Latest / Faster, Please! — The Podcast / ✨ My chat (+transcript) with tech policy expert Neil Chilson on regulating GenAI
Transcript
- 0:06SPEAKER_01: Welcome to Faster Please, the podcast.
- 0:09SPEAKER_01: I'm your host, Jim Pethikoukis.
- 0:12SPEAKER_01: Several times a month,
- 0:13SPEAKER_01: I'll feature a lively conversation with a fascinating and provocative guest about
- 0:18SPEAKER_01: how to make the world a better place through scientific discovery,
- 0:21SPEAKER_01: technological innovation,
- 0:23SPEAKER_01: and economic growth.
- 0:24SPEAKER_01: You're also going to want to check out my Faster Please newsletter.
- 0:27SPEAKER_01: You're on Substack throughout the week for fresh essays,
- 0:30SPEAKER_01: Q&As,
- 0:31SPEAKER_01: and stories from around the internet and around the world.
- 0:36SPEAKER_01: Washington's initial approach to AI regulation has evolved from a knee-jerk fear
- 0:41SPEAKER_01: response to a more nuanced appreciation of its capabilities and potential risks.
- 0:47SPEAKER_01: Today on Faster Please,
- 0:49SPEAKER_01: the podcast,
- 0:50SPEAKER_01: I talk with technology policy expert Neil Chilson about national competition,
- 0:55SPEAKER_01: defense,
- 0:56SPEAKER_01: and federal versus state regulation in this brave new world of artificial intelligence.
- 1:00SPEAKER_01: Chilson is the head of AI policy at the Abundance Institute.
- 1:04SPEAKER_01: He is a lawyer, computer scientist, and former chief technologist at the Federal Trade Commission.
- 1:10SPEAKER_01: He's also the author of Getting Out of Control, Emergent Leadership in a Complex World.
- 1:15SPEAKER_01: Neil, welcome to the podcast.
- 1:17SPEAKER_01: Great to be here.
- 1:18SPEAKER_01: We're going to talk a bit about AI regulation, the future of regulation.
- 1:22SPEAKER_01: So let me start with this.
- 1:25SPEAKER_01: Last summer, the Biden administration put out a big executive order on AI.
- 1:29SPEAKER_01: I assume the Trump administration will repeal that and do their own thing.
- 1:36SPEAKER_01: Any idea what that thing will be?
- 1:39SPEAKER_00: Yeah,
- 1:40SPEAKER_00: so the Biden executive order repealing it is actually part of the GOP platform,
- 1:48SPEAKER_00: which does not say a lot about AI,
- 1:49SPEAKER_00: but it does say that it's definitely going to get rid of the Biden executive order.
- 1:53SPEAKER_00: So I think that's the first order of business.
- 1:55SPEAKER_00: The repeal and replace process,
- 1:58SPEAKER_00: the Trump administration,
- 1:59SPEAKER_00: the previous Trump administration,
- 2:00SPEAKER_00: actually had a couple of executive orders on AI.
- 2:04SPEAKER_00: And they were very...
- 2:08SPEAKER_00: big picture.
- 2:09SPEAKER_00: They were not nearly as pro-regulatory as the Biden executive order, and they saw a lot of the potential.
- 2:16SPEAKER_00: So I'd expect a shift back towards a vision of AI as a force for good.
- 2:22SPEAKER_00: I'd expect a shift towards
- 2:26SPEAKER_00: sort of the international dynamics here, right?
- 2:29SPEAKER_00: That we need to keep ahead of China in AI.
- 2:32SPEAKER_00: We have a lead on the tech.
- 2:33SPEAKER_00: We have the best companies in the world.
- 2:35SPEAKER_00: I think a Trump administration is really going to amp up that rhetoric.
- 2:39SPEAKER_00: And I would expect the executive order to reflect the need to keep the US in the lead on AI technology.
- 2:47SPEAKER_01: And just to be clear, so that emphasis differs from the Biden emphasis in what way?
- 2:55SPEAKER_00: So the Biden emphasis,
- 2:57SPEAKER_00: when you read the executive order,
- 2:58SPEAKER_00: it has some nice language up top about how this is a great new technology.
- 3:03SPEAKER_00: It's very powerful.
- 3:04SPEAKER_00: But overwhelmingly, the Biden executive order is directed at the risk of AI.
- 3:09SPEAKER_00: And in particular, not existential risk,
- 3:12SPEAKER_00: More the traditional risks that academics who have talked about the internet have
- 3:17SPEAKER_00: had for a long time,
- 3:18SPEAKER_00: these risks of bias or risks to privacy or risks to safety or deepfakes.
- 3:23SPEAKER_00: And to be honest, to be clear, there are risks to all of these technologies.
