WEBVTT - The Future of Tech Governance Around the Global

0:00:02.520 --> 0:00:09.800
<v Speaker 1>Bloomberg Audio Studios, Podcasts, radio news. This is Bloomberg Business

0:00:09.880 --> 0:00:14.080
<v Speaker 1>Week with Carol Masser and Tim Steneveek on Bloomberg Radio.

0:00:14.480 --> 0:00:17.440
<v Speaker 2>My AI detector is better than my colleagues because they

0:00:17.440 --> 0:00:20.439
<v Speaker 2>were fooled by a video a couple days ago, and

0:00:20.480 --> 0:00:21.480
<v Speaker 2>I was like, that's AI.

0:00:21.960 --> 0:00:23.680
<v Speaker 3>Okay, Well, that's exactly where I want to go with

0:00:23.720 --> 0:00:26.880
<v Speaker 3>our next guest. DONW. Bradford is Henry el Moses, Professor

0:00:26.880 --> 0:00:29.960
<v Speaker 3>of Law, an international organization at Columbia Law School. She

0:00:30.120 --> 0:00:33.160
<v Speaker 3>was last on with us just over two years ago

0:00:33.960 --> 0:00:37.280
<v Speaker 3>when the Will Smith eating spaghetti test looked not realistic

0:00:37.360 --> 0:00:39.960
<v Speaker 3>at all, just to show how far things have come.

0:00:40.040 --> 0:00:43.440
<v Speaker 3>She's the author of several books, including most recently, Digital Empires,

0:00:43.720 --> 0:00:47.320
<v Speaker 3>The Global Battle to Regulate Technology. This was published back

0:00:47.600 --> 0:00:50.159
<v Speaker 3>in twenty twenty three. And that's exactly what I want

0:00:50.200 --> 0:00:52.640
<v Speaker 3>to start where I want to start with you, Professor Bradford.

0:00:53.120 --> 0:00:55.440
<v Speaker 3>I got an invite today to try out Sora from

0:00:55.480 --> 0:00:58.760
<v Speaker 3>a friend of mine. I think he's a paying subscriber

0:00:58.800 --> 0:01:02.040
<v Speaker 3>of open ais GPT and that's why he has an invitation.

0:01:02.520 --> 0:01:03.880
<v Speaker 3>And I was watching some of the videos that he

0:01:03.960 --> 0:01:07.720
<v Speaker 3>made and I'm thinking to myself, we are so cooked

0:01:08.800 --> 0:01:09.319
<v Speaker 3>are we.

0:01:12.000 --> 0:01:14.400
<v Speaker 4>So? Thanks so much for having me, Tim and Emily.

0:01:15.400 --> 0:01:18.440
<v Speaker 4>We may well be cooked. But the question is what

0:01:18.720 --> 0:01:20.920
<v Speaker 4>are we most worried about. I think there are many

0:01:20.959 --> 0:01:24.720
<v Speaker 4>exciting developments. I think we're certainly more entertained in many ways.

0:01:25.040 --> 0:01:27.839
<v Speaker 4>But at the same time, I think those who warned

0:01:27.920 --> 0:01:31.240
<v Speaker 4>us early on that this is not the kind of

0:01:31.280 --> 0:01:34.520
<v Speaker 4>AI revolution that we should be just leaving for the

0:01:34.600 --> 0:01:39.200
<v Speaker 4>tech companies to govern to manage. We do need governments involved.

0:01:39.200 --> 0:01:41.840
<v Speaker 4>We need some guardrails, we need some regulation to make

0:01:41.880 --> 0:01:45.440
<v Speaker 4>sure that these fast advances that we are witnessing are

0:01:45.520 --> 0:01:47.600
<v Speaker 4>moving to the direction that we're comfortable with.

0:01:47.960 --> 0:01:49.680
<v Speaker 3>It seems like it's too late, though.

