Latest / a16z AI Policy Brief / Marc Andreessen on Betting on America

Transcript

Marc Andreessen on Betting on America · a16z AI Policy Brief

0:00
SPEAKER_02: We could have a revolution in education.
0:01
SPEAKER_02: We could have far better education at far lower cost.
0:03
SPEAKER_02: We could have a revolution in healthcare.
0:05
SPEAKER_02: There's all kinds of things that are possible now that weren't possible before.
0:08
SPEAKER_02: We could be in a world here within a decade where robots are building all the houses
0:11
SPEAKER_02: at far cheaper prices than today.
0:12
SPEAKER_02: Technology is a lever that could cause all those things to happen.
0:15
SPEAKER_02: It is really remarkable that China has decided that open source AI is something
0:19
SPEAKER_02: that is good and that they want to exist and that they want to propagate.
0:21
SPEAKER_02: We're in a weird state of the world where
0:23
SPEAKER_02: the supposedly totalitarian regime is trying to open up the technology.
0:27
SPEAKER_02: AI Policy Brief a16z and the supposedly democratic governance system is trying
0:30
SPEAKER_02: to restrict and control the technology.
0:32
SPEAKER_02: We live in this bifurcated economy where we've decided that some sectors are going
0:36
SPEAKER_02: to be subject to technological change and price declines and productivity growth
0:39
SPEAKER_02: and some sectors are not.
0:40
SPEAKER_02: As the prices for the blue sectors collapse,
0:43
SPEAKER_02: deflation,
0:44
SPEAKER_02: and as the prices for the red sectors inflate dramatically,
0:47
SPEAKER_02: what happens mathematically, right,
0:48
SPEAKER_02: is that the red sectors eat the entire economy,
0:50
SPEAKER_02: which is what's happening, right,
0:51
SPEAKER_02: which is healthcare,
0:52
SPEAKER_02: education, housing, law, government are eating the entire economy.
0:55
SPEAKER_01: Artificial intelligence is often described as a technology story.
0:59
SPEAKER_01: Mark Andreessen sees it as something bigger.
1:02
SPEAKER_01: In this conversation with CSIS's Naveen Girishankar,
1:05
SPEAKER_01: Mark argues that AI has the potential to expand access to intelligence itself,
1:10
SPEAKER_01: putting world-class expertise into the hands of billions of people.
1:14
SPEAKER_01: But realizing that potential will depend on more than just better models.
1:19
SPEAKER_01: The discussion explores productivity growth, infrastructure, regulation, industrial policy,
1:25
SPEAKER_01: U.S.-China Competition and the question of
1:27
SPEAKER_01: whether America's institutions can adapt quickly enough to take advantage of one of
1:31
SPEAKER_01: the most important technological shifts in history.
1:37
SPEAKER_00: Exponential growth is seductive,
1:38
SPEAKER_00: starting slowly and virtually unnoticeably,
1:41
SPEAKER_00: but beyond the knee of the curve it turns explosive and profoundly transformative.
1:46
SPEAKER_00: Those are the words of futurist and author Ray Kurzweil.
1:49
SPEAKER_00: He argues that two world wars,
1:52
SPEAKER_00: the Cold War,
1:52
SPEAKER_00: and every major economic upheaval of the last century failed to make
1:56
SPEAKER_00: the slightest dent in the pace of technological progress The disruptions are real,
2:02
SPEAKER_00: but the curve inevitably wins out That's the accelerationist thesis Now,
2:08
SPEAKER_00: even if we were to accept that society will always yield to technological progress,
2:12
SPEAKER_00: that is a prediction, not a policy
2:15
SPEAKER_00: Predictions,
2:16
SPEAKER_00: however accurate,
2:17
SPEAKER_00: on the trend tell us nothing about the transition itself,
2:20
SPEAKER_00: who wins and loses,
2:22
SPEAKER_00: whether institutions can absorb the shock,
2:25
SPEAKER_00: and what government and the private sector must each do to ensure that
2:28
SPEAKER_00: the gains are broad and the losses are survivable That is the question
2:35
SPEAKER_00: before us today
2:36
SPEAKER_00: I'm Naveen Girishankar and today I'm in conversation with Mark Andreessen,
2:41
SPEAKER_00: co-founder and general partner of Andreessen Horowitz and a member of
2:44
SPEAKER_00: the President's Council of Advisors on Science and Technology, PCAST.
3:06
SPEAKER_00: Welcome to Betting on America Mark Andreessen,
3:09
SPEAKER_00: what a privilege to have you on Betting on America.
3:12
SPEAKER_00: Thank you for doing this.
3:14
SPEAKER_02: Good morning.
3:14
SPEAKER_02: It's great to be here.
3:15
SPEAKER_02: Great to be with you.
3:16
SPEAKER_00: You've been an innovator,
3:18
SPEAKER_00: a technologist,
3:19
SPEAKER_00: an investor,
3:20
SPEAKER_00: and importantly,
3:21
SPEAKER_00: which everybody knows,
3:22
SPEAKER_00: but importantly, such a huge contributor to the public debate on AI and technology.
3:28
SPEAKER_00: And so we wanted to make sure we had the opportunity to speak with you.
3:32
SPEAKER_00: There are a lot of questions around policy that are very pertinent.
3:36
SPEAKER_00: You speak eloquently about
3:39
SPEAKER_00: modern alchemy turning sand into thought.
3:41
SPEAKER_00: I love that metaphor.
3:44
SPEAKER_00: And you talk about the AI boom and that it's actually not quite here yet.
3:48
SPEAKER_00: It's coming.
3:50
SPEAKER_00: So help us, give us your picture on what this looks like when the boom actually arrives.
3:56
SPEAKER_00: What will life look like?
3:59
SPEAKER_02: Yeah, so I mean, so...
4:02
SPEAKER_02: AI Policy Brief a16z
4:21
SPEAKER_02: is everywhere.
4:22
SPEAKER_02: Every field is transformed.
4:23
SPEAKER_02: AI solves every problem.
4:25
SPEAKER_02: There's kind of that view.
4:26
SPEAKER_02: And then,
4:27
SPEAKER_02: of course,
4:27
SPEAKER_02: it's also very easy to get the dystopian view of doom and death and destruction.
4:33
SPEAKER_02: And many people are out selling books based on that idea.
4:37
SPEAKER_02: I think maybe I'd tend to have a little bit more of a nuanced view,
4:41
SPEAKER_02: which is we have the potential for something resembling the utopian view,
4:45
SPEAKER_02: but we have a set of policy choices that are between us and that.
4:51
SPEAKER_02: and many of these policy choices are choices that have been made over
4:54
SPEAKER_02: the preceding 80 years in terms of really sharply restricting the ability
5:00
SPEAKER_02: for technology to actually affect the economy and day-to-day life in many ways.
5:05
SPEAKER_02: AI does not make any of those go away.
5:10
SPEAKER_02: And in fact, it may well be a catalyst for more of those.
5:14
SPEAKER_02: And so I would put myself, I call myself an optimist, not a utopian.
5:19
SPEAKER_02: And then some days when I take a look at what's happening,
5:22
SPEAKER_02: it feels like healthcare,
5:23
SPEAKER_02: education, housing, law, and others, I maybe even become a little bit pessimistic.
5:28
SPEAKER_02: And so anyway, I think this is an actual complex nuanced conversation that needs to happen.
5:33
SPEAKER_02: And I think we'll probably touch on a bunch of that today.
5:35
SPEAKER_00: Yeah, let's take a couple.
5:37
SPEAKER_00: One thing you've said,
5:38
SPEAKER_00: which I find resonates strongly is that AI is going
5:42
SPEAKER_00: to be your new brilliant genius friend,
5:45
SPEAKER_00: whether it's a private tutor for your kids or whether it's your financial advisor.
5:50
SPEAKER_00: AI Policy Brief a16z
6:05
SPEAKER_00: Let's start there because I think it's an important question.
6:21
SPEAKER_02: Yeah, you probably know there's actually been some research studies in this so far
6:24
SPEAKER_02: that kind of,
6:25
SPEAKER_02: you know,
6:25
SPEAKER_02: frame the question consistent with what you just said,
6:28
SPEAKER_02: which is basically,
6:29
SPEAKER_02: you know,
6:30
SPEAKER_02: there are many fields,
6:31
SPEAKER_02: you know,
6:31
SPEAKER_02: in which there's, you know, there's sort of superstars who are like hyperproductive.
6:34
SPEAKER_02: And then there's the sort of rank and file,
6:36
SPEAKER_02: you know, people who are kind of, you know, average levels of productivity.
6:38
SPEAKER_02: And so there's this question of, right, is AI an excel?
6:42
SPEAKER_02: Does AI basically cause the superstars to become,
6:44
SPEAKER_02: you know,
6:45
SPEAKER_02: a thousand X superstars and,
6:46
SPEAKER_02: you know, kind of cause the power law curve to spike way up?
6:49
SPEAKER_02: AI Policy Brief
7:14
SPEAKER_02: There will be a huge amount of focus,
7:15
SPEAKER_02: obviously, politically on the distributional effects, which are important.
7:20
SPEAKER_02: But I think that the dominant thing is I think everybody gets better.
7:25
SPEAKER_02: Having said that,
7:26
SPEAKER_02: the other side of that is the way I describe what you said is,
7:30
SPEAKER_02: yeah, you now have the world's best doctor in your pocket.
7:32
SPEAKER_02: You have the world's best lawyer in your pocket.
7:34
SPEAKER_02: You have the world's best accountant in your pocket.
7:35
SPEAKER_02: You have the world's best teacher in your pocket.
7:37
SPEAKER_02: But that's immediately where, again, I run up against the kind of
7:40
SPEAKER_02: AI actually can't be your lawyer because it can't get admitted to the bar.
7:48
SPEAKER_02: It can't be your doctor because it can't,
7:49
SPEAKER_02: you know, be admitted to the, you know, it can't actually be a doctor.
7:52
SPEAKER_02: It can't, you know, by the way, for example, it can't submit for reimbursement on
7:55
SPEAKER_02: insurance right um right which is a key function of doctors
7:58
SPEAKER_02: and hospitals today um it can't be your cpa like it it can't get licensed as cpa by
8:03
SPEAKER_02: the way it can't be your teacher because you know
8:06
SPEAKER_02: as you know like you know k-12 teachers are
8:08
SPEAKER_02: a you know government sponsored monopoly um
8:10
SPEAKER_02: and uh you know you can't you can't get it yeah i can't get a credit as a teacher so
8:14
SPEAKER_02: so we're going to be in this world in which the software is going to be much better
8:18
SPEAKER_02: than almost anybody you deal with in any of those professions and
8:20
SPEAKER_02: yet those professions as far as i can tell are going to stay completely untouched
8:24
SPEAKER_00: Yeah, you know,
8:25
SPEAKER_00: it's another way of saying if intelligence becomes less of the binding constraint on
8:31
SPEAKER_00: the margin, then what becomes the binding constraint?