- 3:27SPEAKER_00: But the Biden executive order really pounded that home.
- 3:29SPEAKER_00: The emphasis was very much on what are the problems that this tech could cause.
- 3:33SPEAKER_00: And what do we as a federal government need to do to get in here and make sure it's safe for everybody?
- 3:39SPEAKER_00: And I would expect that would be a big change.
- 3:41SPEAKER_00: I don't see,
- 3:41SPEAKER_00: especially on the bias front,
- 3:43SPEAKER_00: I don't see a Trump administration emphasizing that as a primary thing that the
- 3:47SPEAKER_00: federal government needs to fix about AI.
- 3:50SPEAKER_00: In fact,
- 3:50SPEAKER_00: with people like Elon Musk having the ear of the president,
- 3:54SPEAKER_00: I would expect maybe to go in the opposite direction,
- 3:57SPEAKER_00: that these ideas around bias are inflated,
- 4:01SPEAKER_00: that these risks aren't
- 4:03SPEAKER_00: really real.
- 4:04SPEAKER_00: And to the extent that they are,
- 4:06SPEAKER_00: there's no business of the federal government to step in and tell companies how to
- 4:10SPEAKER_00: bias or de-bias their products.
- 4:13SPEAKER_01: Yeah,
- 4:13SPEAKER_01: one thing that sort of confuses me on the Musk angle is that it seemed that he was used,
- 4:17SPEAKER_01: at least used to be very concerned about sort of these somewhat science fictional
- 4:22SPEAKER_01: existential risks to AI.
- 4:25SPEAKER_01: And I guess my concern is like that we'll get that version of Musk again talking to the White House.
- 4:31SPEAKER_01: And maybe he says like, I'm not worried about bias, but I'm still worried about it killing us all.
- 4:36SPEAKER_01: Like, is there any concern like there that like that theme, which I think
- 4:40SPEAKER_01: seems have faded a little bit from the public conversation.
- 4:43SPEAKER_01: Maybe I'm wrong.
- 4:44SPEAKER_01: Like that will, that will like reemerge.
- 4:47SPEAKER_00: Yeah, I agree with you that I think that theme has faded.
- 4:51SPEAKER_00: The early Senate hearings were very much in that vein.
- 4:56SPEAKER_00: They were about the existential risk.
- 4:58SPEAKER_00: And some of that was the people who were up there talking.
- 5:01SPEAKER_00: This is something that's been on the mind of some of the leaders of the cutting edge of the tech space.
- 5:06SPEAKER_00: And it's part of the reason why they got into it.
- 5:08SPEAKER_00: There's always been a tension there.
- 5:10SPEAKER_00: There is some sort of
- 5:13SPEAKER_00: dynamic here where they're like,
- 5:15SPEAKER_00: this stuff is super dangerous and super powerful,
- 5:17SPEAKER_00: so I need to be the one creating it and controlling it.
- 5:21SPEAKER_00: I think Musk still kind of falls in that bucket, so I share a little bit of that concern.
- 5:26SPEAKER_00: But I think you're right that Congress has said
- 5:29SPEAKER_00: oh, those things seem really far-fetched.
- 5:32SPEAKER_00: That's not how we're going to focus our time.
- 5:35SPEAKER_00: And so I would expect that to continue, even with a Musk-influenced administration.
- 5:44SPEAKER_00: I actually don't think that there is necessarily a big tension between that and a
- 5:49SPEAKER_00: pushback against the sort of red tape regulatory approach to AI that is kind of the
- 5:56SPEAKER_00: more traditional approach
- 5:59SPEAKER_00: pessimistic, precautionary approach to technology generally.
- 6:02SPEAKER_00: I think Musk is a guy who hates red tape.
- 6:06SPEAKER_00: I think he's seen it in his own businesses, how it's slowed down, launches of all sorts.
- 6:15SPEAKER_00: So I think you can hate red tape and be worried about this existential risk.
- 6:18SPEAKER_00: It's not necessarily intentioned,
- 6:21SPEAKER_00: but it will be interesting to see how those play out,
- 6:26SPEAKER_00: how Moscow influences the policy of the Trump administration on AI.
- 6:31SPEAKER_01: One issue that seemed to be coming up over and over again is differing opinions
- 6:36SPEAKER_01: among technologists,
- 6:37SPEAKER_01: venture capitalists about sort of the open source issue.
- 6:40SPEAKER_01: How does that play out, you think, heading into a Trump administration?
- 6:46SPEAKER_01: When I listen to the Andreessen Horowitz podcast, those guys seem very concerned.
- 6:50SPEAKER_00: Yeah, so there's a lot of disagreement sort of about how open source plays out.
