Latest / Trumponomics / The Promise and Peril of Trump’s Big AI Plan

Transcript

The Promise and Peril of Trump’s Big AI Plan · Trumponomics

0:02
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.
0:08
Speaker 2: America is the country that started the AI race, and
0:12
Speaker 2: as President of the United States, I'm here today to
0:15
Speaker 2: declare that America is going to win it.
0:28
Speaker 3: I'm Stephanie Flanders, head of Governments and Economics at Bloomberg,
0:31
Speaker 3: and this is Trumppnomics, the podcast that looks at the
0:34
Speaker 3: economic world of Donald Trump, how he's already shaped the
0:37
Speaker 3: global economy. What on earth is going to happen next.
0:41
Speaker 3: One thing President Biden and President Trump and any other
0:45
Speaker 3: elected US politician can agree on is America needs to
0:49
Speaker 3: win the race for AI supremacy. You just heard Donald
0:53
Speaker 3: Trump say as much at the administration's recent AI summit,
0:57
Speaker 3: and he went on to vow to do everything it
0:59
Speaker 3: takes to achieve that. So we're asking a simple question,
1:04
Speaker 3: how's he doing. At that summit, he unveiled a twenty
1:07
Speaker 3: three page AI action Plan, which, to one expert, read
1:11
Speaker 3: like an industrial scale mobilization order. The emphasis was all
1:15
Speaker 3: on freeing up companies to move fast and innovate, getting
1:20
Speaker 3: rid of the red tape in the way of building
1:21
Speaker 3: data storage centers, junking Biden nearer guardrails on safety, and
1:26
Speaker 3: pretty much everything that might be standing in the way
1:28
Speaker 3: of big AI companies doing what they want to do.
1:31
Speaker 3: So will it work and if so, at what potential cost? Well, naturally,
1:36
Speaker 3: we have two brilliant Bloomberg voices to tell us the
1:39
Speaker 3: answer and to consider whether other Trump policies that we
1:43
Speaker 3: often discuss on this podcast might be working against the
1:46
Speaker 3: administration's goals on AI. With me in the London studio
1:51
Speaker 3: is palmey Olsen, the Bloomberg opinion columnist covering technology. She's
1:55
Speaker 3: a former reporter for The Wall Street Journal at Forbes
1:58
Speaker 3: and the author, most recently of Supremacy, AI, Chat, GPT
2:03
Speaker 3: and The Race That Will Change the World. That's the
2:05
Speaker 3: book that won the Financial Times Business Book of the
2:08
Speaker 3: Year award in twenty twenty four. Party is so great
2:10
Speaker 3: to have you on.
2:11
Speaker 4: Thank you for having me.
2:12
Speaker 3: The last time I saw you may have been at
2:13
Speaker 3: the awards ceremony. ALF very pleasant surprise for you to win.
2:17
Speaker 3: And in Washington, D C. Michael DNG joins us. Michael's
2:21
Speaker 3: the geoeconomics technology analyst for Bloomberg Economics, and he previously
2:25
Speaker 3: served in the Biden administration as a policy advisor and
2:28
Speaker 3: Presidential Management Fellow. In the Chips Program Office at the
2:31
Speaker 3: Department of Commerce, among other things. Michael, welcome to Trumponomics,
2:35
Speaker 3: your first time too.
2:37
Speaker 1: Thank you great to be.
2:38
Speaker 3: Here, Parmi, since you have just written this book and
2:46
Speaker 3: we have heard a lot about deep Seek in the
2:48
Speaker 3: last six months or so. If the President wants the
2:52
Speaker 3: US to win the race and AI, is the US
2:56
Speaker 3: still the leader in this technology or did deep Seek
2:59
Speaker 3: shows that China had already caught up?
3:02
Speaker 4: Well, it depends how you define leader. But I would
3:05
Speaker 4: say when it comes to AI capabilities, the models, and
3:09
Speaker 4: how good they are at reasoning and math and coding,
3:14
Speaker 4: then yes, US models are still ahead of the ones
3:17
Speaker 4: coming out of China. The thing is they're not ahead
3:20
Speaker 4: by much. It's a very narrow lead. Now, when it
3:23
Speaker 4: comes to actual adoption of AI services, I would say
3:27
Speaker 4: that in China it's much more sort of readily available.
