Loading summary
Transcript48 lines
- [00:00]
Unknown Host
Are you feeling more fulfilled now that you're back to work this Friday?
- [00:05]
Unknown Co-Host
No, I need a vacation.
- [00:07]
Unknown Host
See the movie that critics are saying is an awesome look at that crowd pleasing, fist pumping all out brawl of a film.
- [00:17]
Unknown Co-Host
You're right about that.
- [00:20]
Dean Ball
They're coming after our family.
- [00:23]
Unknown Host
Go fix this.
- [00:24]
Unknown Co-Host
Oh my.
- [00:26]
Unknown Host
Nobody 2. Rated R. Only in theaters Friday.
- [00:34]
Stephen Overle
Hey, welcome back to Politico Tech. I'm Stephen Overle and on this show I break down tech, politics and policy with the people shaping our digital future. In recent weeks, we've talked a lot about President Donald Trump's AI action plan. It was released last month and basically provides a blueprint for how this administration will bring its American dominance agenda to artificial intelligence. The main author of that plan was Dean Ball, a senior policy advisor in the Office of Science and Technology Policy. At least that was his job until this week. Dean has now joined the think tank foundation for American Innovation and relaunched his AI newsletter called Hyperdimensional. I called him up to find out why he's leaving government now, what he thinks of Trump allowing China to buy American microchips, and whether the current flood of AI investments is causing a market bubble. Here's our conversation. Dean, welcome to Politico Tech.
- [01:48]
Dean Ball
Thanks so much for having me.
- [01:50]
Unknown Co-Host
So you left the White House just this week. So let's start with that decision. I saw that you wrote the private sector is what will kind of lead on governing AI, not the federal government. And so going back into private sector is kind of where you can have the most influence. How do you see that division of power going forward?
- [02:09]
Dean Ball
It's an interesting question. So I think that government is absolutely an incredibly high leverage place to do lots of important work on AI. At the same time, I do believe that a lot of both technical and governance standards fundamentally will be driven by the private sector. I think that my primary value add to the world is probably as someone who can synthesize lots of different things on some complex issue, optimize for different political considerations, and then communicate an effective response and sort of effective policy solution in a strategic way. That's what the job with the action plan was. That's a very different job from implementing the action plan.
- [02:55]
Unknown Co-Host
Right.
- [02:55]
Dean Ball
And I think there I am. I wouldn't say like, oh, I'd be terrible at this, but I would say there's a lot of people who could do a better job than me at things like that. So that was kind of the nut of why I made the decision that I did.
- [03:10]
Unknown Co-Host
Looking ahead, I guess, you know, when you think about AI and where we go from here in terms of whether that's setting ground rules or facilitating the development of what is the role for government as you see it? Because it seemed when AI first was really kind of coming into Washington's consciousness, there was a sense that government would be much more involved and now we've shifted from that.
- [03:32]
Dean Ball
Yeah. So I think government has a number of important roles. First of all, yes, it is in many ways getting out of the way. But also getting out of the way does not, is not so simple as just deregulating everything because the reality is that there are going to be a lot of quite complex industries or other sectors for a variety of different reasons, sometimes just because they have lots of built up regulation already, like financial services, where what you need is re regulation as opposed to deregulation. And I'm not thinking so much of the AI systems themselves, but rather of the rules that, that govern the use of those AI systems in certain applications. So I think government will play a quite profound role there. Of course, government has a role in sort of codifying industry led consensus standards. This is things like what NIST does all day long.
- [04:29]
Unknown Co-Host
Right.
- [04:30]
Dean Ball
I think that will be enormously important. And then also there is a kind of diplomatic role, both ensuring that the export controls that we choose, which is, that's a political question, but once the export controls are settled on, we have to make sure that they're enforced. And that often involves working with other countries. And there's a diplomatic role in ensuring that American AI is kind of the worldwide global standard. And so I think all those things and more, I mean, you know, the action plan in general articulates many things where we take, you know, in the administration, certainly they're taking a light touch approach to these things. And yet the action plan is full of things that government can do. So the two ideas are compatible.
- [05:17]
Unknown Co-Host
Let's unpack the action plan because that was the big work that you, you handled when you were inside the Trump administration. The North Star of that plan really seemed to be kind of beating China. Right. American dominance in AI and China is sort of the chief rival there. Maybe you disagree on that and would be curious to hear your take. I do wonder if that for you was like the big driver behind the action plan.
- [05:42]
Dean Ball
I would say certainly China is a major competitor in AI and certainly we want to achieve dominance. But no, I don't get up in the morning every day and I didn't while I was at the Office of Science and Technology Policy get up in the morning every day dreaming of American dominance over the PRC in particular, I think that there are a lot of competing ways of thinking about AI in the world today. And some of those, I think are quite flawed, quite dangerous, and are coming from not just the European Union, but in fact, many countries around the world. And I think this preemptive regulatory posture that much of the world has assumed is potentially quite dangerous. And I think it's just as important to show the world that it's possible to lead and thrive with AI in a very different way.
- [06:39]
Stephen Overle
Interesting.