- 6:56SPEAKER_00: Open source,
- 6:57SPEAKER_00: should be pointed out first,
- 6:58SPEAKER_00: is a core technology across everything that people who develop software use.
- 7:05SPEAKER_00: And so...
- 7:06SPEAKER_00: Most websites run on open source software.
- 7:10SPEAKER_00: Most development tools have a huge open source component.
- 7:14SPEAKER_00: And it's one of the best ways to develop and test technology is by sharing it with
- 7:19SPEAKER_00: people and having people build on it.
- 7:21SPEAKER_00: So I do think it is a really important technology in the AI space.
- 7:26SPEAKER_00: We've seen that already.
- 7:27SPEAKER_00: People are building smaller models,
- 7:29SPEAKER_00: doing new things in open source that it cost a lot of money to do in the first instance.
- 7:36SPEAKER_00: um maybe in the closed source the concerns that people raise is that this uh
- 7:42SPEAKER_00: especially in the national security space or the national competition that this
- 7:46SPEAKER_00: sort of exposes our best research to other countries and i think there's a couple
- 7:51SPEAKER_00: responses to that the first one is that closed source is no guarantee that those
- 7:57SPEAKER_00: people don't have that technology as well in fact most of these models fit on the
- 8:01SPEAKER_00: thumb drive
- 8:03SPEAKER_00: Most of these AI labs are not run like nuclear facilities,
- 8:10SPEAKER_00: and it's much easier to smuggle a thumb drive out than it is to smuggle a gram of
- 8:16SPEAKER_00: plutonium or something like that.
- 8:20SPEAKER_00: They're going to get this software.
- 8:22SPEAKER_00: They're going to develop it themselves.
- 8:24SPEAKER_00: We can't out-China China.
- 8:26SPEAKER_00: We should lean into what we're really good at,
- 8:27SPEAKER_00: and that is a dynamic software development environment of which open source,
- 8:31SPEAKER_00: I think,
- 8:32SPEAKER_00: is a key component.
- 8:33SPEAKER_00: It also offers,
- 8:35SPEAKER_00: I think,
- 8:36SPEAKER_00: in many ways,
- 8:37SPEAKER_00: an alternative to centralized sources of artificial intelligence models,
- 8:45SPEAKER_00: which can offer a bunch of user...
- 8:49SPEAKER_00: user interface-based benefits.
- 8:51SPEAKER_00: They're just easier to use.
- 8:53SPEAKER_00: It's much easier to log into OpenAI and use their chat GPT than it is to download
- 8:59SPEAKER_00: and build your own model.
- 9:00SPEAKER_00: But it is really nice as a competitive
- 9:04SPEAKER_00: gap filler to have thousands and thousands of other models that might do something
- 9:08SPEAKER_00: specific or have a specific orientation,
- 9:11SPEAKER_00: which you can train on your own.
- 9:13SPEAKER_00: And those exist because of the open source ecosystem.
- 9:15SPEAKER_00: So I think it solves a lot of problems, probably a lot more than it creates.
- 9:19SPEAKER_01: So what would you expect,
- 9:21SPEAKER_01: let's focus on the federal level,
- 9:23SPEAKER_01: for this Congress,
- 9:25SPEAKER_01: for the Trump administration to do other than sort of broadly affirm that we love AI,
- 9:30SPEAKER_01: we hope it continues?
- 9:32SPEAKER_01: Will there be any sort of regulatory rule,
- 9:35SPEAKER_01: any sort of guidance that would be that would anyway constrain or direct this technology?
- 9:41SPEAKER_01: Maybe it's maybe in the area of the open the frontier models.
- 9:44SPEAKER_01: I don't know.
- 9:45SPEAKER_00: Yeah, so I think we're likely to see a lot of action at the use level.
- 9:51SPEAKER_00: What are the various uses of various applications?
- 9:54SPEAKER_00: And how does AI change that?
- 9:56SPEAKER_00: So in transportation, in health care, this is a general purpose technology.
- 10:01SPEAKER_00: And so it's going to be deployed in lots of spaces.
- 10:03SPEAKER_00: A lot of these spaces already have a lot of regulatory
- 10:06SPEAKER_00: frameworks in place.
- 10:08SPEAKER_00: I think we'll see lots of agencies looking at to see like,
- 10:12SPEAKER_00: hey,
- 10:12SPEAKER_00: this new technology,
- 10:13SPEAKER_00: does it really change anything about how we regulate medical devices?
- 10:17SPEAKER_00: If it does, how do we need to accommodate that?
- 10:20SPEAKER_00: What are the maybe unique risks?
- 10:22SPEAKER_00: What are the unique opportunities that maybe the current framework doesn't really allow for?
- 10:26SPEAKER_00: I think we'll see a lot of that.