8:34
SPEAKER_00: Is it the way we interact with each other?
8:35
SPEAKER_00: Is it our values?
8:37
SPEAKER_00: Is it how our institutions function?
8:41
SPEAKER_00: It shines a light on our weaknesses in that realm, right?
8:44
SPEAKER_00: Yeah.
8:45
SPEAKER_02: Yeah, that's right, right, right, exactly.
8:47
SPEAKER_02: If you, yeah, if you remove variables, then you max out the impact of the remaining variables.
8:51
SPEAKER_02: So, you know, that's 100% correct.
8:53
SPEAKER_02: And so, yeah, so, I mean, like,
8:54
SPEAKER_02: you know,
8:54
SPEAKER_02: you know,
8:55
SPEAKER_02: there's,
8:56
SPEAKER_02: as you well know, there,
8:56
SPEAKER_02: you know,
8:57
SPEAKER_02: there's already, like,
8:57
SPEAKER_02: extensive politics around things like, you know, K-12, you know, teacher union,
9:01
SPEAKER_02: as an example like you know if everybody in the world has the world's best teacher
9:05
SPEAKER_02: in their pocket then you know all of a sudden the entire point of being a teacher in
9:08
SPEAKER_02: the K-12 system is going to be the government protection of your job which by
9:13
SPEAKER_02: the way is the direction that that field has been going in for 50 years anyway and
9:17
SPEAKER_02: so it'll just blow it out all the way right so K-12 teachers become
9:21
SPEAKER_02: a purely political function which of course you know to
9:24
SPEAKER_02: a large extent they already are
9:25
SPEAKER_00: Purely political or maybe they teach something else, right?
9:30
SPEAKER_00: No, not at all.
9:30
SPEAKER_02: No, no, no.
9:31
SPEAKER_02: They won't change at all.
9:32
SPEAKER_02: They don't have to.
9:34
SPEAKER_02: They're completely protected.
9:35
SPEAKER_00: Right, you're making a political economy point, fully appreciate the point.
9:39
SPEAKER_00: I'm just saying that ideally,
9:41
SPEAKER_00: if more and more of that function is taken over by AI or supported by AI,
9:46
SPEAKER_00: then what, if anything, do teachers do?
9:50
SPEAKER_00: There's an opportunity for them to do other things, right?
9:53
SPEAKER_00: Teach other things,
9:54
SPEAKER_00: like perhaps more interpersonal skills or values or I don't know what it is,
9:59
SPEAKER_00: but it's not the thing that the AI is doing, right?
10:02
SPEAKER_02: Yeah, so look,
10:03
SPEAKER_02: if we didn't have government,
10:04
SPEAKER_02: let's hypothesize the world where we don't have the government protections
10:07
SPEAKER_02: and controls, right?
10:07
SPEAKER_02: Right, right.
10:08
SPEAKER_02: Somehow K-12 is a free market system like everything else,
10:11
SPEAKER_02: or like people want to imagine it could be.
10:14
SPEAKER_02: So you may know there is a school that is doing what you described.
10:17
SPEAKER_02: There's a private school system called Alpha School.
10:20
SPEAKER_00: I've heard of it, yeah, yeah.
10:23
SPEAKER_02: Yeah, so it's a case study for what you're describing.
10:25
SPEAKER_02: So it's a private school.
10:26
SPEAKER_02: So it's outside the, you know, the public, outside the public system.
10:28
SPEAKER_02: It's a, you know, completely paid, you know, it's a cash pay thing, you know, with parents.
10:32
SPEAKER_02: It's obviously, you know, it's expensive.
10:33
SPEAKER_02: So, you know, most,
10:34
SPEAKER_02: you know,
10:34
SPEAKER_02: it's out of reach,
10:35
SPEAKER_02: obviously, most kids, most parents, but it is a model of what you're saying.
10:38
SPEAKER_02: and I'll just describe it for a moment.
10:40
SPEAKER_02: So the guy who built Alpha Schools,
10:42
SPEAKER_02: this guy,
10:42
SPEAKER_02: Joe Lehmont,
10:42
SPEAKER_02: who's one of the,
10:43
SPEAKER_02: actually like a real software legend,
10:45
SPEAKER_02: you know, in the technology field, you know, from the 90s, a really brilliant guy.
10:50
SPEAKER_02: And he spent the last, I don't know, 15 years or something.
10:52
SPEAKER_02: And I think he's put like a billion
10:53
SPEAKER_02: AI Policy Brief a16z
11:09
SPEAKER_02: You know, just like an existing school, but the day is very different.
11:13
SPEAKER_02: So there's two hours in the morning of actual academic instruction, which is run by AI.
11:19
SPEAKER_02: And so it's AI mediated, you know, sort of computer based instruction.
11:23
SPEAKER_02: The point of that being that the AI is already,
11:26
SPEAKER_02: you know, a better teacher than most human teachers.
11:28
SPEAKER_02: And then specifically, the AI could be in a one to one relationship with each student.
11:31
SPEAKER_02: And so each student...
11:32
SPEAKER_02: AI Policy Brief a16z
11:48
SPEAKER_02: of the day the teachers are primary but that's not academic instruction
11:51
SPEAKER_02: the way you're used to think about in the classroom it's all project based work
11:54
SPEAKER_02: and activity based work and so it's students coming up together to you know whatever
11:58
SPEAKER_02: to have a you know community garden and learn how to take care of plants or
12:01
SPEAKER_02: to you know learn how to start a small business right or learn how
12:04
SPEAKER_02: to do you know whatever it is you know
12:05
SPEAKER_02: that you are you know have you know long projects on you know like Model UN
12:11
SPEAKER_02: or whatever the version of that is today that you know people do for like learning
12:13
SPEAKER_02: about government
12:15
SPEAKER_02: And so, to your point,
12:16
SPEAKER_02: the teachers are hands-on with the kids working on all of these kinds of things that
12:19
SPEAKER_02: in a normal classroom you never get to.
12:21
SPEAKER_02: Now, the challenge is Alpha School is a private system.
12:24
SPEAKER_02: Of course,
12:24
SPEAKER_02: the existing U.S.
12:25
SPEAKER_02: educational system is going to do everything possible to marginalize or destroy it.
12:29
SPEAKER_02: The existing government K-12 system will not do any of what I just described.
12:33
SPEAKER_02: And fundamentally, very little will change.
12:37
SPEAKER_00: But what's really interesting is
12:39
SPEAKER_00: the story you're telling is how technology is providing the motivation and
12:44
SPEAKER_00: the driver for institutional change,
12:47
SPEAKER_00: or at least the impetus for institutional change.
12:49
SPEAKER_02: I appreciate your optimistic framing.
12:51
SPEAKER_02: My observation of institutions is they don't want to change.
12:54
SPEAKER_02: They have no intention of changing.
12:55
SPEAKER_00: I agree.
12:56
SPEAKER_00: And it's a hard thing to make them change, for sure.
13:00
SPEAKER_00: But the opportunity is provided by technology to do something that has not been done before.
13:06
SPEAKER_02: Oh yeah, 100%.
13:07
SPEAKER_02: Look, we could have a revolution education.
13:09
SPEAKER_02: We could have far better education at far lower cost.
13:11
SPEAKER_02: We could have a revolution of healthcare.
13:13
SPEAKER_02: There's all kinds of things that are possible now that weren't possible before.
13:17
SPEAKER_02: By the way, housing construction.
13:18
SPEAKER_02: I mean,
13:19
SPEAKER_02: we could be in a world here within a decade where robots are building all the houses
13:22
SPEAKER_02: at far cheaper prices than today.
13:24
SPEAKER_02: And then self-driving cars open up entire areas of geography in the country.
13:28
SPEAKER_02: for how, you know, for housing, you know, much better housing at much lower cost.
13:31
SPEAKER_02: Yeah, you could, I mean,
13:32
SPEAKER_02: but the revolution of government services,
13:34
SPEAKER_02: you can imagine the government,
13:35
SPEAKER_02: you know,
13:36
SPEAKER_02: literally,
13:36
SPEAKER_02: you know,
13:36
SPEAKER_02: becoming,
13:37
SPEAKER_02: you know,
13:37
SPEAKER_02: state of the art,
13:37
SPEAKER_02: you know,
13:38
SPEAKER_02: if you look at what the Nashville Design Studio,
13:40
SPEAKER_02: for example,
13:40
SPEAKER_02: is doing right now in federal government,
13:41
SPEAKER_02: trying to,
13:42
SPEAKER_02: you know,
13:42
SPEAKER_02: make government services as compelling and easy to use as,
13:46
SPEAKER_02: you know, private sector consumer offerings.
13:49
SPEAKER_02: Yeah, like, yes,
13:51
SPEAKER_02: the sort of modern alchemy of AI is a,
13:54
SPEAKER_02: technology is a lever that could cause all those things to happen.
13:57
SPEAKER_02: I just, just observing the behavior of all these,
14:01
SPEAKER_02: of every single institution that I just referenced,
14:04
SPEAKER_02: They all seem 100% opposed to that I fully appreciate it In fact,
14:08
SPEAKER_00: I want to come back to the question of public sector reform
14:12
SPEAKER_00: through this conversation But I want to just put out the notion that it's not just
14:17
SPEAKER_00: the age of AI It could be the age for institutional reformers And it's something
14:23
SPEAKER_00: for us to consider But I want to come back to regulatory constraints in
14:27
SPEAKER_00: a second One more question for you
14:29
SPEAKER_00: How do you see the productivity boom playing out?
14:32
SPEAKER_00: Because I've heard you talk about it,
14:34
SPEAKER_00: and give me a second here,
14:35
SPEAKER_00: I've heard you talk about that upward sloping curve to the right in terms
14:40
SPEAKER_00: of productivity enhancements in the aggregate.
14:43
SPEAKER_00: And again, that's...
14:44
SPEAKER_00: That resonates strongly Let's see how it plays out through different sectors
14:49
SPEAKER_00: But that's the macro story What about the mezzo and micro story Because
14:54
SPEAKER_00: while it's upward sloping to the right It could be pretty bumpy along the way
14:58
SPEAKER_00: And there could be winners and losers And I just wanted to get your thoughts on that
15:03
SPEAKER_02: The thing with the modern economy,
15:04
SPEAKER_02: the thing with the modern industrialized economy is the productivity growth or,
15:09
SPEAKER_02: by the way, productivity decline, it varies dramatically by sector.
15:13
SPEAKER_02: So there's no longer an economy-wide concept of productivity growth that makes any sense.
15:17
SPEAKER_02: You have to disaggregate by sector and what you find...