3:30
Speaker 4: People use AI much more in their daily lives. And
3:33
Speaker 4: China's ability also to kind of create the energy infrastructure
3:37
Speaker 4: they need to build AI models. They're able to kind
3:40
Speaker 4: of spin up the gigawatts that they need. I think
3:42
Speaker 4: it is something like four hundred gigawatts that they're already
3:45
Speaker 4: planning for AI that they have available versus you know,
3:48
Speaker 4: dozens of gigawatts that the US has that certainly puts
3:51
Speaker 4: them in a very advantageous position. But as it stands,
3:55
Speaker 4: you know, the US is still ahead. Deep Seek of course,
3:58
Speaker 4: caused this kind of freak out for Silicon Valley back
4:01
Speaker 4: in January when that dropped, because here was a model
4:04
Speaker 4: that was not only as good as chat GPT, but
4:08
Speaker 4: it was free and the blueprints had been put on
4:11
Speaker 4: the Internet for anybody to copy and develop on, which
4:15
Speaker 4: is absolutely what people have done. And that really spoke
4:17
Speaker 4: to why China's tech sector has moved so quickly, is
4:22
Speaker 4: there's been this real focus on open source or so
4:25
Speaker 4: called open weight models. You know, when you keep your
4:28
Speaker 4: tech proprietary and secret, you kind of have to start
4:30
Speaker 4: from scratch as a company each time you build AI.
4:32
Speaker 4: But if it's open, companies are collaborating more with each other,
4:36
Speaker 4: and they're building on top of one another's innovations, and
4:39
Speaker 4: then collectively, you know, the whole industry can just move
4:42
Speaker 4: a lot more quickly. And that's exactly what's happened in China.
4:45
Speaker 4: But as I say, they haven't completely caught up, but
4:47
Speaker 4: they're close.
4:48
Speaker 3: Okay, So, Michael, we heard at the beginning there at
4:51
Speaker 3: the summit that Donald Trump at least thought his action
4:54
Speaker 3: plan was all about maintaining US leadership and winning that race.
5:00
Speaker 3: What does this administration think is the answer on this?
5:03
Speaker 4: Yeah?
5:03
Speaker 1: So, I think the AI Action Plan is interesting in
5:06
Speaker 1: that it does correctly identify some of the key areas
5:09
Speaker 1: in which the US really needs to develop further and
5:12
Speaker 1: is falling behind on China if it wants to keep
5:14
Speaker 1: up and win this AI race. One of the areas
5:17
Speaker 1: that it really clearly highlights is in power infrastructure. Parmy
5:20
Speaker 1: has already mentioned that China's being able to build on
5:23
Speaker 1: the one hundreds of gigawatts scale and is bringing on
5:25
Speaker 1: much more power than the US is capable of bringing
5:27
Speaker 1: on to be able to match the demand that data
5:29
Speaker 1: centers will acquire, and this action Plan specifically accelerates a
5:33
Speaker 1: lot of the permitting processes required to build these data centers,
5:36
Speaker 1: though interestingly it doesn't outright call for building new infrastructure.
5:40
Speaker 1: A lot of it is mainly about accelerating interconnection, optimizing grids,
5:44
Speaker 1: stabilizing grids, etc. But it's still headed in the right direction.
5:47
Speaker 1: The contradiction there, however, is that you know, last year
5:50
Speaker 1: ninety percent of new power generation capacity added to the
5:54
Speaker 1: US was all came from renewables, and the Big Beautiful
5:57
Speaker 1: Bill explicitly rolled back a lot of the clean energy
6:00
Speaker 1: tax credits that was going to support that sector of
6:02
Speaker 1: the economy. And another wrinkle in the AI action planet
6:05
Speaker 1: itself is that the aversion to the solar and wind
6:08
Speaker 1: is still very clear. In the text there, you'll notice
6:11
Speaker 1: that they highlight nuclear fission, geothermal and fusion as the
6:15
Speaker 1: only new sources of energy they think should be developed further,
6:18
Speaker 1: while excluding solar, wind, clean battery storage, et cetera. So
6:22
Speaker 1: it doesn't really make sense, especially mentioning fusion, which is
6:25
Speaker 1: not close to being technically viable at this point. On
6:27
Speaker 1: workforce itself, it does target a long standing gap in
6:30
Speaker 1: the US ecosystem in that it's trying to build up
6:33
Speaker 1: that vocational workforce to operate and deploy AI. The issue
6:37
Speaker 1: there being, however, that you know a lot of President
6:40
Speaker 1: Trump's recent actions in the past few months have targeted
6:43
Speaker 1: that high end research talent pipeline that's really critical or
6:46
Speaker 1: for advancing the frontier of AI research, universities, government funding,
6:51
Speaker 1: and also immigration. Jensen Huanga said something of you know,
6:54
Speaker 1: fifty percent of all AI researchers are Chinese, So while
6:58
Speaker 1: you're building up the vocational workforce, you're undercutting or undermining
7:02
Speaker 1: the elite research workforce. You need to really advance that frontier.