0:01:51.360 --> 0:01:53.600
<v Speaker 4>I don't think it's too late. I think we certainly

0:01:53.680 --> 0:01:56.400
<v Speaker 4>are not at the point where we can say that

0:01:56.480 --> 0:01:59.520
<v Speaker 4>AI is done. I think we will continue to see

0:01:59.600 --> 0:02:04.800
<v Speaker 4>massive developments in coming years. And we already have AI

0:02:04.920 --> 0:02:09.519
<v Speaker 4>governance regulation on the rada of many lawmakers. And obviously

0:02:09.560 --> 0:02:13.720
<v Speaker 4>the Europeans have been most proactive, as they usually are,

0:02:14.200 --> 0:02:18.480
<v Speaker 4>and they have the AI Act, a comprehensive piece of

0:02:18.600 --> 0:02:21.959
<v Speaker 4>legislation that is already in force. And now it's then

0:02:22.040 --> 0:02:25.800
<v Speaker 4>the matter of implementing it effectively, and then we also

0:02:25.880 --> 0:02:28.240
<v Speaker 4>need to see what happens in the United States at

0:02:28.280 --> 0:02:32.440
<v Speaker 4>the state level. China is definitely interested in governing AI.

0:02:32.680 --> 0:02:35.480
<v Speaker 4>We have many other jurisdictions, so I think there is

0:02:35.600 --> 0:02:38.280
<v Speaker 4>a lot that is happening and a lot more that

0:02:38.320 --> 0:02:39.240
<v Speaker 4>needs to be done.

0:02:39.560 --> 0:02:44.959
<v Speaker 2>In your new book Digital Empires, you break apart how

0:02:45.200 --> 0:02:51.600
<v Speaker 2>basically different regimes across the globe are governing and regulating

0:02:51.639 --> 0:02:55.840
<v Speaker 2>AI differently. So there's a US, there's Europe, there's China.

0:02:56.480 --> 0:03:00.919
<v Speaker 2>In your research, have you found one need is kind

0:03:00.919 --> 0:03:05.000
<v Speaker 2>of doing that balancing act the best so far in

0:03:05.080 --> 0:03:07.240
<v Speaker 2>terms of balancing out They don't want to stifle innovation,

0:03:07.760 --> 0:03:11.640
<v Speaker 2>but they also want to protect users of AI, consumers

0:03:11.639 --> 0:03:16.160
<v Speaker 2>of AI, companies that are getting involved with AI. Who's

0:03:16.200 --> 0:03:18.239
<v Speaker 2>doing the balancing act the best in your view?

0:03:19.639 --> 0:03:22.600
<v Speaker 4>So I think any regulator really needs to think about

0:03:22.639 --> 0:03:26.320
<v Speaker 4>this balancing to make sure that we harness these tremendous

0:03:26.760 --> 0:03:30.799
<v Speaker 4>benefits that are associated with AI, but also really safeguard

0:03:30.880 --> 0:03:35.720
<v Speaker 4>our citizens and societies from various risks. And in many ways,

0:03:36.320 --> 0:03:39.560
<v Speaker 4>there is a perception that the Europeans are airing on

0:03:39.640 --> 0:03:45.480
<v Speaker 4>the side or preemptively protecting against these risks and maybe

0:03:45.520 --> 0:03:49.040
<v Speaker 4>then foregoing some of the innovation benefits, whereas the Americans

0:03:49.080 --> 0:03:52.200
<v Speaker 4>would be airing on the side of maybe being very

0:03:52.320 --> 0:03:56.280
<v Speaker 4>techno optimists and not thinking about all those potential downsides.

0:03:57.600 --> 0:04:01.360
<v Speaker 4>I think in many ways I do like and endorse

0:04:01.480 --> 0:04:04.800
<v Speaker 4>the European model in the sense that, in my view,

0:04:04.840 --> 0:04:10.120
<v Speaker 4>that best safeguards the public interest and really takes seriously

0:04:10.200 --> 0:04:14.960
<v Speaker 4>the fundamental rights of individuals and democratic structures of the society.