- 10:29SPEAKER_00: I think once you get up to the abstract model level, it's much harder to figure out what problem
- 10:36SPEAKER_00: Both are we trying to solve at the model level and do we have the capability to
- 10:40SPEAKER_00: solve at the model level?
- 10:42SPEAKER_00: So if we're worried about people developing bioweapons with this technology,
- 10:48SPEAKER_00: is making sure the model doesn't allow that,
- 10:52SPEAKER_00: is that useful?
- 10:52SPEAKER_00: Is it even possible?
- 10:54SPEAKER_00: Or should we focus those attentions maybe down on people can't secure the
- 10:58SPEAKER_00: components that they need to execute a biohazard?
- 11:02SPEAKER_00: Would that be a more productive place?
- 11:04SPEAKER_00: And so I don't see a lot of action, honestly, at the model level.
- 11:08SPEAKER_00: Maybe there'll be some reporting requirements or training requirements.
- 11:12SPEAKER_00: The executive order had those, although they did it.
- 11:16SPEAKER_00: They use something called the Defense Production Act,
- 11:18SPEAKER_00: I think probably unconstitutionally how they use that.
- 11:23SPEAKER_00: But that's going to go away.
- 11:25SPEAKER_00: If that gets filled in by Congress, that there's some sort of reporting regime, maybe that's possible.
- 11:30SPEAKER_00: But Congress doesn't seem to be able to get those types of really high level tech
- 11:36SPEAKER_00: regulations across countries.
- 11:38SPEAKER_00: across the line.
- 11:39SPEAKER_00: They haven't done it with privacy legislation for a long time,
- 11:42SPEAKER_00: and everybody sort of seems to think that would be a good idea.
- 11:45SPEAKER_00: But I think we'll continue to see efforts at the agency level.
- 11:50SPEAKER_00: And then there might be one thing Congress might do is they might spend some money in this space, right?
- 11:56SPEAKER_00: Maybe there will be some new investment or maybe we'll get some,
- 11:59SPEAKER_00: the national laboratories will get some money to do additional AI research.
- 12:04SPEAKER_00: That has its own challenges, but most of them are sort of financial challenges.
- 12:08SPEAKER_00: They're not so much like whether or not it's going to impede the industry.
- 12:13SPEAKER_00: And so that's kind of how I think it'll likely play out at the federal level.
- 12:17SPEAKER_01: A report just came out yesterday as we're recording this,
- 12:21SPEAKER_01: you know,
- 12:21SPEAKER_01: from the sort of outside advisory group on U.S.-China relations that advise the government.
- 12:26SPEAKER_01: And they're calling for a Manhattan Project to get to an artificial general intelligence,
- 12:32SPEAKER_01: I assume,
- 12:33SPEAKER_01: before China or anybody else.
- 12:37SPEAKER_01: Is that a good idea?
- 12:38SPEAKER_01: Do you think we'll do that?
- 12:39SPEAKER_01: What do you what do you make of that recommendation, which caused quite a stir when it came out?
- 12:44SPEAKER_00: Yeah, it's a really interesting report.
- 12:46SPEAKER_00: If you read through the body of the report, it's pretty standard international competitiveness analysis.
- 12:55SPEAKER_00: It says, what are the supply chains for chips?
- 12:57SPEAKER_00: How does it look?
- 12:58SPEAKER_00: How do we compare on talent development?
- 13:01SPEAKER_00: compare on the industry backing, investment, things like that.
- 13:05SPEAKER_00: And we compare very well, overall, the report says.
- 13:08SPEAKER_00: But then all of a sudden,
- 13:09SPEAKER_00: at the top level,
- 13:10SPEAKER_00: the first recommendation talks about artificial general intelligence.
- 13:14SPEAKER_00: This is the kind of AI that doesn't exist yet,
- 13:18SPEAKER_00: but it's the kind that could basically do everything a human could do at the
- 13:23SPEAKER_00: intellectual level that a human could do it.
- 13:27SPEAKER_00: It's interesting because that recommendation,
- 13:29SPEAKER_00: it doesn't seem to be founded on anything that's actually in the report.
- 13:33SPEAKER_00: There's no other discussion in the report about artificial general intelligence or
- 13:37SPEAKER_00: how important it is strategically or anything like that.
- 13:40SPEAKER_00: And yet they want to spend Manhattan-level project amounts of money.
- 13:46SPEAKER_00: I think in today's dollars,
- 13:47SPEAKER_00: that would be like $30 billion to create this artificial general intelligence.
- 13:53SPEAKER_00: And I
- 13:55SPEAKER_00: I don't know what to make of that.
- 13:57SPEAKER_00: And more than that, I think it's very unlikely to move the policy discussion.