7:05
Speaker 1: And so again a mixed bag on what it's actually doing.
7:08
Speaker 1: And the last thing I'll bring up is expert controls,
7:11
Speaker 1: which is a massive topic. Lots of implications for foreign policy.
7:14
Speaker 1: How presidentrop has acted so far in the past six
7:17
Speaker 1: months towards a partners Tariffs are involved as well, so
7:19
Speaker 1: very big topic.
7:20
Speaker 3: Right. Well, you have introduced the idea that there might
7:23
Speaker 3: be some tensions within the plan, but certainly between the
7:27
Speaker 3: plan announced at the summit and other aspects of Donald
7:31
Speaker 3: Trump's policies, and we are going to get into that.
7:34
Speaker 3: But Palmi, you described this as a gift to big tech.
7:39
Speaker 3: We know that the people involved in this seem to
7:41
Speaker 3: have played a central role in devising this plan. Your
7:44
Speaker 3: book is all about the people behind AI. I suspect
7:47
Speaker 3: that you have some insight into them.
7:49
Speaker 4: Well, that's the most interesting part. I think one of
7:51
Speaker 4: the most important people behind this plan is Jensen Huang,
7:56
Speaker 4: who is the CEO of Nvidia, and Jensen see him
8:00
Speaker 4: on stage walking around with this leather jacket. He is
8:02
Speaker 4: not someone who likes doing lobbying in Washington, DC, but
8:06
Speaker 4: he's had to do that in the last few months
8:09
Speaker 4: after the Trump administration came into power. And first of all,
8:12
Speaker 4: back in April, the White House enact the ban on
8:15
Speaker 4: in Nvidia's chip sales to China. That's a huge problem
8:18
Speaker 4: for Nvidia because it said it would cost fifteen billion
8:20
Speaker 4: dollars in revenue. And so there came now this effort
8:24
Speaker 4: from Huang to try and turn that view around, and
8:29
Speaker 4: he kind of found an ally in the Trump administration
8:31
Speaker 4: in David Sachs, who is the so called crypto and
8:34
Speaker 4: Ai zar, and both of them were really pushing this
8:37
Speaker 4: idea that actually, we need to continue selling to China
8:41
Speaker 4: because if we don't, China's gonna and Huawei are gonna
8:46
Speaker 4: build their own chips, and then they're going to build
8:48
Speaker 4: their own AI services on top of that, and guess what,
8:51
Speaker 4: They're going to sell all that AI technology to other
8:53
Speaker 4: countries around the world and they're going to be the
8:55
Speaker 4: world's leading AI provider. What we need to do is
8:59
Speaker 4: kind of clutter the market in China with our own
9:01
Speaker 4: tech within Vidia chips, so that Chinese companies become reliant
9:06
Speaker 4: on our own infrastructure. And it's really interesting that this
9:10
Speaker 4: is actually fundamentally a very Silicon Valley worldview, this idea
9:14
Speaker 4: that if you kind of blit scale your product or
9:18
Speaker 4: service dominate the market. Eventually consumers, businesses, your customers get
9:24
Speaker 4: locked in into your ecosystem. And we've all kind of
9:27
Speaker 4: experienced that with Apple Life Songs and Microsoft Cloud the hyperscalers,
9:31
Speaker 4: you know, cloud services. It's actually hard for their customers.