15:20
SPEAKER_02: in the charts basically the chart that I always use is sort of separates
15:23
SPEAKER_02: between those sort of red sectors and blue sectors.
15:26
SPEAKER_02: So the blue sectors are sectors in which there's very rapid productivity growth.
15:32
SPEAKER_02: There's very rapid price declines.
15:34
SPEAKER_02: And there's very rapid
15:35
SPEAKER_02: in technological innovation.
15:36
SPEAKER_02: And these are sectors you could say like television sets,
15:40
SPEAKER_02: consumer electronics television sets as an example,
15:42
SPEAKER_02: software, entertainment content, basically toys, by the way, fall in this category.
15:50
SPEAKER_02: AI Policy Brief a16z
16:05
SPEAKER_02: A16z
16:20
SPEAKER_02: I would put those basically as like five.
16:23
SPEAKER_02: The red sector is characterized by rapidly rising prices,
16:26
SPEAKER_02: rapidly rising prices,
16:27
SPEAKER_02: rapidly rising spend,
16:30
SPEAKER_02: zero or negative productivity growth, and almost no technological
16:35
SPEAKER_02: A16z
16:50
SPEAKER_02: is restrictions on supply.
16:52
SPEAKER_02: So those are sectors of the economy in which there are cartels,
16:55
SPEAKER_02: monopolies, oligopolies, licensing restrictions, inability to fundamentally compete.
17:00
SPEAKER_02: And then because of the spiraling upward prices, there's subsidization of demand.
17:06
SPEAKER_02: You see this with housing policy all the time now,
17:08
SPEAKER_02: which is like, well,
17:08
SPEAKER_02: it's too expensive to buy houses, so therefore we're gonna subsidize home buying.
17:12
SPEAKER_02: Well, if you subsidize a market in which you've restricted supply,
17:15
SPEAKER_02: you just cause prices to rise further.
17:16
SPEAKER_02: Right.
17:18
SPEAKER_02: Right, which is why those sectors have this upward spiral.
17:21
SPEAKER_02: And so we live in this bifurcated economy where we've decided
17:24
SPEAKER_02: that some sectors are going to be subject to technological change and price declines
17:27
SPEAKER_02: and productivity growth, and some sectors are not.
17:30
SPEAKER_02: And then mechanically what happens is
17:32
SPEAKER_02: AI Policy Brief a16z
17:55
SPEAKER_02: AI Policy Brief
18:11
SPEAKER_02: is if you go back a hundred years productivity growth was running two
18:14
SPEAKER_02: or even three times higher than it is today right and so so we think that we live in
18:18
SPEAKER_02: an era of rapid technological change there's endless books and magazine articles
18:22
SPEAKER_02: and news stories about how we live in an area
18:24
SPEAKER_02: of incredible technological change we think
18:26
SPEAKER_02: the computer revolution has been this like huge you know change we think
18:28
SPEAKER_02: the internet's been this huge change we think AI is going to be this huge change and
18:32
SPEAKER_02: if you look at the economic statistics the result is super low productivity growth
18:35
SPEAKER_02: and super low economic growth
18:38
SPEAKER_02: And so this is the problem is you can have the best technology in the world
18:42
SPEAKER_02: that could bend these curves.
18:43
SPEAKER_02: And if the policy setup in those industries prevents that from happening.
18:49
SPEAKER_02: But what's basically another way to think about it is it's just going to be trash.
18:52
SPEAKER_02: We're just going to take all of the monetary gains that we get from AI
18:55
SPEAKER_02: and we're just going to spend them all on healthcare and education.
18:58
SPEAKER_02: So interesting.
18:59
SPEAKER_02: And real estate, right?
19:01
SPEAKER_02: Like that's where all the money is going to go.
19:03
SPEAKER_02: AI Policy Brief a16z
19:28
SPEAKER_02: yeah like where's the where's the economic growth why didn't we get it and of course
19:31
SPEAKER_02: the answer is we didn't want it
19:31
SPEAKER_02: because we'd rather have you know we'd rather have health care education
19:34
SPEAKER_00: and housing work the way that they do today yeah and I think that
19:39
SPEAKER_00: so I now I understand your skepticism about institutional reform yes exactly
19:46
SPEAKER_00: And I'm going to come back to that again.
19:49
SPEAKER_00: So we're in the early innings of this.
19:53
SPEAKER_00: There are many different potential constraints.
19:57
SPEAKER_00: And I think you're pointing to the fact they're ultimately policy and regulatory.
20:01
SPEAKER_00: but there is an infrastructure AI infrastructure build out that's happening
20:06
SPEAKER_00: that has some constraints whether it's
20:07
SPEAKER_00: on energy labor others permitting I should say and then there are these constraints
20:16
SPEAKER_00: on particular sectors even as AI you seek to flow AI through those sectors what are
20:21
SPEAKER_00: the big big constraints like the top two or three we should be thinking about
20:26
SPEAKER_00: when it comes to policy and regs
20:29
SPEAKER_02: Yeah, well,
20:29
SPEAKER_02: so look,
20:30
SPEAKER_02: so on the supply side,
20:31
SPEAKER_02: so on the supply side,
20:32
SPEAKER_02: you know,
20:33
SPEAKER_02: basically what the state of affairs right now is basically every single component
20:38
SPEAKER_02: that goes into the stack of infrastructure and technology.
20:41
SPEAKER_00: Yeah.
20:41
SPEAKER_02: And capabilities that are needed to field AI,
20:43
SPEAKER_02: like there's basically a bottleneck on every single,
20:46
SPEAKER_02: at every single layer of the supply chain.
20:48
SPEAKER_02: Right.
20:49
SPEAKER_02: And so, by the way,
20:50
SPEAKER_02: it starts at the very bottom with energy,
20:53
SPEAKER_02: you know, where there's bottleneck on energy production.
20:56
SPEAKER_02: for reasons that you well understand.
20:58
SPEAKER_02: Then there's bottleneck on literally physical facilities, physical plants, right?
21:03
SPEAKER_02: So the big data center controversy, right?
21:06
SPEAKER_02: And the whole thing on that.
21:10
SPEAKER_02: By the way,
21:10
SPEAKER_02: there's constraints on all the physical infrastructure going
21:12
SPEAKER_02: to building data centers.
21:13
SPEAKER_02: For example, turbines are sold out, I think, for four years.
21:16
SPEAKER_02: like you can't buy turbines you can't buy transformers I know
21:20
SPEAKER_02: of one hyperscaler that's actually milling its own turbine blades to try
21:23
SPEAKER_02: to get new turbines into for power generation cooling systems are sold out you know
21:31
SPEAKER_02: the big HVAC systems that you need big water cooling systems
21:35
SPEAKER_02: And then inside the data center,
21:37
SPEAKER_02: you know,
21:39
SPEAKER_02: NVIDIA, you know, the GPUs and the chips that go in are in very tight constraint.
21:45
SPEAKER_02: Memory chips, you know, the price of memory chips are exploding right now.
21:50
SPEAKER_02: AI Policy Brief
22:05
SPEAKER_02: There's physical constraints actually at every layer.
22:08
SPEAKER_02: And that's important for several reasons.
22:12
SPEAKER_02: One is that it actually means that the AI products and services that you have access
22:17
SPEAKER_02: to today as a consumer or as a business are actually not as capable as they could be
22:21
SPEAKER_02: if the supply chain was more liberated.
22:24
SPEAKER_02: So you're actually getting dumber versions of the AI today than you could get
22:27
SPEAKER_02: if ships were more plentiful.
22:30
SPEAKER_02: because they're constrained.
22:31
SPEAKER_02: They literally,
22:32
SPEAKER_02: these companies don't have enough chips and power and data center space to be able
22:34
SPEAKER_02: to train more advanced models.
22:38
SPEAKER_02: And so you're getting worse versions of the products and then this is going to hit pricing.
22:43
SPEAKER_02: And they're...
22:44
SPEAKER_02: We've been in this world for the last five years where the price per token
22:47
SPEAKER_02: of intelligence has been hyper-deflating because the algorithm has been getting
22:50
SPEAKER_02: so much better.
22:52
SPEAKER_02: But that is rapidly running up against these physical constraints of being unable
22:55
SPEAKER_02: to build new data centers.
22:57
SPEAKER_02: And so I think the price declines in intelligence are going to stop.
23:00
SPEAKER_02: And in fact, it may be that actually intelligence is going to start getting more expensive.
23:06
SPEAKER_00: Such a great insight, such a great insight.
23:08
SPEAKER_00: But let me ask you something there because I could think on that full list
23:12
SPEAKER_00: of problems you identified,
23:14
SPEAKER_00: there are several things that could be done either at the federal level or
23:17
SPEAKER_00: the state level,
23:18
SPEAKER_00: whether it's permitting,
23:19
SPEAKER_00: whether it's, you know, constraints on energy, so on and so forth, maybe even labor.
23:24
SPEAKER_00: But here's one that's sticky.
23:27
SPEAKER_00: and we're looking,
23:27
SPEAKER_00: we look at this often in our institution here at CSIS is the tariff agenda because
23:32
SPEAKER_00: the tariff agenda cuts against some of what we need to do on
23:36
SPEAKER_00: the data center build-out, doesn't it?
23:38
SPEAKER_02: Yeah, so tariffs, I mean, look,
23:40
SPEAKER_02: there's sort of the,
23:41
SPEAKER_02: you know,
23:41
SPEAKER_02: there's sort of the,
23:42
SPEAKER_02: you know,
23:42
SPEAKER_02: I don't know,
23:42
SPEAKER_02: whatever the classical kind of economic view of tariffs,
23:45
SPEAKER_02: you know,
23:45
SPEAKER_02: sort of a form of taxation and then,
23:47
SPEAKER_02: you know,
23:47
SPEAKER_02: and then you get into the,
23:48
SPEAKER_02: you know,
23:48
SPEAKER_02: question of,
23:49
SPEAKER_02: you know,
23:49
SPEAKER_02: re-industrialization question because,
23:51
SPEAKER_02: you know, just as an example, one of the things we haven't touched on yet is Taiwan.
23:54
SPEAKER_02: Right.
23:55
SPEAKER_02: Right, because...
23:58
SPEAKER_02: I'm not laughing because it's funny because it's very serious and it's kind
24:02
SPEAKER_02: of amazing how serious it is which is we are completely dependent on Taiwanese tabs
24:07
SPEAKER_02: for the chips right now to a degree that I think is actually bad for Taiwan
24:13
SPEAKER_02: because the fact that Taiwan is so central for the making
24:16
SPEAKER_02: of advanced AI just makes them an even bigger prize where
24:19
SPEAKER_02: the Chinese government decided to move.
24:21
SPEAKER_02: So I think Taiwan sort of amazingly is like,
24:24
SPEAKER_02: Taiwan's almost like too important for its own good right now.