0:04:15.560 --> 0:04:18.080
<v Speaker 4>But there is always I really reject this notion that

0:04:18.120 --> 0:04:21.960
<v Speaker 4>this comes at the cost of innovation. There definitely is

0:04:22.040 --> 0:04:25.320
<v Speaker 4>a gap where the Americans are doing much better in

0:04:25.440 --> 0:04:29.560
<v Speaker 4>generating AI innovations compared to the Europeans. But the reason

0:04:29.680 --> 0:04:32.400
<v Speaker 4>is not that the Europeans are so keen on regulating.

0:04:32.839 --> 0:04:35.760
<v Speaker 4>I think there are many other reasons that explain why.

0:04:35.800 --> 0:04:38.719
<v Speaker 4>There are just fundamental pillars of the tech ecosystem in

0:04:38.720 --> 0:04:41.560
<v Speaker 4>the US that are much stronger and the Europeans have

0:04:41.640 --> 0:04:46.640
<v Speaker 4>fallen short in replicating that. So regulation as such, the

0:04:46.680 --> 0:04:51.640
<v Speaker 4>protection of those rights is not a choice that needs

0:04:51.680 --> 0:04:55.120
<v Speaker 4>to come at the cost of making beneficial progress in

0:04:55.200 --> 0:04:55.640
<v Speaker 4>this space.

0:04:56.760 --> 0:05:01.160
<v Speaker 3>So, you know, going back to our conversation that I

0:05:01.200 --> 0:05:03.200
<v Speaker 3>had with you two years ago, because the world has

0:05:03.279 --> 0:05:08.240
<v Speaker 3>changed so much since then. Two years ago, Joe Biden

0:05:08.320 --> 0:05:10.920
<v Speaker 3>was president, there were a lot of folks who didn't

0:05:10.960 --> 0:05:15.839
<v Speaker 3>think that Donald Trump would win another term. Fast forward

0:05:15.880 --> 0:05:19.640
<v Speaker 3>two years, Donald Trump is the president, David Sachs is

0:05:19.720 --> 0:05:25.400
<v Speaker 3>crypto and aizar. The regime is thinks about this completely

0:05:25.480 --> 0:05:28.880
<v Speaker 3>differently than I think it's fair to say the Biden administration.

0:05:30.360 --> 0:05:32.440
<v Speaker 3>What do you think the US needs to be doing

0:05:32.520 --> 0:05:36.080
<v Speaker 3>right now to regulate this technology? What would you like

0:05:36.120 --> 0:05:37.159
<v Speaker 3>to see David Sachs do?

0:05:38.320 --> 0:05:41.560
<v Speaker 4>Yeah, so you're so right. There has been a complete

0:05:41.680 --> 0:05:43.840
<v Speaker 4>U turn in many ways. Towards the end of the

0:05:43.839 --> 0:05:50.120
<v Speaker 4>Biden administration, there was closer alignment between the traditional Transatlantic allies,

0:05:50.200 --> 0:05:53.520
<v Speaker 4>where the US was really moving closer to the European

0:05:53.640 --> 0:05:57.760
<v Speaker 4>view that technology like AI needs guardrails, and there was

0:05:57.800 --> 0:06:01.479
<v Speaker 4>a genuine attempt to join force among the world's techno

0:06:01.480 --> 0:06:05.920
<v Speaker 4>democracies in order to halt the advances of Chinese digital

0:06:05.960 --> 0:06:10.320
<v Speaker 4>authoritarian views of governing technology. So I really saw this

0:06:10.839 --> 0:06:13.919
<v Speaker 4>potential for the US and the EU to join forces

0:06:13.960 --> 0:06:17.160
<v Speaker 4>to bring about a very beneficial chase in this space.

0:06:17.640 --> 0:06:19.839
<v Speaker 4>But now the US is doing I think two things.

