. VPVijay Pande31:05

        Yeah. Yeah.

    22. MAMarc Andreessen31:06

        That's an important thing, and that's-

    23. VPVijay Pande31:07

        Yeah, yeah, yeah

    24. MAMarc Andreessen31:07

        ... that, you know, it's worth thinking about. But the way that technological impact actually plays out in human society is not just that. The way it plays out is it lets you basically revisit more fundamental assumptions.

    25. VPVijay Pande31:16

        Yeah. Or what's not being done today.

    26. MAMarc Andreessen31:17

        Or, or what's not being done today-

    27. VPVijay Pande31:18

        Yeah. Yeah

    28. MAMarc Andreessen31:19

        ... that all of a sudden becomes possible.

    29. VPVijay Pande31:20

        Yeah. Yeah. Yeah.

    30. MAMarc Andreessen31:20

        And, and, and, and this comes, this, this always comes up in any sort of discussion about employment-
11. 36:48 – 38:20

    AI as a Creative Superpower

    1. MAMarc Andreessen36:48

       I think AI's gonna play out in a very similar way. Yeah. Like, there, there are people who will argue, no, AI's different- Yeah ... 'cause it will just keep climbing this ladder- Yeah ... and it'll replace everything. I actually think it's gonna be a, the, basically it's the ultimate superpower. Yes. It's the ultimate pairing. A, a good... We were talking about creating screenplays and, and, and scripts. A good example, if I'm a Hollywood screenwriter today, like, GPT is my best friend. Yes. And I'm just sitting there all day long, and I'm just saying, you know, playing out, and it's like, "Okay, I reached this plot point, dot, dot, dot. Give me a list of, like, 10 ideas for what to do." Yeah. It's like, "Oh, okay, that's an interesting one." Yes, yes. Uh, well, um, I'll give you an example of how this could work. So Meth- M- Mad Men is one of my favorite shows. Mad Men 2. Matthew Weiner, you know, is, r- r- ran that show, and he was always praised. He's like, "Wow, that show is so unpredictable," like, you know, you never knew where it was going. And he said, "Yeah, well, the technique we had in the writers' room was at any given time we had to figure out what happened next in the plot." Yeah. Uh, we would brainstorm. We would come up with the f- the five sort of things. Obvious things. Yeah, yeah. The five obvious things. Yeah. And we would rule all those out. Yes. [laughs] So GPT would be the obvious things, and you rule those out? Yeah. Then yeah, yeah, yeah. Exactly. So it pushes creativity. All of a sudden every- Yeah ... individual screenwriter can do that without having to have a whole writers' room to brainstorm. Yeah. You just plug that in, it gives it back to you in two seconds, and you're just like- Yeah ... okay, not those things. Yeah, yeah. I'm gonna do something else, and now I am more creative than I was before. Yeah. Right? Well, your comment about music's really interesting 'cause, uh, now we've got Spotify, so we got everything in your pocket. Could imagine, like, the AI Spotify, which is, like, the doctor, the personal trainer, the educator. Like, all those different things in my pocket available right now for whatever I need to do. Yeah, that's right. Yeah. Yeah. And with a human escalation path, right? Yeah. It's like, you know- Yeah ... the, the, the AI, the AI therapist, uh, or whatever. Yeah. Yeah. But, you know, with, with the thing of, like, well, okay. Yeah. Yeah. I, I already have a thing here. Especially if it gets really serious, to escalate immediately. Yeah, that's right. Yeah, yeah. Okay, so what's, what's gonna hold us back? What, what do we need to change? So