24:26
SPEAKER_02: Right.
24:28
SPEAKER_02: And then there's all the strategic kind of aspects,
24:30
SPEAKER_02: which is if the Chinese do ultimately move to Taiwan,
24:32
SPEAKER_02: it's like, okay, are we going to be able to get chips?
24:34
SPEAKER_02: Are we going to be able to build anything?
24:36
SPEAKER_02: And so there is this need to reindustrialize for several reasons,
24:42
SPEAKER_02: not least of which is natural security.
24:43
SPEAKER_02: And so then you get the industrial policy debate.
24:48
SPEAKER_02: But I will tell you that the other thing about the tariff thing,
24:51
SPEAKER_02: which I find fairly amazing in the discussion is,
24:54
SPEAKER_02: you know, a tariff, it's really funny.
24:56
SPEAKER_02: A tariff is, of course, it's a tax on international, you know, financial transactions trade.
25:01
SPEAKER_02: And
25:02
SPEAKER_02: AI Policy Brief a16z
25:27
SPEAKER_00: No, it's both.
25:29
SPEAKER_00: I'm just saying that if we're trying to solve the problem you're talking about,
25:33
SPEAKER_00: which is the data centers,
25:35
SPEAKER_00: the physical infrastructure is now going to become a constraint on AI.
25:41
SPEAKER_00: Don't you want to remove all the obstacles to it?
25:44
SPEAKER_00: That means the permitting stuff, but also it means tariffs, right?
25:47
SPEAKER_02: Yeah, but like 99% of the practical restrictions,
25:50
SPEAKER_02: the practical restrictions and constraints are not the tariffs.
25:53
SPEAKER_02: 99% are on the things we do to ourselves inside our own country.
25:55
SPEAKER_00: Fair point, behind the border.
25:58
SPEAKER_02: Yeah, so I would just, I would, yes.
26:00
SPEAKER_02: So I would just, every...
26:04
SPEAKER_02: What I'm reacting to is not you.
26:05
SPEAKER_02: I'm reacting to,
26:06
SPEAKER_02: you know, five,
26:06
SPEAKER_02: you know,
26:07
SPEAKER_02: four years right now of sort of this kind of,
26:09
SPEAKER_02: you know,
26:09
SPEAKER_02: hysterical kind of frenzy in the press and among the pundit class on
26:12
SPEAKER_02: the tariff topic from people who think it's a great idea
26:14
SPEAKER_02: to have all equivalent taxes and restrictions on internal trade.
26:18
SPEAKER_02: So it's like the only thing that people get upset about in the public discussion
26:21
SPEAKER_02: on this is trade with foreigners.
26:22
SPEAKER_02: Like trade domestically is far more constrained and controlled.
26:25
SPEAKER_00: Yeah, and we should be upset about both.
26:27
SPEAKER_00: I'm just saying like there are two dimensions to it.
26:30
SPEAKER_00: There's the cost dimension and
26:32
SPEAKER_00: And then there's the volatility,
26:33
SPEAKER_00: the erratic nature in which these things have been implemented,
26:37
SPEAKER_00: none of which is good for, I mean, you tell me, is it good for investors?
26:41
SPEAKER_02: You know, I'm just saying 99% of the issues are internal.
26:43
SPEAKER_02: Yeah.
26:43
SPEAKER_02: Fair point.
26:44
SPEAKER_02: It's almost internal.
26:54
SPEAKER_02: what's happening literally in the US right now county by county with the ability
26:57
SPEAKER_02: to build data centers is like profoundly destructive no true yeah
27:01
SPEAKER_02: and that's like that's entirely domestic and a large number
27:05
SPEAKER_02: of politicians are like feeding that hysteria as much as they possibly can
27:09
SPEAKER_02: and a lot of our leading public figures and a lot of intellectuals and a lot of
27:12
SPEAKER_02: the press and a lot of the analysts and the rest of it is just like this kind
27:15
SPEAKER_02: of hyper paranoia about building data centers and the consequence of data.
27:19
SPEAKER_02: I'll just give you an example,
27:20
SPEAKER_02: this completely fake meme about water use,
27:23
SPEAKER_02: which is just like factually not true,
27:25
SPEAKER_02: which is just like running wild through the public discussion
27:27
SPEAKER_02: that somehow these data centers are like basically destroying all the water,
27:30
SPEAKER_02: which is like this completely insane idea.
27:31
SPEAKER_02: Like that factor is like so much a bigger factor holding us back
27:35
SPEAKER_02: than anything involving internal trade.
27:38
SPEAKER_02: so it's like external trade is the thing that's easy to talk about it's all
27:42
SPEAKER_02: of our internal issues that are like much much more important that's
27:46
SPEAKER_00: a very fair point let's just talk about models for a second so
27:54
SPEAKER_00: The Mythos case and the export controls that were put on it.
27:57
SPEAKER_00: It's really interesting because whether or not the Commerce Department has
28:03
SPEAKER_00: the legal authority to do it is a separate question.
28:06
SPEAKER_00: I look at this and I pose it as a question for you.
28:08
SPEAKER_00: AI Policy Brief a16z
28:34
SPEAKER_00: how do you assess that whole thing because that's like as far
28:37
SPEAKER_00: as models are concerned separate from leading-edge chips that's
28:41
SPEAKER_00: an important question that we would have to answer as well no
28:45
SPEAKER_02: Yeah, so I think there's a whole bunch of,
28:46
SPEAKER_02: you know, very complicated topics, a whole bunch of factors.
28:50
SPEAKER_02: I would start with a very high-level kind of view on this, though,
28:52
SPEAKER_02: which is we have,
28:56
SPEAKER_02: as is often the case with anything,
28:57
SPEAKER_02: you know,
28:57
SPEAKER_02: complicated in the real world,
28:58
SPEAKER_02: you know,
28:58
SPEAKER_02: there are multiple contradictory goals,
29:00
SPEAKER_02: you know,
29:00
SPEAKER_02: that we would like to be able to solve them all at the same time,
29:03
SPEAKER_02: but it's hard because they conflict.
29:05
SPEAKER_02: And so let's just start with the U.S.
29:07
SPEAKER_02: versus China part because I think that drives a lot of this.
29:09
SPEAKER_00: Yeah, okay.
29:10
SPEAKER_02: Because I think if China didn't exist, I think we'd be having a different discussion.
29:13
SPEAKER_02: We'd be having a different and simpler discussion because it would just be about us.
29:18
SPEAKER_02: To start with, AI right now is a two-horse race.
29:20
SPEAKER_02: It's US and China.
29:21
SPEAKER_02: Effectively, there's no other player.
29:23
SPEAKER_02: By the way, there could be other players specifically in Europe.
29:27
SPEAKER_02: They've decided to make everything illegal.
29:29
SPEAKER_02: So they've suicidally taken themselves out of the race,
29:32
SPEAKER_02: which is a whole other thing we could talk about.
29:34
SPEAKER_02: They've taken every bad idea that we have and kind of maxed it out to 11%.
29:39
SPEAKER_02: and so they're maybe becoming a case study of what not to do,
29:42
SPEAKER_02: but be that as it may, it's basically a two-horse race now.
29:44
SPEAKER_02: It's U.S. versus China.
29:47
SPEAKER_02: So to start with, we have two contradictory goals, one of which is we want to...
29:53
SPEAKER_02: We want to make sure that the US wins the global technology race.
29:58
SPEAKER_02: So we want to make sure that when we wake up in a decade,
30:01
SPEAKER_02: the world is running on American AI and not on Chinese AI.
30:06
SPEAKER_02: And in fact, ideally, what we would like to do is live in a world
30:08
SPEAKER_02: AI Policy Brief a16z
30:37
SPEAKER_02: To do that, what do we have to do?
30:39
SPEAKER_02: We have to export.
30:41
SPEAKER_01: We have to take our technology and we have to make it available to the world.
30:45
SPEAKER_02: We have another goal,
30:46
SPEAKER_02: which is,
30:47
SPEAKER_02: as far as I can tell,
30:48
SPEAKER_02: just as important, which is AI is an extremely disruptive new technology.
30:52
SPEAKER_02: It has profound, not just economic implications, but also national security implications.
30:57
SPEAKER_02: Also, by the way, competitiveness implications.
31:00
SPEAKER_02: And with that goal,
31:02
SPEAKER_02: we need to control and restrict and constrain and maybe even hoard AI to ourselves.
31:10
SPEAKER_02: We need to make sure that the AI is this magic technology that only we have,
31:14
SPEAKER_02: and we need to make sure other people don't get it.
31:16
SPEAKER_02: And we have to absolutely make sure that China doesn't get it.
31:20
SPEAKER_02: and this goes straight to topics like for example chip export controls which
31:24
SPEAKER_02: of course were in place even before the mythos issue
31:29
SPEAKER_02: but right away there you can see these are directly contradictory goals
31:32
SPEAKER_02: and I think what
31:34
SPEAKER_02: And I think what you have in
31:35
SPEAKER_02: the US government is I think you have extremely well-meaning people who have
31:41
SPEAKER_02: the country's best interests apart,
31:42
SPEAKER_02: some of whom have the first goal as a primary goal,
31:45
SPEAKER_02: some of whom have the second goal as a primary goal,
31:47
SPEAKER_02: and those goals are exactly contradictory with each other.
31:53
SPEAKER_02: And so I think that's actually the underlying kind of logical question that you have to have.
32:01
SPEAKER_02: Then the other example that directly on your Mythos point,
32:04
SPEAKER_02: I would say is another example of sort of diametrically opposed goals,
32:08
SPEAKER_02: which is now you have a level of capability with this technology,
32:12
SPEAKER_02: starting with the current models and the next set of models like Mythos,
32:16
SPEAKER_02: where they are better than human at both attacking
32:21
SPEAKER_02: Cyber Systems.
32:22
SPEAKER_02: And they are better at defending cyber systems than human beings are.
32:28
SPEAKER_02: And so you have these models, they're a threat of disruption.
32:32
SPEAKER_02: And of course,
32:33
SPEAKER_02: this is where the current US government is very worried about disruption of
32:35
SPEAKER_02: the financial system,
32:36
SPEAKER_02: mythos models being used by bad guys,
32:38
SPEAKER_02: criminals or terrorists or foreign governments to,
32:41
SPEAKER_02: for example,
32:42
SPEAKER_02: break into and really wreck US banks or US stock market or whatever,
32:45
SPEAKER_02: which is like, I think a very legitimate concern.
32:47
SPEAKER_00: Yeah.
32:48
SPEAKER_02: But you also have this diametrically opposed thing where the same tool that's good
32:52
SPEAKER_02: at penetrating is also very good at defending.
32:54
SPEAKER_02: Right.