9:34
Speaker 4: They feel they're locked in. And so this was the
9:37
Speaker 4: idea that Huang and Saks really pushed in the Trump administration. Eventually,
9:41
Speaker 4: Howard Lutnik, the Commerce Secretary, also picked up on that,
9:44
Speaker 4: and we saw that play out with this big U
9:47
Speaker 4: turn that Trump made on the ban on in Nvidia chips.
9:51
Speaker 4: He lifted it, so in Vidia is now able to
9:53
Speaker 4: continue selling specialized sort of export controlled chips called the
9:57
Speaker 4: AH twenties. They're kind of framed as being low power,
10:00
Speaker 4: not as good as the ones that American companies can use,
10:03
Speaker 4: but actually they're still pretty powerful and Chinese companies can
10:06
Speaker 4: still do a lot with them. And so I think
10:09
Speaker 4: that's raised a lot of questions in Washington among kind
10:12
Speaker 4: of traditional China hawks and the administration about is this
10:16
Speaker 4: really a good idea?
10:17
Speaker 3: And I think, Michael, you've described this as a strategic
10:20
Speaker 3: pivot from technology denial to export dominance. You know, this
10:24
Speaker 3: idea that instead of trying to keep China from getting
10:26
Speaker 3: the technology, you make everybody, the whole world, including China,
10:29
Speaker 3: absolutely dependent on it. I mean that seems to me
10:32
Speaker 3: a major shift, not just from policy a few months ago,
10:34
Speaker 3: but actually from the Biden administration's kind of instincts on
10:38
Speaker 3: this of actually trying to put controls on technology. But
10:41
Speaker 3: I know that there's different levels of export control involved here.
10:45
Speaker 3: So just talk me through it a bit, Michael. Is
10:47
Speaker 3: this a complete change of strategy or is it a
10:50
Speaker 3: bit murkier than that?
10:52
Speaker 1: Absolutely, I would characterize it as very murky in the
10:55
Speaker 1: sense that if you notice in the Air Action Plan itself,
10:59
Speaker 1: there's a spection specifically for export controls, and that section
11:02
Speaker 1: almost looks like it was written by the same Biden
11:05
Speaker 1: administration export control specialists and thinkers use alongside exactly In
11:12
Speaker 1: that one, it hints and signals at targeting component subsystems
11:16
Speaker 1: of semiconductor tools. And what that means is, if you
11:19
Speaker 1: think about the progression of export controls over the past
11:21
Speaker 1: few years. We essentially went from controlling the chips then
11:24
Speaker 1: to controlling the tools, and now the administration is thinking
11:27
Speaker 1: about controlling the components that go into those tools. So
11:30
Speaker 1: again with those components, there's a heavy reliance on the
11:32
Speaker 1: US and Western suppliers. So it is an avenue of
11:35
Speaker 1: vulnerability for China in terms of building up their domestic
11:38
Speaker 1: semiconductor manufacturing ecosystem. But this has been a direction that
11:41
Speaker 1: export controls have been going into for a while and
11:44
Speaker 1: those in the field have been seeing if this was
11:46
Speaker 1: going to happen. But it really does seem to be
11:48
Speaker 1: at odds with letting H twenty chips go to China,
11:51
Speaker 1: and now I can see maybe like a targeted strategy
11:55
Speaker 1: of will deny them access to building up their own
11:58
Speaker 1: manufacturing ecosystem while still letting them have access to chips,
12:03
Speaker 1: so a Huai competitor doesn't build up. But on the
12:06
Speaker 1: flip side, what it now means is you've enabled the
12:08
Speaker 1: entire Chinese AI ecosystem to build off of that. And
12:10
Speaker 1: if you notice what Anthropic and Open Ai have said,
12:14
Speaker 1: they've always been fairly pro export control. Anthropic in particular
12:17
Speaker 1: has gotten into it publicly with Nvidia sometimes around export controls,
12:21
Speaker 1: because they understand that if you give age twenties and
12:23
Speaker 1: advanced compute to China, you're empowering their AI companies to
12:27
Speaker 1: become more competitive, and Tom directly with deep Seek with
12:31
Speaker 1: Ali Baba with USAI companies.