- [06:39]
Unknown Co-Host
So it's not just about competing with China on the development of the technology. It's also about American dominance in the regulatory landscape and some of these jurisdictions, like the eu, that have been very proactive in regulating the tech.
- [06:54]
Dean Ball
Yeah. My view is that, you know, the standards for global technology today do not get set by people like me sitting around conference tables in the Eisenhower Building. You know, the reason that, you know, all smartphones in the world look the way they do is not because people in at nist in Gaithersburg, Maryland, or in the Commerce Department in D.C. or at the White House sat around a table and designed, you know, this is what the specs for the iPhone should be. This is the standard. Right?
- [07:23]
Unknown Co-Host
Yeah.
- [07:23]
Dean Ball
No, that happened because an American company won on product. And fundamentally, I think that most of. Not most, but many of the most important AI standards will just fundamentally be set in similar ways by American institutions and American products and American companies ways of doing things, those things winning around the world, not because some law said so, but. But because it's the obvious thing that everybody wants to do.
- [07:57]
Unknown Co-Host
You mentioned export controls. And so obviously I want to ask about that. The AI action plan, it does call for tougher and tighter export controls. That is sort of one of the ways in which not just this administration, but the last administration have kind of approached competition in tech more broadly. We've heard in recent days, obviously, that administration is allowing or planning to allow Nvidia and AMD to sell chips to China in exchange for a cut of the money. Is that arrangement in line with this action plan?
- [08:30]
Dean Ball
Yeah, I think it absolutely is. And I'll tell you. I'll tell you why a lot of people have asked me this question both before I left government and now that I'm not in, hoping that I'll say something juicy. But the reality is that the action plan doesn't really say very much about what the substance of AI compute export control should be. What it says is that regardless of what the export controls are, they have to be enforced in an even but robust way all over the world. And in addition to that, it does make some specific in terms of where it gets into substance of export controls that is primarily related to semiconductor manufacturing equipment export controls. And I think there's an entirely defensible and coherent strategy one can have where you say we will sell chips to China, we will sell the advanced AI compute in some quantity and maybe not the most cutting edge stuff, but we're going to sell something. And where we're going to double down is on ensuring, on preventing China from developing their own domestic semiconductor manufacturing capability. And ultimately those decisions about the export controls are made well above my pay grade from when I was in government.
- [09:52]
Unknown Co-Host
Sure.
- [09:53]
Dean Ball
So I had very little to no role in the making of those decisions in particular. But I would say the action plan was written with the idea of giving the President and other senior policymakers sort of significant flexibility. It's not supposed to tie us down to specific things. We want there to be flexibility on these things. And so it's more of a, on the export control stuff, I'd say the action plan is much more of a framework for how to implement them rather than missive about what the substance of those should be.
- [10:26]
Unknown Co-Host
Gotcha. And I do take your point that there is a difference between selling the equipment to China to make chips and selling them the finished chips. I guess though, the end result is ultimately China having these chips that they need to develop AI. And so doesn't this just sort of fuel the AI machine of China that we are competing with?
- [10:47]
Dean Ball
Ultimately it depends on what you view as sort of the fundamental objective here. Right. One objective might be, you know, we want to prevent China from developing an AI industry at all or an AI ecosystem at all. Another might be we understand that they're going to do that either way. And instead we want to sort of try to ensure that China develops its AI ecosystem on our compute stack. And the other way to put it might be like, well, what we really care the actual threat in the long term, the actual threat is China developing domestic semiconductor manufacturing capacity. Because at that point you will need to coordinate worldwide tariffs. Because we know what the playbook is. We know the playbook is indigenize the tech. The Chinese playbook is indigenize the technology, sell at subsidized prices, take all the profit out of the market, thereby run western competitors out of business. I don't think that's a strategy that's going to be very good for the world in the long term. But for sure it's not going to be good for lots of American and other Western countries as to like what the specific goals are behind the latest moves. You know, again, I was not involved in formulating those policies particularly so hard for me to say.
- [12:20]
Lifelock Sponsor
This episode is brought to you by Lifelock. When you visit the doctor, you probably hand over your insurance, your ID and contact details. It's just one of the many places that has your personal info and if any of them accidentally expose it, you could be at risk for identity theft. LifeLock monitors millions of data points a second. If you become a victim, they'll fix it, guaranteed or your money back. Save up to 40% your first year@lifelock.com podcast terms apply.
- [12:53]
Unknown Co-Host
You know, one thing I'm just curious, I did not see in the AI action plan, I wonder if it was ever contemplated, was anything around copyright? You know, this is an issue that looms with AI. There's a number of big court cases happening right now that could set some precedent in this space about what is protected and whether artists for instance, should be compensated. Was it a deliberate decision not to weigh in on that issue?
- [13:16]
Dean Ball
It did definitely occur to us. Of course. It's a very prominent issue. Lots of people met with us about it. We received about 10,000 comments from the public and I believe by a pretty significant margin, copyright was the plurality most cited issue of all the issues in AI policy.
- [13:32]
Unknown Co-Host
That's interesting.