32:54
SPEAKER_02: And so the other thing you need to do is you need to get those tools in the hands
32:57
SPEAKER_02: of every existing company and business everywhere in the West,
33:00
SPEAKER_02: everywhere in the US,
33:01
SPEAKER_02: and you need to fix all the security holes in all the systems and have new kinds
33:04
SPEAKER_02: of AI cyber defenses and everything.
33:05
SPEAKER_00: Yeah.
33:06
SPEAKER_02: but again here you can see this thing where these are directly contradictory because
33:09
SPEAKER_02: the more scared you are of it legitimately scared you are of it worried about it
33:14
SPEAKER_02: the more you want to restrict it but the more you want to actually use it as
33:17
SPEAKER_02: a prophylactic to make sure that all of our banks for example aren't subject
33:21
SPEAKER_02: to cyber attack the more you want to deploy it
33:24
SPEAKER_02: Yeah.
33:25
SPEAKER_02: And so anyway,
33:26
SPEAKER_02: so I just,
33:27
SPEAKER_02: you know,
33:27
SPEAKER_02: a lot of people when they engage on these issues,
33:29
SPEAKER_02: it's sort of, you know, they question people's motives.
33:32
SPEAKER_02: I think in this case,
33:33
SPEAKER_02: you've got in both cases,
33:34
SPEAKER_02: you've got these like directly contradictory motives and you have to go straight to
33:37
SPEAKER_02: the underlying conversation of like,
33:39
SPEAKER_02: which is actually the most important goal before you can figure out what
33:42
SPEAKER_02: the right policies are.
33:44
SPEAKER_00: And by the way,
33:45
SPEAKER_00: you described two groups of people,
33:47
SPEAKER_00: one that holds the kind of innovation goal,
33:50
SPEAKER_00: the other one that holds the safety goal.
33:53
SPEAKER_00: I would say like a lot of times the same person is trying
33:56
SPEAKER_00: to balance those two objectives in government.
33:59
SPEAKER_00: having served in government I recognize many people like struggle and wrestle with
34:02
SPEAKER_00: that so I mean I'm going to ask you because in a way you're now on
34:07
SPEAKER_00: the President's Council of Advisors for Science
34:10
SPEAKER_00: and Technology what would you advise them to do
34:12
SPEAKER_00: because how do you weight these goals because I can imagine at any given point
34:17
SPEAKER_00: in time one becomes more important than the other and you described
34:21
SPEAKER_00: the like really challenging situation that we're in
34:25
SPEAKER_02: Yeah, so my view,
34:27
SPEAKER_02: my normal view on these things is basically it's sometimes called
34:29
SPEAKER_02: the technological imperative,
34:30
SPEAKER_02: which is basically this idea like you don't uninvent new technologies, right?
34:35
SPEAKER_02: Like once a new technology exists, it exists and it's going to make its way into the world.
34:40
SPEAKER_02: Like it is going to have a way of making its way out.
34:43
SPEAKER_02: And, you know,
34:43
SPEAKER_02: there may be physical constraints or whatever that prevent it
34:45
SPEAKER_02: from being fully realized everywhere.
34:47
SPEAKER_02: But like fundamentally...
34:48
SPEAKER_02: You know,
34:48
SPEAKER_02: I don't know,
34:49
SPEAKER_02: once the process for making steel, like,
34:50
SPEAKER_02: was a known thing,
34:52
SPEAKER_02: you know,
34:52
SPEAKER_02: it was inevitable that all military equipment was going to get made out of steel,
34:54
SPEAKER_02: like,
34:55
SPEAKER_02: and by the way,
34:56
SPEAKER_02: and so was all civilian equipment going to get made out of steel,
34:58
SPEAKER_02: and that was going to happen.
34:59
SPEAKER_02: And the same thing, same,
35:00
SPEAKER_02: same thing for steam power,
35:01
SPEAKER_02: and the same thing for electricity,
35:02
SPEAKER_02: and the same thing, right,
35:03
SPEAKER_02: the same thing for,
35:04
SPEAKER_02: you know,
35:05
SPEAKER_02: was it the Haberbach process,
35:06
SPEAKER_02: and you just go right down the list of all these innovations,
35:09
SPEAKER_02: and, you know, the computer chip,
35:11
SPEAKER_02: and they were going to happen.
35:13
SPEAKER_02: They may happen faster or slower, but they're going to happen.
35:16
SPEAKER_02: And so if you're going to live in that future world,
35:18
SPEAKER_02: in my view,
35:18
SPEAKER_02: you want to be as strong and powerful and dominant as you can possibly be
35:22
SPEAKER_02: when those things do happen, right?
35:24
SPEAKER_02: You want to win.
35:25
SPEAKER_02: And to me,
35:25
SPEAKER_02: victory,
35:26
SPEAKER_02: like if I were,
35:27
SPEAKER_02: you know, king for a day, victory would be, like I said, you would set a vision.
35:32
SPEAKER_02: You would say,
35:32
SPEAKER_02: we're going to have a world in which the entire world is going to run
35:34
SPEAKER_02: an American AI.
35:36
SPEAKER_02: and American AI is going to be so good and proliferated so broadly and going to be
35:40
SPEAKER_02: so universal in the world that even China's not at some point they're just going
35:42
SPEAKER_02: to say this isn't even worth competing with us like this is a complete waste of time
35:46
SPEAKER_02: and so I would come out very strongly on the side of you kind
35:48
SPEAKER_02: of want you want maximum export right you want
35:51
SPEAKER_02: to just like basically turbocharge exports you want
35:53
SPEAKER_02: the US government you know working hand in hand with with the companies
35:56
SPEAKER_02: to figure out optimal policies to make sure that AI American AI at
36:00
SPEAKER_02: the software level chip level and so forth you know basically proliferates and runs
36:03
SPEAKER_02: the entire world yeah
36:05
SPEAKER_02: Now, having said that,
36:06
SPEAKER_02: I think the people who are arguing,
36:08
SPEAKER_02: for example,
36:08
SPEAKER_02: for chip export controls against that are doing so in completely good faith.
36:12
SPEAKER_02: And I think they have very reasonable arguments for what they're doing.
36:17
SPEAKER_02: But I would go in that direction.
36:18
SPEAKER_02: And then on things like Mythos,
36:21
SPEAKER_02: the direction I would go is I would say, look,
36:23
SPEAKER_02: the whole reason why we're worried about like cyber exploitation of systems is,
36:28
SPEAKER_02: so this is actually very important.
36:29
SPEAKER_01: So
36:30
SPEAKER_02: AI hacking does not create new security vulnerabilities
36:33
SPEAKER_02: that don't already exist AI hacking exploits existing security vulnerabilities
36:37
SPEAKER_02: that already exist and those security vulnerabilities are subject
36:39
SPEAKER_02: to being exploited both by AI but also by non-AI hackers and of course banks
36:44
SPEAKER_02: and all these other government agencies are getting hacked all the time even
36:47
SPEAKER_02: without AI
36:49
SPEAKER_02: AI hacking is going to be much more effective.
36:51
SPEAKER_02: And so I think you need to get the defenses.
36:54
SPEAKER_02: We need to focus on the defenses.
36:55
SPEAKER_02: We need to get the defenses in place.
36:56
SPEAKER_02: And the way to get the defenses in place is we need to use these new advanced AI models.
36:59
SPEAKER_02: We need to put them in the hands of all companies as fast as possible to
37:03
SPEAKER_02: and be able to basically armor up and have AI defenses against AI hacking
37:07
SPEAKER_02: and non AI hacking, right?
37:09
SPEAKER_02: Like for example, this is also how you solve the ransomware crisis, right?
37:12
SPEAKER_02: Which is you need to go fix all the systems in the hospital so
37:15
SPEAKER_02: that they can't be held hostage by ransomware.
37:18
SPEAKER_02: AI Policy Brief a16z
37:41
SPEAKER_02: You know, I guess I would say this.
37:45
SPEAKER_02: The winds are going against me on both of those topics.
37:48
SPEAKER_02: And so it feels like I'm not going to be king for a day.
37:51
SPEAKER_02: And so it feels like we're going to be living in as one in which probably
37:57
SPEAKER_02: the opposite arguments are going to prevail.
37:59
SPEAKER_00: Yeah.
38:00
SPEAKER_00: If I might just offer a couple of reflection on that.
38:02
SPEAKER_00: That was a great rundown.
38:03
SPEAKER_00: A couple of reflections on that.
38:05
SPEAKER_00: One is that when it comes to chips, for example, the
38:10
SPEAKER_00: Fully accept your point that over time diffusion is going to happen, you can't prevent it.
38:15
SPEAKER_00: But whether you can change the timetable is an open question, especially when it comes to chips.
38:20
SPEAKER_00: And that timetable can be critical depending on where you are in your competition
38:25
SPEAKER_00: with China, for example.
38:27
SPEAKER_00: Would you agree with that point?
38:29
SPEAKER_02: So I think that's true,
38:30
SPEAKER_02: but also I think something else is true,
38:32
SPEAKER_02: which is if you deny them chips, you incent them to create their own chips.
38:35
SPEAKER_00: And you see that happening already.
38:37
SPEAKER_00: Yeah, we do see that happening already, yeah.
38:40
SPEAKER_00: and they can start creating ecosystems that prevent us
38:44
SPEAKER_00: from entering them they can advance faster than us and then now we're not close to
38:48
SPEAKER_00: where innovation is happening all of those things are true um
38:52
SPEAKER_00: but at the same time taking that off the table is
38:54
SPEAKER_00: a real challenge like I find it very difficult to say we're going
38:59
SPEAKER_00: to unilaterally not use this instrument if we can use it I think the challenge is or
39:05
SPEAKER_00: the question is how do you use it where it really hits the mark rather than sort
39:09
SPEAKER_00: of taking a buckshot approach and using export controls all over the place
39:13
SPEAKER_00: for every problem which is really what I think sometimes we over-index on that
39:19
SPEAKER_02: Yeah, I mean,
39:19
SPEAKER_02: you know,
39:19
SPEAKER_02: you can try, you know, all the discussions around industrial policy, right?
39:22
SPEAKER_02: You can try.
39:24
SPEAKER_02: You know, I'll just give you my background here.
39:26
SPEAKER_02: So my first commercial product I ever built and took to market was the Netscape Browser in 1994.
39:35
SPEAKER_02: It was expert controlled.
39:37
SPEAKER_02: Yeah.
39:38
SPEAKER_02: It was classified by ITAR as ammunition.
39:40
SPEAKER_02: It was in the same classification category as a Tomahawk missile.