12. 38:20 – 43:08

    Fear, Prometheus, and Regulating Math

    1. MAMarc Andreessen38:20

       I think it's mostly fear. So this is where maybe I'm- Yeah ... I'm, I'm a radical on it. Yeah. 'Cause, you know, this is usually where people start talking about, like, regulation. Um- Yeah ... I, I, I think it's f- like, we have these, we have these fear-driven reactions. I, I always think of there's this deep-seated myth in, in human society. It's the, the Prometheus myth, right? Yeah, yeah. And the Prometheus, Prometheus myth is all about new technology, right? Right, right. And the, 'cause the Prometheus myth is, like, basically this new technology of fire. Right. A- and, you know, and fire is one of these classic technologies where- Yeah ... like, it can be used for good. It can, or it can burn the village down. It can keep you alive, right? Or it can be used very badly. [laughs] Yeah. Right? And, yeah, it can destroy your, your, your whole world. Um, and so, uh, you know, Prometheus famously, you know, goes and retrieves, um, you know, fire from the gods. Yep. And his punishment for it is to be chained to a rock- Yeah ... and have his liver pecked out every day- Yes, yes ... for the rest of eternity, right? Yes, yes. So embedded in there is, like, the anxiety about the new technology, and then the arrival of the new technology maybe is like, you know, the, the fear, right, is that it's net bad- Yes ... and that the person who does that should be punished. Yes. And, and so I always find that myth kind of plays out over and over again in all these discussions- Yes ... about regulation. Yes. That this stuff, you know. Especially it's the gods who punish him, right? The existing gods. Yes, yes. Well, on behalf- Yes, yes, yes ... on behalf of, on behalf of, uh- Yeah, yeah ... on behalf of existence. Yeah, yeah. But, um, yeah, so, um, I, yeah, I, I think generally it's this. It's, it's just you get, you get these fears. If you look at the history of, uh, we, we talked about some of this, if you look at the history of new technologies- Yeah, yeah ... you generally have had these fears- Yes ... uh, every step along the way. Yes. Every new technology's been greeted with some prediction that it's gonna upend the social order and cause the, you know- Yeah ... enormous chaos. Well, it does upend to some degree, right? I mean- It, it will do that. Yeah. But, you know, generally speaking- In a positive way ... on balance- Yeah ... you know, tech- tech- I mean, technology is why we live much better lives today than we did- Certainly people f- now would not want what people had 50 years ago. Nobody would make that- Yes, right ... and you could do, you could go back in time infinitum- Yes ... and it'd be like that, right? Nobody would ever make the trade. Yes. Yes, yes. Nobody would ever make the trade to go back in time. Yeah. It would never happen. Yeah. And it, and, and, yeah, right, that's literally, it's because you would not wanna- Yeah ... you would not wanna lose the technologies- Yeah ... and the consequences- Yes, yes ... of the technologies we have today. So I think that's true. And so I actually think, like, fear may be the, to rip off FDR, fear, fear may be the, the actual biggest threat. Yeah. You know, fear leads to the kind of, uh, you know, reach for regulation. Yes. I, I'm a skeptic. I, I don't... It's like, I don't know, reg- regulating math. [laughs] Like, are we really gonna regulate math? Well, but it's not gonna look like regulating math, right? Uh... It's gonna look like regulating this superpower. Well, that's what they're gonna say. Yeah. Right? Right. But then the actual implementation of it- Yes ... is regul- you know, regulating- Yes. Regulating- Matrix algebra. Yeah. Yeah, regulating algebra. Yeah. Regul- regul- regulating linear algebra. Yeah, yeah. Like, are we really gonna regulate linear algebra- Yeah ... matrix multiplication? Yeah. Like, really? Seriously? Yeah. Um, and, and, and then even if we do, are we gonna possibly do it- You know ... in a, in a way that makes any sense? Sense. Yeah. Well, okay, but it won't, obviously it won't look like that. It'll be saying, "Well, we can't have computers drive cars." Right? Or, or, like, what, what, what's, what's... Well, how do you give the computer a test? Yeah. Or how do you know? Like, okay, you make this... I'll be the cynic. So okay, you make this claim that the computer AI is better than human. Yeah. Like, how do I know that? Yeah, 'cause that... Well, as it turns out, because the cars are driving. Yeah. So- Yeah. [laughs] So there, it's okay. Yeah. So here was inter- Yeah ... okay, so here's how that played out in self-driving cars- Yeah, yeah ... which is interesting. There was one category company that said, "We're gonna basically wait until it's perfect- Yes, yes ... and we're gonna basically try to validate- Yes ... and we're gonna work with the regulators and do this stuff." Yes, yes. There's this other- They're not driving. Yeah. And they're not on the road. Yeah. They're still not on the road. Yeah. Um, there's another category company that said, "You know what? Let's evolve out of basically the cruise control." Yes, yes. Um, and, you know, security cruise control, and then it's radar-based cruise control. A- and you get humans driving with it, and you label data and- Exactly, and you don't expect the car to drive itself in, in from the very beginning. The c- the car is, like, an autopilot kind of thing. Yeah. The expectation is that you pay attention. Like, you know, Te- Te- Tesla's the company I'm, I'm alluding to. Yeah. And if you, if you turn on full self, self-driving on Tesla, you're still, you know, you're still told, like, you're not supposed to be watching a movie. Yeah. You're supposed to be actually paying attention, and the car will, like, alert you when it's time to pay attention. Yeah. Um, but you know, notwithstanding that, Tesla has been climbing the ladder on self-driving car functionality capability. Yeah. They do new software releases pushed live to car at night anytime they want. Yes. Those new releases are not being tested by any federal- Yeah, yeah, yeah ... the federal, you know, NHTSA, whatever. Yeah, yeah, yeah. It's not, we're doing these things. Yeah. There's no actual test happening. Yeah, yeah. Um, and that has, that, and that has led to incredible progress- Oh, yeah ... including- Yeah ... as, as we said, clearly in the data, this is now safer than human drivers. Yeah, because you can't, you can't make it work just magically, right? [laughs] Right, right. It has to do, happen gradually. Right, 'cause it's, it's- Yeah ... you know, it, it actually much like medicine is. Yeah. You're centering into a complex system- Yeah, yeah ... with a lot of variables- Yeah ... in the real world. Also, like medicine too, it's like life or death. Yeah. You know, it's just serious. But- Yeah ... but it, it, yeah. Yeah. The, the, and we can go back to how we started the conversation. Yeah, yeah. The, the wait for permission thing- Yeah, yeah ... the, the binary zero or one, wait for permission- Yeah ... wait for perfection thing versus the incremental- Yes ... let's get better and better and better- Yes, yes ... and the threshold is, is it better than humans? Yeah, yeah. Is it, is it an, is it a net improvement? Yes. I mean, clearly in self-driving cars, that second approach is the approach that's working. Yeah. If you just, you know, observe, observe the real world. And do you think you get to the tipping point where look, let's look at the statistics we have because we, we have all this happening right now. We have the statistics, and it's, like I suspect we're gonna get there in medicine pretty quick.