12:34
Speaker 3: So Pami, I would imagine other parts of the sort
12:36
Speaker 3: of AI ecosystem are a bit concerned. I guess on
12:40
Speaker 3: two fronts that there's this quite abrupt change of strategy
12:43
Speaker 3: from the US, and also that it is in itself
12:46
Speaker 3: a bit inconsistent and seemingly unstable. We often talk about
12:50
Speaker 3: in Trumponomics about the sort of uncertainty being costly for
12:54
Speaker 3: trade and for the real economy, but I imagine it's,
12:58
Speaker 3: if anything, more costly if you're trying to think about
13:00
Speaker 3: AI investments.
13:01
Speaker 4: Absolutely, I'm sorry you said inconsistent, seemingly unstable. I feel
13:04
Speaker 4: like that should be the tagline for the Trump administration.
13:07
Speaker 4: And that's precisely like as Michael was saying. You know,
13:09
Speaker 4: when you read the action plan, the actual behavior of
13:13
Speaker 4: Trump doesn't correspond to what is being said in this plan.
13:16
Speaker 4: You know, tough talk on export controls, and yet he hasn't.
13:20
Speaker 3: On the most important one, all the biggest, most obvious.
13:23
Speaker 4: Arguably the most important one. Yeah, I think this kind
13:25
Speaker 4: of uncertainty is perhaps not great for companies. But on
13:29
Speaker 4: the other hand, in Silicon Valley there's this term accelerationist.
13:32
Speaker 4: So if you're an AI accelerationist, you want to move
13:35
Speaker 4: fast and break things. Essentially, you believe the AI development
13:38
Speaker 4: should speed up because the benefits to humanity will outweigh
13:41
Speaker 4: whatever risks and problems come up along the way. And
13:45
Speaker 4: what seems to be happening is the White House is
13:48
Speaker 4: adopting this accelerationist approach that has taken hold in Silicon Valley.
13:52
Speaker 4: You know, just even a year and a half ago,
13:55
Speaker 4: so called AI doomers we're lobbying very hard in Washington
14:00
Speaker 4: for more guardrails out of concern that AI could bring
14:03
Speaker 4: existential risk to human civilization. Those voices are now completely quiet.
14:08
Speaker 4: They've been drowned out by the accelerationists. So yes, well,
14:11
Speaker 4: the uncertainty isn't great I think for tech companies. You know,
14:14
Speaker 4: AMD is another one in Silicon Valley, anyone making hardware,
14:18
Speaker 4: anyone selling to China. This is really good news.
14:21
Speaker 3: Michael, I do want to go back to you on
14:23
Speaker 3: how some of this sits within the broader economic policies,
14:26
Speaker 3: broader trumponomics that we've seen over the last few months,
14:31
Speaker 3: and you reference some of it at the beginning. Just
14:34
Speaker 3: taught me through how important the negatives have been so
14:38
Speaker 3: far on the issues that really matter to the development
14:41
Speaker 3: of AI, and how that might sort of stack up
14:43
Speaker 3: against anything that was said at the summit.
14:46
Speaker 1: Yeah, absolutely so. I think the biggest issue with a
14:49
Speaker 1: lot of this is the Air Action Plan lays out
14:51
Speaker 1: very ambitious objectives and correctly diagnoses the areas on which
14:55
Speaker 1: the US needs to work, but the environment and the
14:58
Speaker 1: ecosystem currently that the Trump administration has created with many
15:02
Speaker 1: of their policies is not conducive to running as fast
15:05
Speaker 1: as possible on AI. I've talked about the workforce issues already.
15:09
Speaker 1: I've talked about the energy issues already. But even if
15:11
Speaker 1: we're speaking about globally, for example, on export controls, you
15:15
Speaker 1: need close collaboration with a lot of these suppliers in
15:18
Speaker 1: other countries to be able to set up a regime
15:21
Speaker 1: that actually keeps China out of these technology choke points.
15:25
Speaker 1: And the aaction plane itself have some very coercive and
15:27
Speaker 1: muscular language and essentially saying that we welcome allies to
15:31
Speaker 1: collaborate with US, and if not, we're going to use
15:33
Speaker 1: secondary tariffs the foreign direct Product rule to induce allies
15:36
Speaker 1: to participate in this export control regime. And so that
15:41
Speaker 1: was something that really surprised me when I read it.