39:44
SPEAKER_01: it was explained to us by our lawyers in no uncertain terms
39:46
SPEAKER_01: that we could not possibly let this outside the US right and by the way
39:50
SPEAKER_02: when it started you'll enjoy this actually when it started encryption was such
39:53
SPEAKER_02: a sensitive topic of the 1990s that we were actually expert controlled not just
39:58
SPEAKER_02: on strong encryption but also on weak encryption
40:01
SPEAKER_02: we couldn't even ship web server software that had weak encryption it took years
40:09
SPEAKER_02: to get the government to basically come to grips with the idea that
40:11
SPEAKER_02: if we were not allowed to do that what was happening of course is what you'd expect
40:15
SPEAKER_02: A16z
40:36
SPEAKER_02: A16z
40:47
SPEAKER_02: And having said that,
40:48
SPEAKER_02: you know, again,
40:49
SPEAKER_02: back to the core argument is,
40:50
SPEAKER_02: do you really want to live in a world in which that means that U.S.
40:52
SPEAKER_02: technology loses?
40:54
SPEAKER_02: Because encryption was going to happen, right?
40:58
SPEAKER_02: I used to own a t-shirt.
40:59
SPEAKER_02: I probably still have it somewhere.
41:01
SPEAKER_02: I used to own a t-shirt.
41:02
SPEAKER_02: I remember the RSA algorithm was like the key encryption algorithm of that era.
41:07
SPEAKER_02: And there was an implementation of the RSA algorithm, which was just math.
41:10
SPEAKER_02: There was an implementation of it in four lines of code.
41:13
SPEAKER_02: it was a sort of very complicated hard to read code but there were four lines
41:17
SPEAKER_02: of code and I had a t-shirt that had the four lines of code on it and of course
41:20
SPEAKER_02: the joke which which wasn't a joke was
41:22
SPEAKER_02: that t-shirt was ammunition like right it was actually illegal like
41:28
SPEAKER_02: in theory I never by the way I never tested this but in theory if I had worn
41:31
SPEAKER_02: that t-shirt and board an international flight I could have been put
41:33
SPEAKER_02: in jail wow yeah so so so there's that what a great story yeah
41:39
SPEAKER_02: Yeah, yeah, yeah.
41:40
SPEAKER_02: It took years.
41:41
SPEAKER_02: It took years to work through that, right?
41:42
SPEAKER_02: And so it is kind of amazing.
41:44
SPEAKER_02: Okay, so then on AI, like, AI is math.
41:48
SPEAKER_02: Like, at the end of the day, it's math.
41:50
SPEAKER_02: Like, it's actually, by the way, it's actually remarkable.
41:53
SPEAKER_02: It's actually quite straightforward and simple math.
41:54
SPEAKER_02: It's basically linear algebra,
41:56
SPEAKER_02: and then it's a handful of algorithms with names like gradient descent,
41:58
SPEAKER_02: reinforcement learning.
41:59
SPEAKER_02: It's math.
42:00
SPEAKER_02: And you've probably been watching this.
42:02
SPEAKER_02: I
42:02
SPEAKER_02: I know your organization has been tracking this which is the version of the math
42:07
SPEAKER_02: that implements a model at whatever GPT 5.0 or 5.5 level or mythos level
42:13
SPEAKER_02: or whatever like that math looks hard and expensive for about six months and
42:19
SPEAKER_02: then somebody figures out a way to run it on a PC
42:22
SPEAKER_02: they figure out a way to shrink it down and basically run it on a piece
42:25
SPEAKER_02: of consumer hardware increasingly by the way these things just run
42:27
SPEAKER_02: on your cell phone and the lag time between the new version of the AI
42:32
SPEAKER_02: the new capability being something rare and special that you can control
42:35
SPEAKER_02: because you can control where the data centers get built to being something that is
42:37
SPEAKER_02: an open source
42:39
SPEAKER_02: by the way open source from the US open source from China or open source you know
42:42
SPEAKER_02: in theory from anywhere in
42:43
SPEAKER_02: the world you know any academic institution could do this now you create
42:47
SPEAKER_02: the open source version and then you have a version that can run on a PC or can run
42:51
SPEAKER_02: on a phone and so this goes back to like in theory you can calibrate who gets access
42:56
SPEAKER_02: to what and when and in theory you can kind of do this dance like
42:59
SPEAKER_02: in practice you do find yourself in both in Christian case and
43:02
SPEAKER_02: the AI case you find yourself trying to control the propagation of math which is an
43:08
SPEAKER_02: It's an extremely difficult thing.
43:09
SPEAKER_02: And then,
43:09
SPEAKER_02: by the way,
43:10
SPEAKER_02: there's another kind of dimension on this that I would put out there,
43:12
SPEAKER_02: which is if you really want to make sure the powerful AI doesn't proliferate,
43:16
SPEAKER_02: and if you talk to the people who are very worried about this,
43:18
SPEAKER_02: they will say this with complete seriousness.
43:20
SPEAKER_02: You have to start to watch what people do on all computer systems.
43:24
SPEAKER_02: You have to start to watch what happens on every chip.
43:27
SPEAKER_02: And so,
43:29
SPEAKER_02: the policy recommendations that ultimately flow out of this line
43:32
SPEAKER_02: of thought include things like putting a software agent on every chip
43:35
SPEAKER_02: on every computer everywhere in the world including all the computers
43:38
SPEAKER_02: in your house including your kid's laptop and you put an agent on that and
43:42
SPEAKER_02: that agent reports back to the government what that computer is being used for and
43:46
SPEAKER_02: if it's used to run AI that's too powerful then there needs to be some set
43:50
SPEAKER_02: of consequences to it and then of course the very next thing is that needs to be
43:54
SPEAKER_02: a global regime
43:55
SPEAKER_02: right and in fact you have to hear this from a lot of people in
43:57
SPEAKER_02: the industry they're like well we need
43:58
SPEAKER_02: a global governance regime it's like well what does that mean well it means like
44:03
SPEAKER_02: a UN you know like a UN with teeth that like controls global use of software yeah
44:09
SPEAKER_02: and you find yourself walking down this cut
44:10
SPEAKER_02: in my view my view you find yourself walking down this kind
44:13
SPEAKER_02: of 1984 Orwellian totalitarian you know playbook
44:17
SPEAKER_02: where Big Brother is watching what happens on everybody's
44:19
SPEAKER_02: on everybody's computers like all the time and then stepping in
44:21
SPEAKER_02: when you're running unapproved software and
44:23
SPEAKER_02: and so again it's just like in theory you can kind of play this game you can kind
44:28
SPEAKER_02: of do the dance I think in practice the sort of downstream effects get
44:33
SPEAKER_00: to be quite scary Such a great rundown of so many different issues
44:37
SPEAKER_00: and how they're connected I would just say that
44:41
SPEAKER_00: you make a very compelling case like focus on innovation
44:44
SPEAKER_00: and innovating faster that's really the only long-term way of staying ahead
44:49
SPEAKER_00: and remove the obstacles to doing that because trying to I'm trying to apply
44:53
SPEAKER_00: an export control on
44:54
SPEAKER_00: a model is like exceedingly difficult I don't know how you would implement that
44:58
SPEAKER_00: and enforce it effectively
45:02
SPEAKER_00: but the second thing I would say is just look at what is happening in the PRC
45:07
SPEAKER_00: which is a real commitment to diffusion and a real commitment to using AI
45:11
SPEAKER_00: in various realms of the economy and I wonder
45:14
SPEAKER_00: whether our challenge now especially coming on the heels of a you know on the run up
45:19
SPEAKER_00: to an election skepticism about AI and the fears about it are like
45:26
SPEAKER_00: the overwhelming thing and I think it might be getting in the way of our
45:30
SPEAKER_00: staying ahead in the tech race and getting all kinds of economic benefits from that.
45:36
SPEAKER_00: Do you agree with that?
45:39
SPEAKER_02: Yeah, I agree for sure.
45:40
SPEAKER_02: And by the way, I'd start just to stay on the geopolitics for a moment.
45:43
SPEAKER_02: It is really remarkable that China has decided that open source AI is something
45:46
SPEAKER_02: that is good and that they want to exist and that they want to propagate.
45:49
SPEAKER_02: Like we're in a weird state.
45:52
SPEAKER_02: We're in a weird state of the world where
45:54
SPEAKER_02: the supposedly totalitarian regime is trying to open up the technology and
45:59
SPEAKER_02: the supposedly democratic governance system is trying to restrict and control
46:03
SPEAKER_02: the technology.
46:05
SPEAKER_02: Like it's the opposite.
46:07
SPEAKER_02: We're in like opposite world from what you would think.
46:09
SPEAKER_00: But that might just be a reflection of where they are in the race, right?
46:13
SPEAKER_00: So they're also restricting critical minerals in a pretty coercive way.
46:19
SPEAKER_00: So I think that the minute if the balance,
46:22
SPEAKER_00: God forbid,
46:22
SPEAKER_00: the balance were to shift,
46:24
SPEAKER_00: then I can't imagine that they would be committed to open AI models just because of
46:30
SPEAKER_00: the goodness of their heart, right?
46:31
SPEAKER_00: So it might be just a reflection of where we are, no?
46:33
SPEAKER_00: No.
46:34
SPEAKER_02: Oh, I mean, I would take it a step further.
46:36
SPEAKER_02: I think it's a deliberate strategy.
46:38
SPEAKER_02: I really agree with what you just said, which is I think it's a deliberate strategy.
46:41
SPEAKER_02: I think the Chinese,
46:41
SPEAKER_02: and by the way,
46:42
SPEAKER_02: I think the US government believes this very specifically,
46:46
SPEAKER_02: which is that the Chinese are very deliberately,
46:48
SPEAKER_02: the Chinese CCP is very deliberately encouraging or mandating its companies
46:52
SPEAKER_02: to create AI open source and to advance it as fast as possible.
46:56
SPEAKER_01: Precisely to prevent the success of American industry.
47:01
SPEAKER_02: It's like a turbo dumping strategy.
47:05
SPEAKER_02: So flood the market with basically free AI to prevent the American companies
47:09
SPEAKER_02: from being able to make money on it.
47:11
SPEAKER_00: Yeah.
47:12
SPEAKER_02: So I totally agree with what you're saying.
47:14
SPEAKER_02: I just think it is fairly amazing,
47:16
SPEAKER_02: at least for now, that they're the proponents of free and open.
47:24
SPEAKER_00: Let me take 30 seconds.
47:25
SPEAKER_00: I got to ask you something,
47:26
SPEAKER_00: because a lot of the people who argue for the pro-innovation stance on national,
47:32
SPEAKER_00: let's just call it economic competitiveness and national security, right?
47:35
SPEAKER_00: The argument you've made.
47:37
SPEAKER_00: It resonates with me.
47:38
SPEAKER_00: But
47:39
SPEAKER_00: Invariably people who hold that position,
47:41
SPEAKER_00: when you ask them what about deep civil military fusion in China and the risks that
47:48
SPEAKER_00: a broader set of commercial technologies are dual use,
47:52
SPEAKER_00: I don't get a strong answer for them.