- 14:02SPEAKER_00: Maybe it moves the Overton window so people are talking like, yeah, we need a Manhattan Project.
- 14:07SPEAKER_00: But I don't think I don't think that it's likely to do anything for the for the most part.
- 14:14SPEAKER_00: Artificial general intelligence, I don't understand what the appeal of that is, frankly.
- 14:18SPEAKER_00: I mean,
- 14:19SPEAKER_00: it has a sort of theoretical appeal that we could have a computer that could do all
- 14:22SPEAKER_00: the things that a person could do.
- 14:24SPEAKER_00: But in the modern economy,
- 14:26SPEAKER_00: it's actually better to have things that are really good at doing a specific set of
- 14:29SPEAKER_00: things rather than having a generalist that you can deploy lots of different places.
- 14:34SPEAKER_00: Especially if you're talking about software,
- 14:35SPEAKER_00: why not train something that can do something specific really well?
- 14:38SPEAKER_00: And so I think that would slot into our economy better.
- 14:41SPEAKER_00: I think it's much more likely to be the most productive value of the intense
- 14:46SPEAKER_00: computation time and money that it takes to train these types of models.
- 14:50SPEAKER_00: And so it seems like a strange thing to emphasize in our federal spending,
- 14:59SPEAKER_00: even if we're talking about the national security implications.
- 15:02SPEAKER_00: It would seem like it'd be much better to train a model that's specifically built for
- 15:06SPEAKER_00: You know,
- 15:06SPEAKER_00: some some type of drone warfare or something that rather than trying to,
- 15:10SPEAKER_00: like,
- 15:10SPEAKER_00: make it good at everything and then say,
- 15:12SPEAKER_00: oh,
- 15:12SPEAKER_00: now we're going to use you to fly drones.
- 15:14SPEAKER_00: That doesn't seem to make a ton of sense.
- 15:16SPEAKER_01: So you talked about the federal level, certainly not that the states seem to need a nudge.
- 15:21SPEAKER_01: But if if they see that Washington doing less,
- 15:24SPEAKER_01: I'm sure there'll be plenty of state governments saying,
- 15:26SPEAKER_01: well,
- 15:26SPEAKER_01: then we need to do more.
- 15:27SPEAKER_01: We need to fill up the gap with our state regulation.
- 15:30SPEAKER_01: That already seems to be happening.
- 15:33SPEAKER_01: Will that continue to happen?
- 15:35SPEAKER_01: And can the Trump administration stop that?
- 15:39SPEAKER_00: So I think it will continue to happen.
- 15:41SPEAKER_00: The question is whether or not what kind of gap is left by the Trump administration.
- 15:46SPEAKER_00: I would say what the Biden administration left was a vision gap.
- 15:49SPEAKER_00: They didn't really have like an overarching vision for how the U.S.
- 15:53SPEAKER_00: was going to engage with this technology at the federal level.
- 15:58SPEAKER_00: Unlike
- 15:59SPEAKER_00: You know,
- 15:59SPEAKER_00: the Clinton administration would set out like a pretty clear vision for how the
- 16:03SPEAKER_00: federal government plan to engage on the early version of the Internet.
- 16:09SPEAKER_00: And what it said was,
- 16:10SPEAKER_00: we're going to let for some really good reasons,
- 16:13SPEAKER_00: we're going to let the commercial sector lead on development here.
- 16:17SPEAKER_00: I think sending a signal like that could have,
- 16:19SPEAKER_00: you know,
- 16:19SPEAKER_00: sort of bully pulpit effect,
- 16:22SPEAKER_00: especially I think in redder States,
- 16:25SPEAKER_00: you'll still see States like California and New York.
- 16:27SPEAKER_00: They're gonna, they're, they're very, they're listening to Europe on how to do stuff in this space.
- 16:32SPEAKER_00: Still.
- 16:33SPEAKER_00: Are we still listening to you?
- 16:35SPEAKER_01: I mean, who are the people out there and think like they've got it figured.
- 16:38SPEAKER_01: I, I, I,
- 16:39SPEAKER_01: I understand that maybe that's your initial impulse when you have a new technology
- 16:43SPEAKER_01: and you're like,
- 16:43SPEAKER_01: I don't know what to do.
- 16:44SPEAKER_01: So who is doing something on it?
- 16:46SPEAKER_01: But like we've had a little bit of time and I just don't get like anybody who would
- 16:52SPEAKER_01: default to be like,
- 16:53SPEAKER_01: man,
- 16:53SPEAKER_01: we're just going to pull off like,
- 16:54SPEAKER_01: you know,
- 16:54SPEAKER_01: we're going to look at a couple of EU white papers and off to the races here in our state.
- 16:59SPEAKER_00: I think a lot of this is a hangover sort of, you know, the social media fights.