15:43
Speaker 1: It wasn't something that the Biden administration when I was there,
15:46
Speaker 1: ever seriously considered, though when negotiations got tough it came
15:49
Speaker 1: close a few times. I will say it essentially shows
15:52
Speaker 1: this Trump administration's view that they can force US tech
15:56
Speaker 1: supremacy on the rest of the world. They're going to
15:58
Speaker 1: do that through the expert packages that they've contemplated, despite
16:01
Speaker 1: the fact that this is at the end of the day,
16:03
Speaker 1: a very global industry, a very global ecosystem, and so
16:06
Speaker 1: it's very unclear to me how exactly they're going to
16:10
Speaker 1: make the US outrun China on this a supremacy race
16:13
Speaker 1: if they're going to be so unilateral in their approach.
16:17
Speaker 3: Just a small footnote to that, and you've made the
16:20
Speaker 3: point in other analysis that you've done, if you're also
16:24
Speaker 3: making the US itself a more closed economy, it's going
16:27
Speaker 3: to be this sort of inside the Fortress America. You're
16:30
Speaker 3: going to build all this stuff, and that's going to
16:32
Speaker 3: sustain the supremacy, but you're making it harder for those
16:35
Speaker 3: crucial inputs. I remember, Michael, you highlighted just that copper tariff.
16:38
Speaker 3: Just remind us that even that one piece is going
16:41
Speaker 3: to make things harder for a certain bit of the
16:43
Speaker 3: chips industry.
16:45
Speaker 1: Yeah. Absolutely, Donald Trump in this Air Action Plan talks
16:48
Speaker 1: about building out AI infrastructure, data center semiconductor fabs. All
16:52
Speaker 1: of those require a lot of inputs from a broad
16:54
Speaker 1: imported data centers use immense amounts of copper in power
16:58
Speaker 1: and cooling. Sem enter fabs even more so, you know,
17:02
Speaker 1: copper just for power and cooling, but coppers found in
17:04
Speaker 1: ships themselves, and not just to talk about copper. Were
17:07
Speaker 1: contemplating Section two thirty two sectoral tariffs on semiconductors, and
17:10
Speaker 1: those might involve semiconductor tools themselves. All the different consumables
17:15
Speaker 1: and materials that will go into these fabs being built
17:17
Speaker 1: in the US that they need to actually produce the
17:19
Speaker 1: chips that go into AI. At the end of the day,
17:22
Speaker 1: chip manufacturing and aidentist data centers require a lot of
17:24
Speaker 1: global inputs that just doesn't square with this vision of
17:27
Speaker 1: a fortress America, this unilateral vision that Trump is all in.
17:31
Speaker 3: For Palmy, I mean, you weren't at the summit, though,
17:34
Speaker 3: I imagine that you had your spies there in one
17:36
Speaker 3: for or another. And was any of this these sort
17:39
Speaker 3: of tensions, parent contradictions voice there. I mean, even the
17:44
Speaker 3: sort of basic fact that all that he's talking about
17:47
Speaker 3: building this stuff and getting rid of red tape and
17:49
Speaker 3: know the bureaucracy at the same time as actually the
17:53
Speaker 3: important components that you really need to make this stuff
17:55
Speaker 3: as becoming more and more expensive and harder to get.
17:58
Speaker 3: Was there anybody who was saying, hang on a minute.
18:00
Speaker 4: Not that I could see. I mean, I was checking
18:02
Speaker 4: out a lot of social media. There's a there's a
18:05
Speaker 4: website called tech Meme that aggregates all the chatter about
18:08
Speaker 4: any particular topic, and everything that was going on Twitter
18:13
Speaker 4: was pretty much praising the bill, And I mean that's
18:15
Speaker 4: just a kind of cacophony of voices from Silicon Valley
18:18
Speaker 4: are just happy to see things moving head so quickly,
18:21
Speaker 4: you know, because the whole plan covers so many different things,
18:24
Speaker 4: so it's it's kind of hard to look deep into
18:27
Speaker 4: the weeds, as Michael has noted, all these very important contradictions,
18:31
Speaker 4: and so my sense is that actually people aren't taking
18:35
Speaker 4: much note of those potential limitations.