47:54
SPEAKER_00: Give me your strong answer to that because it's true.
47:57
SPEAKER_00: They have a policy of deep civil military fusion.
48:00
SPEAKER_00: My question is that your argument that the only way to get ahead of it is
48:04
SPEAKER_00: to out-innovate and have them use American AI Like we need to stay ahead But
48:10
SPEAKER_00: in doing that,
48:11
SPEAKER_00: this would be my counter-argument with others who say just open the doors
48:15
SPEAKER_00: and just let us trade Let us kind of export American technology to
48:20
SPEAKER_00: the Chinese You know,
48:22
SPEAKER_00: what about the civil-military fusion risk which is very real there?
48:27
SPEAKER_02: Meaning that,
48:29
SPEAKER_02: if I understand properly,
48:29
SPEAKER_02: you're saying that the Chinese take American AI,
48:32
SPEAKER_02: they use it to build better military weapons.
48:34
SPEAKER_00: Weapons, yes, because they have deep fusion across their commercial and their military sectors.
48:40
SPEAKER_00: That's their policy.
48:41
SPEAKER_02: Oh, yeah, yeah, for sure.
48:42
SPEAKER_02: Yeah, I mean, look, I think that's a real, I mean, 100%.
48:45
SPEAKER_02: I think that they absolutely would do that, by the way.
48:48
SPEAKER_02: I think they're likely doing that today.
48:49
SPEAKER_02: Right.
48:50
SPEAKER_00: Yeah.
48:51
SPEAKER_02: This is the other thing, which is, are we actually successfully embarking on chips?
48:57
SPEAKER_02: There's a lot of chips in the world.
48:59
SPEAKER_02: It's hard to control where they go.
49:01
SPEAKER_02: And by the way, here's a question.
49:06
SPEAKER_02: Do we think the Chinese already have Mythos?
49:10
SPEAKER_02: All Mythos is, it's a set of numbers on a hard drive.
49:14
SPEAKER_02: It's a file.
49:15
SPEAKER_00: Yes.
49:16
SPEAKER_02: Like how incompetent is the MSS if they haven't already figured out a way
49:19
SPEAKER_01: to download that file?
49:20
SPEAKER_02: And by the way, any data center that runs an AI model has a copy of that file.
49:24
SPEAKER_02: Like that is how the systems work.
49:26
SPEAKER_02: It's a giant matrix of numbers.
49:29
SPEAKER_02: And so I would say to start with what you're describing is probably already happening.
49:34
SPEAKER_02: A, B, because we have to question whether the controls can actually hold.
49:41
SPEAKER_02: Another way to put it is there are no American AI companies
49:44
SPEAKER_02: that have anything resembling counterintelligence or any security control system
49:48
SPEAKER_02: that anybody with a government background would possibly find
49:51
SPEAKER_02: to be even remotely acceptable.
49:54
SPEAKER_02: They all employ large numbers of Chinese nationals.
49:57
SPEAKER_02: They all employ large numbers of, frankly, Chinese Americans with relatives in mainland China.
50:02
SPEAKER_02: who are subject to exploitation.
50:07
SPEAKER_02: They run open and collaborative R&D environments.
50:09
SPEAKER_02: They don't have internal stow piping.
50:11
SPEAKER_02: They don't have classification.
50:13
SPEAKER_02: They don't have counterintelligence.
50:15
SPEAKER_02: By the way,
50:15
SPEAKER_02: it's actually illegal for American AI companies to not employ Chinese engineers
50:18
SPEAKER_02: under civil rights law.
50:20
SPEAKER_02: so even if you try to control for
50:22
SPEAKER_02: that you can't like it's not allowed I mean SpaceX got prosecuted by
50:27
SPEAKER_02: the previous administration's Justice Department for not hiring enough refugees as
50:31
SPEAKER_02: a federal military contractor that's only allowed to in theory that's only allowed
50:34
SPEAKER_02: to have US citizens work on its systems so
50:38
SPEAKER_02: Anyway, so first of all,
50:38
SPEAKER_02: it's likely the Chinese have everything that we're describing anyway, A.
50:43
SPEAKER_02: And then B, yeah, 100%.
50:45
SPEAKER_02: Like, yeah,
50:45
SPEAKER_02: if they get free and unalloyed access to everything,
50:49
SPEAKER_02: then yeah, they're going to use it.
50:50
SPEAKER_02: But again,
50:50
SPEAKER_02: you're back to the question of trade-off,
50:52
SPEAKER_02: which is, okay,
50:53
SPEAKER_02: if they're not using the American technology to do that,
50:55
SPEAKER_02: then they're building domestic technology to do it.
50:57
SPEAKER_02: And then do you really want to live in the world in
50:59
SPEAKER_02: which their domestic technology is their military technology?
51:01
SPEAKER_02: Yeah.
51:02
SPEAKER_02: Wouldn't it be better from a national security standpoint if
51:05
SPEAKER_02: the US government always knew that they could go talk
51:08
SPEAKER_02: to any American technology company for anything happening anywhere in the world,
51:13
SPEAKER_02: as opposed to having black box companies in mainland China that they have no access
51:16
SPEAKER_02: to and no way into?
51:18
SPEAKER_02: Yeah.
51:19
SPEAKER_02: But again,
51:19
SPEAKER_02: I would say, look,
51:21
SPEAKER_02: I think it's a completely legitimate question and observation
51:23
SPEAKER_02: because I think there are real trade-offs.
51:25
SPEAKER_02: And if American AI wins all over the world,
51:27
SPEAKER_02: then yeah, American AI will be the basis of everybody's military systems.
51:31
SPEAKER_02: And yes, that could lead to faster advances in enemy military systems.
51:34
SPEAKER_02: And I think that's a completely real question.
51:36
SPEAKER_00: Yeah.
51:36
SPEAKER_00: Yeah.
51:37
SPEAKER_00: I mean I think this is the moment we're in right like we the the economic policies
51:42
SPEAKER_00: the national security policies of the last 75 years 80 years really are not crafted
51:49
SPEAKER_00: for the moment that we're in and this raises a question
51:53
SPEAKER_00: for me you know I think we need a significant public sector reform effort I put out
51:58
SPEAKER_00: a piece in foreign affairs saying America needs economic warriors and it was kind of
52:02
SPEAKER_00: a
52:04
SPEAKER_00: The title was the title but the main argument was like we need to retool government
52:07
SPEAKER_00: to do the things that it has to do in this moment and
52:11
SPEAKER_00: that means not only efficiencies but also new capabilities
52:14
SPEAKER_00: that we do not currently have The administration deserves some credit for doing that
52:18
SPEAKER_00: with TechForce and other things like that but we're far from that and I just wanted
52:23
SPEAKER_00: to get your thoughts I think you were pretty optimistic about what Doge could do
52:28
SPEAKER_00: and some of these other things but where do you think we are now?
52:32
SPEAKER_02: Yeah, so I think there's a bunch of people like in this administration
52:35
SPEAKER_02: who are trying very hard.
52:36
SPEAKER_02: And I just mentioned earlier,
52:38
SPEAKER_02: the National Design Studio,
52:39
SPEAKER_02: Joe Gabbia,
52:40
SPEAKER_02: who's one of the great Silicon Valley founders,
52:41
SPEAKER_02: co founder of Airbnb,
52:42
SPEAKER_02: you know, who's literally in the White House trying to do what you're describing.
52:45
SPEAKER_02: I think he and his team are doing great work.
52:46
SPEAKER_02: By the way, a lot of the
52:47
SPEAKER_02: doge people.
52:48
SPEAKER_02: A lot of the really sharp table doge people are still in government
52:51
SPEAKER_02: and I think having pretty big impact.
52:54
SPEAKER_02: And so, you know, I think there are examples of that.
52:57
SPEAKER_02: You know, having said that, again, the main issue is not sort of, you know, what's possible.
53:02
SPEAKER_02: The main issue is do these institutions want to be reformed?
53:05
SPEAKER_02: And what are the levels of the antibodies that come out,
53:07
SPEAKER_02: you know, whenever there's any suggestion of reform?
53:10
SPEAKER_02: And, you know, and as you know, the antibodies are extremely strong and vigorous in
53:16
SPEAKER_00: I totally agree with that but there are better and worse ways
53:19
SPEAKER_00: of doing reform don't you think like I mean I think that some people would argue
53:23
SPEAKER_00: that the doge effort has left some bureaus like Swiss cheese and the holes are not
53:28
SPEAKER_00: where you need those holes to be right you know I would love
53:32
SPEAKER_02: to like you I would love to see other approaches to reform that work yeah
53:38
SPEAKER_00: And I think that's the moment that we're in, Mark.
53:40
SPEAKER_00: Like, I feel that we need American institutional reformers like par excellence
53:45
SPEAKER_00: who know how to do this,
53:47
SPEAKER_00: who can face the interests that are going to resist against it,
53:52
SPEAKER_00: but also are imaginative in terms of thinking about the capabilities
53:55
SPEAKER_00: that government needs in this era of AI, which we're far from thinking.
54:01
SPEAKER_00: We're not there yet, right?
54:03
SPEAKER_02: And I'll also add,
54:05
SPEAKER_02: and I say this hopefully on an optimistic note,
54:07
SPEAKER_02: this doesn't have to be a partisan issue, what you're saying.
54:10
SPEAKER_02: And in fact,
54:12
SPEAKER_02: you may remember,
54:12
SPEAKER_02: you know,
54:13
SPEAKER_02: there was actually the Clinton-Gore administration in the 1990s had a big effort
54:16
SPEAKER_00: in this direction.
54:17
SPEAKER_00: Yes.
54:18
SPEAKER_00: Called Reinventing Government.
54:19
SPEAKER_00: Reinventing Government.
54:20
SPEAKER_00: Yeah.
54:21
SPEAKER_02: and Al Gore in particular put a lot of time and effort into it
54:24
SPEAKER_02: and got some ways down the field.
54:26
SPEAKER_02: And so like, yes, one can certainly imagine what you're describing.
54:31
SPEAKER_02: I think you're 100% right that we need it and it would be great.
54:37
SPEAKER_02: Having said that,
54:37
SPEAKER_02: I would just say the people who have tried to do it with whatever method have a lot
54:42
SPEAKER_02: of scar tissue.
54:44
SPEAKER_02: And so it's, yes, it's a, yeah, I was going to say this, it's never been harder.
54:48
SPEAKER_00: Yeah, I hear that.
54:50
SPEAKER_00: I want to ask you, there's a lot of talk about what policy do we need for AI.
54:57
SPEAKER_00: What about AI for public policy?
54:59
SPEAKER_00: What's your view on that?
55:01
SPEAKER_02: How so?
55:02
SPEAKER_00: Well, I just think that there are a lot of policies that different politicians promote.