- 17:04SPEAKER_00: AI,
- 17:04SPEAKER_00: if you pull it just at that level,
- 17:06SPEAKER_00: if you're like,
- 17:06SPEAKER_00: hey,
- 17:07SPEAKER_00: do you think AI is going to be good or bad for your job or for the economy?
- 17:10SPEAKER_00: Americans are somewhat skeptical.
- 17:13SPEAKER_00: It's because they think of AI in the cultural context that includes like Terminator and automation.
- 17:19SPEAKER_00: And so they think of it that way.
- 17:20SPEAKER_00: They don't think about, you know, the thousand of...
- 17:23SPEAKER_00: thousands of applications on their phones that use artificial intelligence.
- 17:27SPEAKER_00: I think there's a political moment here around this.
- 17:32SPEAKER_00: The Europeans jumped in and said, hey, we're the first to regulate in this space comprehensively.
- 17:37SPEAKER_00: I think they're dialing that back since some of their member states are like,
- 17:40SPEAKER_00: hey,
- 17:40SPEAKER_00: this is killing our own homegrown AI industry.
- 17:45SPEAKER_00: But
- 17:46SPEAKER_00: For some reason,
- 17:48SPEAKER_00: California and New York seem to be embracing that,
- 17:50SPEAKER_00: and I think they probably will continue to.
- 17:52SPEAKER_00: At the very local level,
- 17:55SPEAKER_00: at the state level,
- 17:56SPEAKER_00: there's just weird incentives to do something,
- 18:00SPEAKER_00: and then you don't really pay a lot of consequences down the road.
- 18:06SPEAKER_00: Having said that,
- 18:07SPEAKER_00: there was a controversial bill that was very aggressively pushed,
- 18:12SPEAKER_00: SB 1047,
- 18:13SPEAKER_00: in California over the summer,
- 18:15SPEAKER_00: and it got killed.
- 18:17SPEAKER_00: It got canned by Gavin Newsom in the end.
- 18:21SPEAKER_00: And I think that's sort of a unique artifact of California's go-along-to-get-along legislature process.
- 18:29SPEAKER_00: Where even people who don't support bills sort of vote for them,
- 18:32SPEAKER_00: kind of knowing that Gavin or that the governor will bring down the veto when it
- 18:37SPEAKER_00: doesn't make political sense.
- 18:38SPEAKER_00: So anyways, all of this is to say California is going to California.
- 18:42SPEAKER_00: I think we're starting to see what and what concerns me is we're starting to see
- 18:46SPEAKER_00: the shopping of bills that look a lot like.
- 18:50SPEAKER_00: the way privacy has worked across the states.
- 18:52SPEAKER_00: And in some cases are being pushed by the same organizations that represent
- 18:57SPEAKER_00: compliance companies saying,
- 19:00SPEAKER_00: hey,
- 19:00SPEAKER_00: yeah,
- 19:01SPEAKER_00: we need to do all this algorithmic bias auditing or safety auditing and states
- 19:07SPEAKER_00: should require it.
- 19:08SPEAKER_00: There's a Texas draft bill that has been floated right now.
- 19:12SPEAKER_00: And you wouldn't think that Texas would be on the frontier of
- 19:17SPEAKER_00: banning differential effects in bias from AI.
- 19:22SPEAKER_00: It doesn't really sound particularly red statey.
- 19:26SPEAKER_00: But these things are getting shopped around.
- 19:28SPEAKER_00: And if it moves in Texas, it'll move other places too.
- 19:31SPEAKER_00: So I worry about that level of red tape coming at the state level.
- 19:36SPEAKER_00: And that's just going to be a ground fight, ground warfare on the legislative front at the state level.
- 19:43SPEAKER_01: So federal preemption, what is that and how would that work?
- 19:47SPEAKER_01: And is that possible?
- 19:48SPEAKER_00: Yeah, it's really hard in this space because the technology is so general.
- 19:52SPEAKER_00: I would say Congress could,
- 19:54SPEAKER_00: of course,
- 19:55SPEAKER_00: they could write something that was very broad and preempted certain types of
- 19:59SPEAKER_00: regulation of models.
- 20:00SPEAKER_00: And maybe that's a good idea.
- 20:02SPEAKER_00: I've seen some draft language around that.
- 20:05SPEAKER_00: On the other hand, I do believe in federalism.
- 20:08SPEAKER_00: And these aren't quite the same sort of network-based technologies that only make sense in federalism.