0:06:19.880 --> 0:06:22.640
<v Speaker 4>So it's first of all, giving a lot more power

0:06:22.680 --> 0:06:27.159
<v Speaker 4>to the tech companies walking away from regulation, impraising these

0:06:27.520 --> 0:06:33.520
<v Speaker 4>deregulatory zeal That really reflects the very strong form of

0:06:33.600 --> 0:06:37.520
<v Speaker 4>techno libertarian, techno optimist firl wheel. But in many ways,

0:06:37.520 --> 0:06:41.599
<v Speaker 4>the US is also playing Beijing's game and becoming very

0:06:41.640 --> 0:06:47.240
<v Speaker 4>state driven. We see our massive state investment in some

0:06:47.279 --> 0:06:51.440
<v Speaker 4>of these leading tech companies. We see our export controls,

0:06:51.520 --> 0:06:56.839
<v Speaker 4>investment restrictions, we see subsidies. So the US is to

0:06:56.960 --> 0:07:01.080
<v Speaker 4>me are losing some of its own And if you

0:07:01.120 --> 0:07:04.520
<v Speaker 4>think about how that will also impact the US's adamant

0:07:04.560 --> 0:07:07.159
<v Speaker 4>goal of being a leader in AI, what is happening

0:07:07.160 --> 0:07:10.520
<v Speaker 4>in the space of immigration, I think that is really

0:07:10.600 --> 0:07:14.920
<v Speaker 4>counterproductive if you think about where all those AI innovation comes,

0:07:15.120 --> 0:07:17.480
<v Speaker 4>innovations come from the US. So what the US would

0:07:17.520 --> 0:07:21.240
<v Speaker 4>need to do first? The US would need to regulate

0:07:21.280 --> 0:07:24.000
<v Speaker 4>this space. We need to make sure that fundamental rights

0:07:24.040 --> 0:07:27.400
<v Speaker 4>are protected, we need to make sure that those societal

0:07:27.600 --> 0:07:31.080
<v Speaker 4>risks are under control, and we need to also at

0:07:31.120 --> 0:07:33.440
<v Speaker 4>the same time make sure that we will continue to

0:07:33.480 --> 0:07:36.360
<v Speaker 4>invest in the development of the AI by retaining the

0:07:36.400 --> 0:07:40.720
<v Speaker 4>world's best talent, which often is immigrant talent, including then

0:07:40.840 --> 0:07:45.240
<v Speaker 4>Chinese data scientists who have been contributing to advances in

0:07:45.280 --> 0:07:46.440
<v Speaker 4>the space in the US.

0:07:46.800 --> 0:07:51.640
<v Speaker 2>Is there any specific regulation that comes to mind that

0:07:51.760 --> 0:07:56.480
<v Speaker 2>you would want to see in the US that would

0:07:56.840 --> 0:08:00.720
<v Speaker 2>prevent this kind of idea that ten percent in the

0:08:00.760 --> 0:08:05.080
<v Speaker 2>beginning of the segment about this the spaghetti test. It

0:08:05.160 --> 0:08:07.440
<v Speaker 2>sounds silly, the will Smith spaghetti test, but it gets

0:08:07.480 --> 0:08:12.360
<v Speaker 2>at the heart. Yeah, concerned that people have that suddenly

0:08:12.480 --> 0:08:16.160
<v Speaker 2>the Internet is going to be, you know, filled with

0:08:16.200 --> 0:08:16.880
<v Speaker 2>these videos.

0:08:16.960 --> 0:08:21.280
<v Speaker 3>We're not and it's the end.

0:08:21.400 --> 0:08:24.800
<v Speaker 2>I'm sorry, Well, maybe we can have a professor help.

0:08:25.440 --> 0:08:25.760
<v Speaker 4>She said.

0:08:26.080 --> 0:08:28.840
<v Speaker 3>She said, it's not the end already, which I'm great, right, it's.

0:08:28.680 --> 0:08:29.280
<v Speaker 1>Not the end.

0:08:29.960 --> 0:08:32.319
<v Speaker 2>Is there a specific piece of regulation that that comes

0:08:32.320 --> 0:08:35.200
<v Speaker 2>to mind? Is it about I don't know, digital privacy,

0:08:35.280 --> 0:08:38.760
<v Speaker 2>people needing disclaimers on top of every video that you

0:08:39.080 --> 0:08:40.040
<v Speaker 2>see on the Internet.