55:10
SPEAKER_00: but we don't know if they're effective or not and I wonder if we have
55:13
SPEAKER_00: an opportunity now to really accelerate evaluation in real time of what's working
55:19
SPEAKER_00: and what's not working so it improves the quality of the debate
55:22
SPEAKER_00: on what policies we should undertake whether it's in healthcare or housing
55:26
SPEAKER_00: or whatever it is so I mean AI for policy and policy evaluation and design
55:34
SPEAKER_02: Yeah, I think that's a great idea.
55:36
SPEAKER_02: I think the current tools are actually quite good at this.
55:38
SPEAKER_02: I think optimistically,
55:40
SPEAKER_02: you could say that this is kind of happening in the field,
55:41
SPEAKER_02: in the academic field of economics.
55:44
SPEAKER_02: Right.
55:44
SPEAKER_02: In a way that's sort of analogous or maybe even directly on point,
55:47
SPEAKER_02: which is,
55:48
SPEAKER_02: you know,
55:48
SPEAKER_02: my sense of like academic economics is sort of shifting from,
55:53
SPEAKER_02: call it the post-World War II method of sort of physics,
55:55
SPEAKER_02: you know, kind of the physics of economics where everything is formulas.
55:59
SPEAKER_02: A16z
56:19
SPEAKER_02: I would love to see if the GAO and CBO and others really jump into this in
56:22
SPEAKER_02: a significant way because they could really help us
56:40
SPEAKER_00: AI Policy Brief a16z
56:58
SPEAKER_00: A16z
57:20
SPEAKER_02: I think there's a lot.
57:21
SPEAKER_02: I'm reasonably optimistic.
57:23
SPEAKER_02: I think there's a lot going on.
57:26
SPEAKER_02: It happens on a number of fronts.
57:28
SPEAKER_02: So one that's just very specific is reindustrializing on the defense side.
57:34
SPEAKER_02: And so as your organization has studied at length,
57:37
SPEAKER_02: there are very real issues about the physical supply chain that goes into the U.S.
57:40
SPEAKER_02: military.
57:42
SPEAKER_02: and national security.
57:42
SPEAKER_02: And so we are intensely proud of our companies that are in that space.
57:47
SPEAKER_02: And I would say the current administration has been incredibly supportive
57:51
SPEAKER_02: of those efforts and is working very aggressively with young companies,
57:55
SPEAKER_02: really for the first time.
57:56
SPEAKER_02: A16z
58:19
SPEAKER_02: A16z
58:42
SPEAKER_02: Yeah.
58:42
SPEAKER_02: And of course, you know, Elon, of course, was a, you know, has been an incredible leader there.
58:46
SPEAKER_02: But there are many,
58:46
SPEAKER_02: many other founders that are inspired by Elon,
58:49
SPEAKER_02: inspired by Palmer Luckey, and the team at Anderil and these other companies.
58:53
SPEAKER_02: And, you know,
58:53
SPEAKER_02: their startups,
58:54
SPEAKER_02: you know,
58:55
SPEAKER_02: many of whom are backing,
58:56
SPEAKER_02: but their startups doing new nuclear,
58:57
SPEAKER_02: you know,
58:57
SPEAKER_02: nuclear fission reactors for the first time in decades,
59:00
SPEAKER_02: their startups doing we have multiple companies going after rare earth,
59:04
SPEAKER_02: you know, mineral discovery, extraction processing,
59:08
SPEAKER_02: There's energy.
59:09
SPEAKER_02: We've actually backed a company, by the way.
59:10
SPEAKER_02: I mentioned electrical transformers are sold out.
59:12
SPEAKER_02: We backed
59:12
SPEAKER_02: a new generation electrical transformer company that's building electrical
59:15
SPEAKER_02: transformers in the US.
59:18
SPEAKER_02: And so, yeah, so I think there's optimism there.
59:22
SPEAKER_02: There, by the way,
59:22
SPEAKER_02: I would say even in California,
59:24
SPEAKER_02: you know,
59:24
SPEAKER_02: where there's all kinds of, you know, both a lot of good and bad things happening.
59:27
SPEAKER_02: But, you know,
59:28
SPEAKER_02: there's like a new industrial,
59:30
SPEAKER_02: I don't know,
59:30
SPEAKER_02: even like manufacturing ecosystem,
59:32
SPEAKER_02: entrepreneurial cluster in and around Los Angeles.
59:34
SPEAKER_02: Yeah.
59:36
SPEAKER_02: You know, around El Segundo and Hawthorne and these places where SpaceX was born and so forth.
59:41
SPEAKER_02: Yeah.
59:41
SPEAKER_02: um and where Andrew is based and
59:43
SPEAKER_02: so you know you like optimistically we maybe get actually two Silicon Valleys
59:47
SPEAKER_02: in California we get kind of software AI Silicon Valley up north
59:50
SPEAKER_02: and we get like defense and industrial Silicon Valley around LA yeah um
59:55
SPEAKER_02: so I think I think that's a possibility um like let's put it this way
59:58
SPEAKER_02: the entrepreneurs all want to do it um the money by the way is lined up but
1:00:01
SPEAKER_02: the money is available to do it the
1:00:03
SPEAKER_02: at least this uh government um uh really wants this to happen
1:00:07
SPEAKER_02: and is doing everything that it can to
1:00:09
SPEAKER_02: AI Policy Brief
1:00:25
SPEAKER_02: you want jobs and all these communities have gotten hollowed out you know and
1:00:28
SPEAKER_02: in many cases have gone sharply to the right as
1:00:30
SPEAKER_02: a result like you actually want reindustrialization
1:00:32
SPEAKER_02: because you know you want you want those people to have good new jobs and
1:00:35
SPEAKER_02: so optimistically this could be a bipartisan effort yeah and I would say even
1:00:39
SPEAKER_00: the we can there's a debate on what tools are the best tools but
1:00:44
SPEAKER_00: the previous administration made efforts around chips and you talked
1:00:47
SPEAKER_00: about chip making that were important and similarly in under
1:00:52
SPEAKER_00: the IRA I would just say
1:00:55
SPEAKER_00: Across both administrations, this is a huge priority across parties, I would say.
1:01:01
SPEAKER_00: The thing that I find interesting from an investor's perspective is that are we in
1:01:06
SPEAKER_00: a moment where you could pursue financial objectives as an investor
1:01:11
SPEAKER_00: and non-financial objectives around,
1:01:12
SPEAKER_00: say, national security or national interest without giving up returns?
1:01:16
SPEAKER_00: And it seems like you're saying we are in that moment.
1:01:19
SPEAKER_02: So I don't think it's the case that there's a direct tradeoff, at least for what we do.
1:01:27
SPEAKER_02: There's not a case that there's a direct tradeoff of financial investing versus
1:01:30
SPEAKER_02: the larger goals.
1:01:31
SPEAKER_02: What you do in our world is you organize the entire purpose of the company.
1:01:34
SPEAKER_02: A16z
1:01:51
SPEAKER_02: of wanting to do something specific and then they're basically saying like what's
1:01:54
SPEAKER_02: the way to turn that into a mission that then attracts the smartest people in
1:01:57
SPEAKER_02: the field right that attracts people who are the most ambitious about
1:02:00
SPEAKER_02: about undertaking your programs you know you infuse the company
1:02:03
SPEAKER_02: with patriotism you get a completely different kind of energy than you get
1:02:06
SPEAKER_02: if you're just outsourcing everything to China
1:02:08
SPEAKER_02: You get,
1:02:09
SPEAKER_02: by the way,
1:02:09
SPEAKER_02: co-located R&D happening with manufacturing,
1:02:11
SPEAKER_02: which is actually what everybody actually wants when they're trying
1:02:15
SPEAKER_02: to manufacture anything complicated.
1:02:17
SPEAKER_02: You then bring customers into this and the customers have their own incentives
1:02:21
SPEAKER_02: to want to buy more American-produced goods.
1:02:23
SPEAKER_02: So what you do in our world is you create the strategy first and then you line up
1:02:30
SPEAKER_02: the financial plan behind that.
1:02:32
SPEAKER_02: and so from that standpoint you basically just like set out like these these are
1:02:36
SPEAKER_02: the kinds of companies you're building you're not building companies
1:02:38
SPEAKER_02: that just defaulted Chinese you know contract manufacturing like that's not anywhere
1:02:41
SPEAKER_02: in the DNA of the company and then you see if you can actually build
1:02:44
SPEAKER_02: a superior model with a new approach and I think I think we have you know probably
1:02:48
SPEAKER_02: at this point dozens of companies that are doing what I just described
1:02:52
SPEAKER_00: Thank you so much for spending all this time today
1:02:55
SPEAKER_00: and we'll be watching what you're doing and also what you keep saying
1:02:58
SPEAKER_00: about these things including in your role at PCAST so thank you for contributing to
1:03:03
SPEAKER_00: the national debate Mark it's really fantastic Good thank you so much
1:03:07
SPEAKER_00: for having me I really appreciate it Thank you for listening to today's conversation
1:03:11
SPEAKER_00: with Mark Andreessen you can find this episode and more on CSIS.org YouTube
1:03:16
SPEAKER_00: or wherever you get your podcasts
1:03:18
SPEAKER_00: This is Naveen Girishankar reminding you that everyone has a role to play in winning
1:03:22
SPEAKER_00: the tech race.
1:03:25
SPEAKER_01: Thanks for listening to this episode of the A16Z Podcast.
1:03:29
SPEAKER_01: If you liked this episode,
1:03:30
SPEAKER_01: be sure to like,
1:03:31
SPEAKER_01: comment,
1:03:31
SPEAKER_01: subscribe, leave us a rating or review, and share it with your friends and family.
1:03:36
SPEAKER_01: For more episodes, go to YouTube, Apple Podcasts, and Spotify.
1:03:40
SPEAKER_01: Follow us on X at A16Z and subscribe to our Substack at a16z.substack.com.
1:03:46
SPEAKER_01: Thanks again for listening and I'll see you in the next episode.
1:03:50
SPEAKER_01: This information is for educational purposes only and is not a recommendation
1:03:54
SPEAKER_01: to buy, hold, or sell any investment or financial product.
1:03:57
SPEAKER_01: This podcast has been produced by a third party
1:03:59
SPEAKER_01: and may include paid promotional advertisements,
1:04:01
SPEAKER_01: other company references, and individuals unaffiliated with a16z.
1:04:05
SPEAKER_01: Such advertisements,
1:04:06
SPEAKER_01: companies,
1:04:06
SPEAKER_01: and individuals are not endorsed by AH Capital Management LLC,
1:04:10
SPEAKER_01: a16z, or any of its affiliates.
1:04:12
SPEAKER_01: Information is from sources deemed reliable on the date of publication,
1:04:16
SPEAKER_01: but a16z does not guarantee its accuracy.