- 20:15SPEAKER_00: in a national sweep so maybe maybe there's an argument that states we should let
- 20:19SPEAKER_00: states suffer the consequences of their own regulatory approaches um that hurts my
- 20:25SPEAKER_00: heart a little bit uh just to think about the future because there are a lot of
- 20:29SPEAKER_00: talented people in those states who are gonna you know find out it's the lawyers
- 20:33SPEAKER_00: who are their main constraint those types of transaction costs just they will slow
- 20:38SPEAKER_00: us down so i think if it looks like
- 20:42SPEAKER_00: We're falling behind in the US because we can't get out of our own way regulatorily.
- 20:48SPEAKER_00: I think there will be more impulse to fix things.
- 20:51SPEAKER_00: There are some other creative solutions such as interstate compacts to try to get
- 20:56SPEAKER_00: people to level up across multiple states.
- 21:00SPEAKER_00: about how they're going to treat AI and allow innovation to flourish.
- 21:04SPEAKER_00: And so I think we'll see more of those experiments,
- 21:06SPEAKER_00: but it is really hard at the federal level to preempt just because there's so many
- 21:10SPEAKER_00: state-based interests who are going to push back against that sort of thing.
- 21:15SPEAKER_01: As far as AI sort of influencing what we do elsewhere,
- 21:19SPEAKER_01: one thing you wrote about recently in a really great essay,
- 21:21SPEAKER_01: which I've already sort of drawn upon in some of these questions,
- 21:25SPEAKER_01: is thinking about immigration and AI talent coming to the United States.
- 21:33SPEAKER_01: What I think is now a widely accepted understanding that this is an important
- 21:38SPEAKER_01: technology and we certainly want to be the leader in this technology.
- 21:41SPEAKER_01: Does that change how we think about
- 21:45SPEAKER_01: at least sort of very high-skilled immigration.
- 21:49SPEAKER_00: I think it should.
- 21:50SPEAKER_00: I mean, frankly, we should have changed our minds about some of this stuff a long time ago.
- 21:55SPEAKER_00: We should be wanting the most talented people to come here and stay here.
- 21:58SPEAKER_00: The most talented people in the world already come here for school often.
- 22:01SPEAKER_00: I know when I was in computer science grad school,
- 22:05SPEAKER_00: it was full of people who really desperately wanted to stay in the US and build
- 22:09SPEAKER_00: companies and build products.
- 22:12SPEAKER_00: And some of them struggled really hard to figure out a way to do it.
- 22:17SPEAKER_00: And so I think that making it easier for those people to stay is essential to
- 22:22SPEAKER_00: keeping not just our lead in the world.
- 22:25SPEAKER_00: I don't want to say it that way.
- 22:26SPEAKER_00: I mean, that's important.
- 22:27SPEAKER_00: I think national competitiveness is sort of underrated.
- 22:30SPEAKER_00: I think that is valuable.
- 22:34SPEAKER_00: Those people are the most productive in the US system where they can get access to
- 22:39SPEAKER_00: venture capital that's unlike any other part of the planet.
- 22:42SPEAKER_00: They can get access to networks of talent that are unavailable on other parts of the planet.
- 22:48SPEAKER_00: And so keeping them here is good for the US, but I think it's good overall for technological development.
- 22:54SPEAKER_00: And we should really, really, really focus on how to make that easier and more attractive.
- 23:00SPEAKER_01: You know, this isn't necessarily a specific AI issue.
- 23:03SPEAKER_01: But again,
- 23:04SPEAKER_01: as you said earlier,
- 23:04SPEAKER_01: it seems like a lot of the debate initially is really a holdover from sort of the
- 23:09SPEAKER_01: social media debates about moderation and bias and all that.
- 23:13SPEAKER_01: And a lot of those sorts of people in many cases
- 23:16SPEAKER_01: And sort of frameworks just kind of got glommed on to AI.
- 23:21SPEAKER_01: Another aspect is sort of like, you know, the antitrust.
- 23:24SPEAKER_01: And now, you know, worried about these big companies owning these platforms and they're biased.
- 23:29SPEAKER_01: Do we begin to look at issues,
- 23:31SPEAKER_01: how we look at our big companies who have been leading in AI,
- 23:34SPEAKER_01: doing a lot of R&D?
- 23:36SPEAKER_01: There's a politics on big tech change if we begin to see them as sort of like our
- 23:40SPEAKER_01: vanguard companies that will keep us ahead of China.
- 23:44SPEAKER_00: I think it already has, honestly.
- 23:47SPEAKER_00: You saw early on the Senate hearings around AI were totally inflected with the
- 23:54SPEAKER_00: language of social media and that network effects type of ecosystem.
- 24:02SPEAKER_00: AI does not work like that.
- 24:03SPEAKER_00: It doesn't work the same way.
- 24:05SPEAKER_00: The feedback loops are so much faster from these models.