18:38
Speaker 3: And is it just because they're so lofty and big
18:41
Speaker 3: picture that they don't really think about how do you
18:43
Speaker 3: actually build lotty?
18:44
Speaker 4: They're not lofty, they're galactic. I mean, people in Silicon
18:47
Speaker 4: Valley are. They're living in the future. And that's kind
18:50
Speaker 4: of what makes the regions so successful is they don't
18:53
Speaker 4: see themselves constrained by traditional limits that might limit people
18:58
Speaker 4: and entrepreneurs in other parts of the world. There's just
19:00
Speaker 4: an incredible amount of belief that has helped propel so
19:02
Speaker 4: many of these businesses, not alongside anti competitive practice, of course,
19:06
Speaker 4: as they get bigger.
19:08
Speaker 3: I mean historically people animated by just a vision of
19:11
Speaker 3: the future. It hasn't always gone well, but aside, oh absolutely,
19:15
Speaker 3: But the I mean, just to come back on this
19:18
Speaker 3: safety point, and I just noticed, I mean, there's lots
19:20
Speaker 3: of people who are writing about it, but asimas Are,
19:22
Speaker 3: who's the British kind of blogger and tech entrepreneur who
19:25
Speaker 3: writes often about these issues, he had a sort of
19:27
Speaker 3: quite a nice line. He said, the upside of all
19:29
Speaker 3: this is speed. The downside, mainly for the rest of
19:32
Speaker 3: the world, is a regulatory vacuum primed for social, environmental,
19:37
Speaker 3: and geopolitical blowback. Do you think that's right, Oh one.
19:40
Speaker 4: But I think at the same time, even in a
19:42
Speaker 4: place like Europe, where we have the European Union's AI
19:45
Speaker 4: Act that still has a lot of problems. The act
19:48
Speaker 4: in itself is incredibly vague, It's got loopholes. I think,
19:52
Speaker 4: just very briefly, one of the biggest problems we're seeing
19:55
Speaker 4: today with AI is addiction to AI companions, people forming
20:00
Speaker 4: emotional bonds with them, getting emotionally manipulated, some people experiencing psychosis.
20:05
Speaker 4: There have been many, many stories about people getting chatchypt
20:08
Speaker 4: induced psychosis.
20:10
Speaker 3: No, you wrote a very good I'd recommend everybody in
20:12
Speaker 3: your column about it, which made me think about it.
20:14
Speaker 3: But it was also even for people who don't get
20:16
Speaker 3: psychotic episodes, you lose your critical thinking capacity.
20:19
Speaker 4: But that's because these models are kind of in much
20:22
Speaker 4: the same way social media was optimized for engagement to
20:25
Speaker 4: keep you talking, and so many of them in Chatgypt
20:28
Speaker 4: in particular, tend to validate the user, and so you
20:32
Speaker 4: have these unintended consequences that it would have been impossible
20:35
Speaker 4: to predict a year ago even and so writing regulation
20:39
Speaker 4: is really like chasing a bullet. And so, yes, we're
20:42
Speaker 4: in a regulatory vacuum, made worse by the fact now
20:45
Speaker 4: that these companies are moving so quickly and deploying so quickly,
20:49
Speaker 4: and we're really I mean, that's why I think the
20:51
Speaker 4: UK really needs to step up its own regulatory efforts here,
20:55
Speaker 4: because I think the UK is a great tech regulator.
20:57
Speaker 3: Well, I know that there were some of these basic
20:59
Speaker 3: issues they've really struggled with, including AI safety. So I
21:02
Speaker 3: wouldn't hold your breath, but hope, maybe Michael was one
21:06
Speaker 3: thing I wanted to mention that we haven't touched on.
21:08
Speaker 3: And then we're going to run out of time. We've
21:09
Speaker 3: said all the way through. The trade off that's been
21:11
Speaker 3: made with this action plan is over sort of speed
21:15
Speaker 3: over safety and the junking of the sort of a
21:17
Speaker 3: lot of the guardrails around ethics and AI safety that
21:22
Speaker 3: President Biden at least tried to begin to develop, But
21:25
Speaker 3: there's one area where they are still focused on safety.
21:28
Speaker 3: They want to keep us safe from being forced to
21:30
Speaker 3: think about climate change and from the woke mind virus.