    2. VPVijay Pande43:06

       Yeah, yeah, yeah, yeah, yeah.

    3. MAMarc Andreessen43:07

       Like,

13. 43:08 – 46:20

    AI in Medicine Without Waiting for Permission

    1. MAMarc Andreessen43:08

       I, I would say I'm, I'm an optimist-

    2. VPVijay Pande43:08

       Yeah

    3. MAMarc Andreessen43:08

       ... on that. And again-

    4. VPVijay Pande43:09

       Yeah, yeah

    5. MAMarc Andreessen43:09

       ... I'm not, I'm not an optimist 'cause I think the AI's gonna be perfect.

    6. VPVijay Pande43:11

       Yeah.

    7. MAMarc Andreessen43:11

       I'm an optimist 'cause I think the status quo is not that great.

    8. VPVijay Pande43:13

       Yeah, yeah. Well, and that might be, like you start empowering doctors, you give them tools, they start using them.

    9. MAMarc Andreessen43:17

       Yeah.

    10. VPVijay Pande43:17

        And you start empowering patients.

    11. MAMarc Andreessen43:19

        Yeah.

    12. VPVijay Pande43:19

        Patients start using them.

    13. MAMarc Andreessen43:20

        Yeah.

    14. VPVijay Pande43:20

        And actually, here, I think it's even different than a car because you're not on a road.

    15. MAMarc Andreessen43:23

        Right.

    16. VPVijay Pande43:24

        It's, and it's your body or whatever, and actually patients are driving their own healthcare more than ever. I think COVID was another sort of tailwind there. So maybe you start, maybe it's just about developing the tools and giving them out.

    17. MAMarc Andreessen43:34

        Right. Well, here would be an example-

    18. VPVijay Pande43:35

        Yeah.

    19. MAMarc Andreessen43:35

        So let's use our screenwriting example-

    20. VPVijay Pande43:36

        Yeah, yeah

    21. MAMarc Andreessen43:37

        ... as applied to medicine, which is, you know, a given set of conditions. There, there may be many possible diagnoses.

    22. VPVijay Pande43:41

        Yeah, yeah.

    23. MAMarc Andreessen43:41

        Uh, an experience I've had-

    24. VPVijay Pande43:43

        Yeah

    25. MAMarc Andreessen43:43

        ... is there's a set of, a set of symptoms.

    26. VPVijay Pande43:45

        Yes.

    27. MAMarc Andreessen43:45

        One doctor comes up with one diagnosis.

    28. VPVijay Pande43:46

        Yes.

    29. MAMarc Andreessen43:46

        Another doctor comes up with a different diagnosis.

    30. VPVijay Pande43:48

        Yes.
14. 46:20 – 50:26

    Moral Panic, Trolley Problems, and Fait Accompli Adoption

    1. MAMarc Andreessen46:20

       look, the, the, the big thing that goes wr- I mean, look, the big thing, t- I think two things go... So one is just that, the expectation of perfection.

    2. VPVijay Pande46:25

       Yeah, yeah.

    3. MAMarc Andreessen46:26

       Right? Uh, k- that, that... And, and look, it's very, you know, it's, it's very easy to generate the negative headline.

    4. VPVijay Pande46:31

       Yes.