- 24:09SPEAKER_00: We saw things like Google Gemini that had some,
- 24:12SPEAKER_00: you know,
- 24:13SPEAKER_00: a historical renderings of the founding fathers and that got.
- 24:18SPEAKER_00: So much shouting on on Twitter and on X and lots of other places that Google very
- 24:24SPEAKER_00: quickly adjusted,
- 24:25SPEAKER_00: tweaked its path.
- 24:26SPEAKER_00: So I think.
- 24:28SPEAKER_00: I think we're seeing the toning down of that rhetoric and the recognition that
- 24:33SPEAKER_00: these companies are creating a lot of powerful,
- 24:37SPEAKER_00: useful products and that they are sort of national champions.
- 24:41SPEAKER_00: I mean,
- 24:42SPEAKER_00: Trump on the campaign trail,
- 24:44SPEAKER_00: when asked about breaking up Google for an ongoing antitrust litigation,
- 24:49SPEAKER_00: was like,
- 24:50SPEAKER_00: hold on,
- 24:51SPEAKER_00: guys.
- 24:51SPEAKER_00: Breaking up these companies might not be in our best interest.
- 24:54SPEAKER_00: There might be other ways we can solve these types of problems.
- 25:01SPEAKER_00: in contrast to the Biden sort of big is bad rhetoric that they sort of leaned into entirely.
- 25:06SPEAKER_00: I think a Trump administration is going to bring more nuance to that in some ways.
- 25:11SPEAKER_00: And I do think that there will be more of a look towards our innovative companies
- 25:17SPEAKER_00: as being the vanguard of what we do in the U.S.
- 25:20SPEAKER_00: Now, having said that, obviously, I think
- 25:23SPEAKER_00: there's tons of AI development that is not inside of these largest companies in the
- 25:28SPEAKER_00: open source space,
- 25:29SPEAKER_00: and especially in the application layer,
- 25:32SPEAKER_00: building on top of some of these foundation models.
- 25:34SPEAKER_00: And so I think that ecosystem is also extremely important.
- 25:38SPEAKER_00: And so things that sort of preference the big companies over the small ones,
- 25:43SPEAKER_00: I would have a lot of concerns about.
- 25:45SPEAKER_00: And there have been
- 25:47SPEAKER_00: Regulatory regimes propose that,
- 25:50SPEAKER_00: even while opposed by maybe some of the bigger companies,
- 25:53SPEAKER_00: would certainly be possible for them to comply with in a way that small companies
- 25:57SPEAKER_00: would struggle to comply with.
- 25:59SPEAKER_00: And open source developers just don't have any sort of nexus.
- 26:04SPEAKER_00: with which to comply, since there is no actual business model that's propping that type of approach up.
- 26:12SPEAKER_00: So I'd want to keep it pretty neutral between the big companies,
- 26:17SPEAKER_00: the small companies,
- 26:17SPEAKER_00: and open source,
- 26:19SPEAKER_00: while having the sort of cultural recognition that big companies are extremely
- 26:24SPEAKER_00: valuable to the US innovation ecosystem.
- 26:28SPEAKER_01: If you had some time with,
- 26:30SPEAKER_01: I don't know,
- 26:30SPEAKER_01: somebody,
- 26:30SPEAKER_01: the president,
- 26:31SPEAKER_01: the vice president,
- 26:32SPEAKER_01: the secretary of commerce,
- 26:33SPEAKER_01: someone in an elevator going from the first to the 10th floor,
- 26:36SPEAKER_01: you had to quickly say like,
- 26:37SPEAKER_01: here's what you need to be keeping in mind about AI over the next two to four years.
- 26:44SPEAKER_01: What would you say?
- 26:45SPEAKER_00: Yeah,
- 26:46SPEAKER_00: so I think the number one thing I would say is that at the state level,
- 26:50SPEAKER_00: we're wrapping a lot of red tape around innovative companies and individuals,
- 26:57SPEAKER_00: and that we need to find a way to clear that thicket or stop it from growing any further.
- 27:04SPEAKER_00: That's the number one challenge that I see facing this.
- 27:07SPEAKER_00: Secondary to that,
- 27:08SPEAKER_00: I would say,
- 27:08SPEAKER_00: and related,
- 27:10SPEAKER_00: frankly,
- 27:11SPEAKER_00: the US government needs to figure out how to take advantage of these tools.
- 27:16SPEAKER_00: The federal government is slow to adopt new technologies,
- 27:19SPEAKER_00: but this technology has a lot of applications to the types of government work that
- 27:26SPEAKER_00: hundreds of thousands of federal employees do every day.
- 27:30SPEAKER_00: And so finding ways to streamline using AI to do the job better, I think is really valuable.
- 27:36SPEAKER_00: And I think it would be worth some investment at the federal level to think about how to do that well.
- 27:46you