21:35
Speaker 3: So just talk us through that, because there's a lot
21:36
Speaker 3: in here which is actually about getting that out of AI.
21:40
Speaker 1: Yeah. So I remember reading that section and trying to
21:44
Speaker 1: understand how exactly they were going to implement it, and
21:48
Speaker 1: the executive order also just came out that still doesn't
21:51
Speaker 1: go into much detail about it. I think this fundamentally
21:54
Speaker 1: is entirely a political move to, you know, highlight a
21:58
Speaker 1: issue that I think with an AI is a major
22:02
Speaker 1: area of concern.
22:03
Speaker 3: Palmi, I guess just to give you the last word,
22:06
Speaker 3: and I may be miss characterizing, but one of the
22:09
Speaker 3: columns that you wrote, I think immediately after Deep Seat
22:12
Speaker 3: came out or soon afterwards, and you're sort of reflecting
22:15
Speaker 3: on how significant that was and changing our view of
22:18
Speaker 3: how the AI race might turn out. And you suggested
22:23
Speaker 3: that actually the AI not being US dominated and being
22:28
Speaker 3: more open source or at least open weight, might be
22:31
Speaker 3: healthier for the world. Do you still think that, And
22:33
Speaker 3: I guess there's a risk that you know, China could
22:35
Speaker 3: have won, it might be China dominated.
22:37
Speaker 4: Yeah, And I think I don't agree with the kind
22:40
Speaker 4: of state controlled aspect of Chinese AI. But I think
22:44
Speaker 4: the best thing about Deep seek was that it raised
22:47
Speaker 4: the specter of more competition, and it raised this notion
22:50
Speaker 4: that maybe the Western Silicon Valley companies weren't the guaranteed
22:54
Speaker 4: winners of this AI race. Here were these other companies
22:58
Speaker 4: coming out of China that were cheap, that we're free,
23:01
Speaker 4: and that we're just as good. This isn't really an
23:03
Speaker 4: issue about American companies. This is more about companies that
23:07
Speaker 4: dominate the market, as we've seen already with cloud companies
23:11
Speaker 4: from Silicon Valley e commerce companies. You know, it's almost
23:14
Speaker 4: impossible now to open an e commerce business because of Amazon,
23:17
Speaker 4: and so many parts of the wider market have absolutely
23:20
Speaker 4: been warped by the anti competitive and monopolistic behavior of
23:24
Speaker 4: big tech. And what would be terrible would to just
23:28
Speaker 4: see that continue. With AI and AI being so transformative,
23:33
Speaker 4: so emotionally powerful, cognitively powerful, it can influence how people
23:38
Speaker 4: think about things. People are using it to make daily
23:41
Speaker 4: life decisions and write their own emails and reports. That
23:45
Speaker 4: leaves an incredible amount of power in the hands of
23:47
Speaker 4: just a few companies who are building these models. And
23:50
Speaker 4: so if there are more on the market, I think
23:52
Speaker 4: that's a great thing. I think there needs to be
23:55
Speaker 4: more competition. Whether that will actually continue to play out
23:58
Speaker 4: as it's still playing out now, I think is an
24:01
Speaker 4: open question.
24:03
Speaker 3: Well, it's still playing out. There's some very wise words
24:06
Speaker 3: from both of you, some of which I suspect the
24:08
Speaker 3: President would think we're bordering on trees, treacherous, oh the
24:12
Speaker 3: prospect of you know, non American leadership. But thank you
24:16
Speaker 3: very much, both of you. Thank you, thank you, thanks
24:26
Speaker 3: for listening to Trumponomics from Bloomberg. It was hosted by
24:28
Speaker 3: me Stephanie Flanders, and I was joined by Bloomberg's Pame
24:31
Speaker 3: Olsen and Michael Deng. Trumponomics is produced by Samasadi and
24:37
Speaker 3: Moses and with help from Amy Keene and special thanks
24:41
Speaker 3: this week to Nervy Shavin. Sound design is by Blake
24:44
Speaker 3: Maples and Sage Bowman is the head of Bloomberg Podcast.
24:48
Speaker 3: Please help others find it and enjoy it by rating
24:51
Speaker 3: and reviewing Trumponomics wherever you listened.