    5. MAMarc Andreessen46:31

       It's very easy to set off the scare-

    6. VPVijay Pande46:32

       Yes, yes

    7. MAMarc Andreessen46:32

       ... the, the moral panic basically.

    8. VPVijay Pande46:34

       Yeah, yeah.

    9. MAMarc Andreessen46:34

       Right? With a single instance goes wrong and it gets extrapolated. Um, you know, we talk a lot about thalidomide, like you know-

    10. VPVijay Pande46:39

        Yeah

    11. MAMarc Andreessen46:39

        ... it'd be very easy to have that kind of moment.

    12. VPVijay Pande46:40

        Or, or like the, the, the person on a bike that got hit by a Tesla or something like that.

    13. MAMarc Andreessen46:45

        Right, yeah, exactly.

    14. VPVijay Pande46:45

        That I think he was biking across a freeway.

    15. MAMarc Andreessen46:48

        Right, right. [laughs]

    16. VPVijay Pande46:48

        Right? [laughs]

    17. MAMarc Andreessen46:49

        Exactly.

    18. VPVijay Pande46:49

        And so like a human probably hit him too.

    19. MAMarc Andreessen46:52

        Yeah, that's right.

    20. VPVijay Pande46:53

        Yeah.

    21. MAMarc Andreessen46:53

        Oh, well, that's a good point. Yeah.

    22. VPVijay Pande46:54

        Yeah, yeah, yeah.

    23. MAMarc Andreessen46:55

        So the, the trolley problem. Um-

    24. VPVijay Pande46:56

        Yeah, yeah

    25. MAMarc Andreessen46:56

        ... you know, the trolley problem's been in the press a little more recently-

    26. VPVijay Pande46:58

        Yeah, yeah

    27. MAMarc Andreessen46:58

        ... 'cause it turns out that, uh, Sam Bankman-Fried was an expert in the trolley problem.

    28. VPVijay Pande47:01

        Oh, really? Okay.

    29. MAMarc Andreessen47:02

        So that shows you that maybe that's not the route to ultimate-

    30. VPVijay Pande47:03

        [laughs]
15. 50:26 – 55:23

    Insurance Letters, Do Not Pay, and Bots Leveling the Field

    1. MAMarc Andreessen50:26

       Now, now look, y- having said that, healthcare is very sophisticated, right?

    2. VPVijay Pande50:29

       Yes.

    3. MAMarc Andreessen50:29

       There's lots of regulations.

    4. VPVijay Pande50:31

       Yes.

    5. MAMarc Andreessen50:31

       There's lots of payment.

    6. VPVijay Pande50:31

       Yes.

    7. MAMarc Andreessen50:31

       Right? All these things. So I saw this thing on Twitter the other day-

    8. VPVijay Pande50:34

       Yeah, yeah

    9. MAMarc Andreessen50:34

       ... that blew my mind, right?

    10. VPVijay Pande50:35

        Yeah.

    11. MAMarc Andreessen50:35

        'Cause this whole time I've been thinking in terms of like, you know, diagnosis-

    12. VPVijay Pande50:37

        Yeah, yeah

    13. MAMarc Andreessen50:38

        ... all this stuff in my life. So this doctor posted a video, um, and, um-

    14. VPVijay Pande50:42

        I think I saw that one. [laughs]

    15. MAMarc Andreessen50:43

        You probably saw it. So the doctor posts a video, and he said, "Look," he said, "The problem it is," from whatever, diagnosis or whatever, like, "I do the diagnosis, I do the prescription. Then it's a question of whether or not I can get the, uh, the insurance company to-

    16. VPVijay Pande50:52

        Yeah, yeah

    17. MAMarc Andreessen50:52

        ... to reimburse-

    18. VPVijay Pande50:53

        Yeah, yeah

    19. MAMarc Andreessen50:53

        ... to actually pay for the thing."

    20. VPVijay Pande50:54

        Yeah.

    21. MAMarc Andreessen50:55

        To do that for anything even slightly out of the ordinary, I have to write a let- I, the doctor-

    22. VPVijay Pande50:58

        Yeah

    23. MAMarc Andreessen50:58

        ... has to write a letter to the insurance company.

    24. VPVijay Pande51:00

        Yes.

    25. MAMarc Andreessen51:00

        And that letter needs to be in a specific format, and it needs to outline the case-

    26. VPVijay Pande51:02

        Has to make the case, right?

    27. MAMarc Andreessen51:03

        ... make the case, and it needs to have the scientific citations.

    28. VPVijay Pande51:05

        Yeah.

    29. MAMarc Andreessen51:06

        And if I do the letter really well, it's gonna get paid for.

    30. VPVijay Pande51:08

        Yes.

Episode duration: 55:23