- [13:33]
Dean Ball
And certainly I would not say that copyright is something that it's an issue that many folks in the administration care about and that included me while I was there. I would say there was never a deliberate decision not to include copyright per se. Instead, when you have a blank sheet of paper and such a capacious topic, all good creative work is done within the bounds of constraints. So you have to constrain yourself from the get go in some ways because otherwise you'll just produce something mealy mouthed. And so one of the constraints that we chose very early on was to focus on executive actions only. And so there's not that much that the executive branch, there aren't that many levers that the executive branch directly can pull with respect to copyright pre training issues. Now IP is mentioned in terms of lab IP protecting lab intellectual property, this is to say in particular foreign adversary, corporate espionage style stuff. But there were some other issues that were bouncing around in my mind that just sort of were left on the cutting room floor, not related necessarily to pre training fair use as a question. But for example, there are some refinements I believe one could make to the ways in which AI assisted patents are granted. That's an issue that I worked on a little bit, ultimately decided wasn't anything against the issue. You just. We said no to a lot of things. We cut a lot of things for a lot of different reasons.
- [15:13]
Unknown Co-Host
Got it. That's interesting. I'm curious, you know, as you're writing this thing, how much is kind of the market dynamics around AI hanging out in the back of your mind? Because there's all this chatter right now, as I'm sure you know, about whether AI is a bubble. Right. Whether this thing is going to pop and it's all this inflated value. I mean, is that idea on your mind as you're drafting this thing?
- [15:37]
Dean Ball
Well, it's on my mind as an observer and you know, until well now and also quite recently before commentator on the AI industry. So it's on my mind, but. And I don't think any of my colleagues really ever let that sort of thing weigh too heavily into the policy of the action plan. I think regardless of what happens to, you know, individual stocks or companies or products or whatever else in the next year or two, I think the reality is that AI is a transformative technology that's here to stay with us. So I think it's about setting America up for success in the long term, kind of regardless of how. I mean, maybe even the current LLM paradigm peters out. I really doubt that. Again, maybe that's the perception, but I think that there will be. And maybe the data center investment slows down or something. But I think the technology that we have right now has already unlocked so much productivity potential that it's just a matter of diffusing it into the economy, which even if the technology froze in place right now, still tons of potential.
- [16:52]
Unknown Co-Host
Got it. That's interesting because, I mean, obviously the plan is very focused on kind of driving the development, bringing all this energy to bear to meet the energy demands. And I think some folks worrying about a bubble might see the plan as like inflating that even further because of all the investments and all the money and energy going into it. Sounds like what you're saying is like, regardless of the near term, there's a lot in here that is really long term focused.
- [17:20]
Dean Ball
Yeah, I mean, we, we, you know. Well, certainly I don't think government should be putting its thumb on the scale for particular, you know, companies or investment directions or things like this. You know, I think we should let private capital allocators make those calls. Government should procure AI when AI is useful to it and advances the mission of A particular agency. But yeah, I mean again like we'll kind of see what happens. But in general, I'm sort of relatively old fashioned person who sees in things like this, I see the role of government as being to create the conditions for success regardless of of what happens in the short term.
- [18:05]
Unknown Co-Host
One last question for you here, Dean, and that's just you're obviously back now to being an observer of how AI is developing and observer of how this administration now kind of takes this plan you've written and carries it forward. I wonder what you'll be watching for. What are the obstacles that will most be in your mind now that they take these ideas and take them from paper to reality?
- [18:27]
Dean Ball
The most important obstacle is not specific to this administration or any current political dynamics. It's the obstacle that always faces government, which is sacrificing the important at the expense of the urgent. You're always deal. I mean, trust me, there are things in the action plan where I think we need to develop some of the policies and some of the intellectual frameworks around some of those recommended actions better than they currently are. And it's easier to do that work outside of government. And that's what I get really excited about inside government. Like you are being bombarded, especially the White House, but everywhere in government you're being bombarded with like, you know, a thousand things a day.
- [19:10]
Unknown Co-Host
There's always a fire to put out. Yeah, yeah.
- [19:12]
Dean Ball
And it can just be quite hard to focus on these sort of long term objectives that like you're not getting actively get in trouble or something if you don't do so. That would be one. You know, another important one that I think of course is I hope that everybody in government and other interested parties outside of government sees that the action plan is both a plan for the government with the bullet points that are like, you know, here are the five things that we're going to do right now. But that also it's a strategy and there are subheaders in the action plan that say things like build world class data sets or something. Those strategic objectives I think can be contributed to by a wide variety of parties, including Congress, including private industry, including civil society. And so I hope that people from all stripes all over the country see those strategic objectives as a calling card to do their own work and contribute their own things because there's much more to do beyond just the bullet points in the action plan.
- [20:23]
Unknown Co-Host
Got it. Well, listen, Dean, I appreciate you being.
- [20:25]
Stephen Overle
Here on Politico Tech.
- [20:27]
Dean Ball
Thank you so much for having me. It was fun.
- [20:30]
Stephen Overle
That'S all for this week's Politico Tech. If you enjoy Politico Tech, go on and subscribe and recommend the show to a friend or colleague. And for more tech news, subscribe to our newsletters, Digital Future Daily and Morning Tech. Norma Malaiko produced today's episode. I'm Stephen Overle. See you back here next week.