0:08:41.800 --> 0:08:44.360
<v Speaker 4>So I think there are many aspects, and there's no

0:08:44.520 --> 0:08:46.319
<v Speaker 4>easy way to say that you just need to do

0:08:46.360 --> 0:08:49.520
<v Speaker 4>one thing in order to then address this multitude of

0:08:49.559 --> 0:08:53.880
<v Speaker 4>different hams. But one thing. It does start from the

0:08:53.920 --> 0:08:57.360
<v Speaker 4>protection of privacy and our agency and our ability to

0:08:57.400 --> 0:09:00.200
<v Speaker 4>be able to tell what is fiction and what is not,

0:09:00.920 --> 0:09:04.200
<v Speaker 4>and our ability to engage in conversation based on real

0:09:04.240 --> 0:09:09.640
<v Speaker 4>information that is not manipulated by our AI and this

0:09:09.800 --> 0:09:16.400
<v Speaker 4>information obviously existed even without JUDGPT type of tools, but

0:09:16.679 --> 0:09:22.760
<v Speaker 4>they are now fueled with this AI driven ability to

0:09:22.920 --> 0:09:26.240
<v Speaker 4>manipulate our sense of reality. So in many ways, I

0:09:26.280 --> 0:09:29.160
<v Speaker 4>think it does need labeling. It does need the kind

0:09:29.200 --> 0:09:33.560
<v Speaker 4>of transparency and accountability that we have a sense of

0:09:34.240 --> 0:09:37.440
<v Speaker 4>how these AI systems are built and how we engage

0:09:37.480 --> 0:09:40.959
<v Speaker 4>with them. But then there are also risks around and

0:09:41.160 --> 0:09:44.480
<v Speaker 4>just protecting content content creators. We need to take copyright

0:09:44.600 --> 0:09:47.760
<v Speaker 4>seriously and the idea of how we actually train these

0:09:47.840 --> 0:09:51.960
<v Speaker 4>models with the data that has been generated by individual authors,

0:09:52.000 --> 0:09:55.679
<v Speaker 4>by journalists, and that needs to be also compensated well

0:09:55.800 --> 0:09:58.520
<v Speaker 4>so that we still have the incentive to engage in

0:09:58.600 --> 0:10:02.400
<v Speaker 4>that kind of content production. But privacy is obviously very

0:10:02.480 --> 0:10:05.320
<v Speaker 4>high on my list, This information is very high on

0:10:05.360 --> 0:10:08.480
<v Speaker 4>my list. Then there are questions that are more about

0:10:08.559 --> 0:10:13.320
<v Speaker 4>existential risks, more about systemic risks, and even if it's

0:10:13.360 --> 0:10:16.880
<v Speaker 4>hard to sometimes know the probabilities of some of the

0:10:16.920 --> 0:10:22.360
<v Speaker 4>most severe risks and how likely they are to materialize,

0:10:22.480 --> 0:10:25.800
<v Speaker 4>we still need to be prepared to also as a

0:10:25.840 --> 0:10:29.520
<v Speaker 4>society to confront that kind of reality when AI advance

0:10:29.679 --> 0:10:32.840
<v Speaker 4>is really fast and we reach the point when we

0:10:32.880 --> 0:10:35.640
<v Speaker 4>find even harder to govern that technology. So I think

0:10:35.640 --> 0:10:38.720
<v Speaker 4>there are all these layers and we are not even

0:10:38.800 --> 0:10:41.400
<v Speaker 4>really having, at least at the better level, a real

0:10:41.480 --> 0:10:44.040
<v Speaker 4>conversation in how we go about regulating this space.

0:10:44.600 --> 0:10:47.560
<v Speaker 3>Professor ONW brad for Henry el Moses, Professor of Law

0:10:47.559 --> 0:10:50.520
<v Speaker 3>and an International Organization at Columbia Law School. She's the author

0:10:50.559 --> 0:10:53.600
<v Speaker 3>of several books, including her most recent, Digital Empires, The

0:10:53.640 --> 0:10:57.319
<v Speaker 3>Global Battle to Regulate Technology, published back in twenty twenty three,

0:10:57.360 --> 0:11:00.480
<v Speaker 3>but as relevant right now as it was to years ago.

0:11:03.480 --> 0:11:03.520
<v Speaker 3>M