Loading summary
Transcript38 lines
- [00:01]
State Farm Ad
This episode is brought to you by State Farm. Knowing you could be saving money for the things you really want, like that dream house or ride, is a great feeling. That's why the State Farm Personal Price Plan can help you save when you choose to bundle home and auto bundling. Just another way to save with a personal price plan. Prices are based on rating plans that vary by state. Coverage options are selected by the customer. Availability, amount of discounts, and savings and eligibility vary by state.
- [00:36]
Stephen Oberle
Hey, welcome back to Politico Tech. I'm your host, Stephen Oberle. Like many in the tech world, I spent my Wednesday digging into President Donald Trump's new AI Action plan, which effectively outlines his strategy for keeping the US at the forefront of artificial intelligence.
- [00:57]
Donald Trump
Whether we like it or not, we're suddenly engaged in a fast paced competition to build and define this groundbreaking technology that will determine so much about the future of civilization itself.
- [01:12]
Stephen Oberle
That's Trump speaking after the plan was released.
- [01:15]
Donald Trump
America is the country that started the AI race, and as President of the United States, I'm here today to declare that America is going to win it.
- [01:28]
Stephen Oberle
For the most part, the AI Action Plan is a gift to the tech industry. It directs federal agencies to slash regulations that slow down the development of AI and the construction of new data centers. And it pushes the state and Commerce departments to promote the sale of American AI technology abroad. It also deters states from from imposing AI regulations of their own. But one part of the plan could give Silicon Valley heartburn because it drags the industry further into Trump's war on wokeness. To help me break all of this down, I called up Neil Chilson. Neil is the head of AI policy at the Abundance Institute, a libertarian tech policy nonprofit, and he was previously the chief technologist at the Federal Trade Commission. I knew Neil would be a fan of the plan. He's generally against AI regulation, but I wanted to hear how he thinks some of Trump's big ideas on paper could play out in reality. Here's our conversation. Neil, welcome to Politico Tech.
- [02:52]
Neil Chilson
Thanks for having me.
- [02:54]
Stephen Oberle
So this action plan came out yesterday and frankly, it covers a lot of ground in just 28 pages. For you, what's the big headline here?
- [03:02]
Neil Chilson
For me, the big headline is this continues the course correction of this administration from the previous administration and frames up AI as a giant opportunity for America that needs us to seize it and that it's really important that we do. So that's my big takeaway from the AI Action plan.
- [03:21]
Stephen Oberle
Yeah, I know many folks criticize the Biden administration for being Too focused on risks of AI, not focused enough on opportunity. And we have seen a real sea change with Trump coming into office and how we talk about AI with this plan in particular, let's start with the big focus on deregulation because there's calls for agencies to review and repeal regulations that they find hinder, you know, the development and deployment of AI. I know that you are someone who's generally opposed to those AI regulations. I mean, how much regulation is there to get rid of, though? And what's the action here that could most move the needle? You think?
- [03:59]
Neil Chilson
So at the federal level, I think there are a lot of. Because AI is a general purpose technology, I think we don't quite know the full scope of what regulation regulatory barriers might be stopping entrepreneurs from deploying in some of these highly regulated sectors like healthcare or transportation, for example. And so I think this plan directs the agencies to help figure out what that is. It's going to be more complicated than just doing a search for AI in the Federal Register.
- [04:29]
Stephen Oberle
Right, Right.
- [04:30]
Neil Chilson
Most of these regulatory barriers probably aren't going to say much about artificial intelligence, but they might talk about privacy or they might talk about types of approval processes that don't make a ton of sense for AI solutions. And so I think there's some real work to be done here at the federal level. I will say one of the other categories of regulatory or legislative concern that the action plan highlights is state level regulatory environments. And we have seen a lot of action in state legislatures about artificial intelligence. And the plan points out that those could be problematic for people who want to build this technology and, and take some steps to try to mitigate some of those concerns.
- [05:15]
Stephen Oberle
Let's dig into that because this was language that was flagged to me in particular about, you know, potentially withholding AI related funds from states that have regulation considered burdensome. This obviously comes on the heels of a moratorium on state AI laws that was struck down by Congress. Do you read this AI action plan to be a de facto moratorium on states enforcing AI laws?
- [05:40]
Neil Chilson
No, it's much more limited, I think, than even, I think the Senate proposal that was considered as part of the big beautiful bill. But it does show that the administration is worried about this and is exploring some options. So that language talks about considering the regulatory environment before spending AI related funds is. It's pretty cabin. Right. Like that's a lot of discretion to the agency. There isn't really a strict command there. So, you know, and some of the framing even says talks about, you know, both the need to make sure that regulation isn't getting in the way, but also the need to ensure that states can do the types of regulation that they traditionally have done, provided it doesn't affect AI innovation and the US's competitiveness overall.
- [06:36]
Stephen Oberle
Well, that's. There was some pushback to the moratorium, even from small government folks who just felt like it was overreach on the part of the federal government. And I guess one question in my mind is, does this new approach sort of allay some of those concerns, do you think?
- [06:52]
Neil Chilson
I think so. We were supportive generally of the idea of federal preemption or a moratorium of some kind, just given the level and overlap of all of these state regulations. But I do think that this kind of carves a middle path on some of that. And it, most importantly, I think it signals to the states that the administration sees a strong federal role in policy in this space and that they're not going to sit back and let the states do this, that it is by themselves, that it is a, it's going to have to be a cooperative effort between the federal government and the states for this very, you know, broad general purpose technology.
- [07:36]
Stephen Oberle
The other piece of this plan that really did stick out to me is the stipulation that the government will only contract with AI providers whose models, quote, are free from top down ideological bias. And this fits into kind of a pattern we've seen from this administration. You know, criticisms of so called WOKE technologies or accusations about conservative bias. How do you see that playing out from a practical perspective? Just because you start to get into questions about how trainable AI really is and questions about the First Amendment, I wonder how you see this new requirement playing out.
- [08:13]
Neil Chilson
There's a couple different directions you can take that issue. It is a very complicated one. So to the extent that the action plan is talking about conditioning federal grants on models being fit for purpose, you can, you can see how it makes some sense. And in fact, there is some uses of AI in government that probably actually have a requirement that they not be biased politically in one way or another because the government is using them. So the question is, is the AI action plan talking purely about the models that the government uses or is it talking about contracts? Any company and any model that a company that also contracts with the government produces. And it's not quite clear from the action plan which of those is the issue and those raise different constitutional concerns. And so I think that will make a big difference on how, I think frankly, sustainable it would be in the courts. There is the second dimension of this, which is like, practically what is possible here. And I've written some about the fact that a lot of the concerns that sort of land on the left of center about AI bias tend to be about implicit bias, where, oh, the data set is full of different ideas, and therefore the outputs might. After you train the AI, the outputs might reflect that biased distribution of ideas in the data set. That's very hard, actually, to fix. Technically, that's very hard because people have different ideas about what bias is. And so it can be very hard to fix that type of bias. Often when we hear about bias on the right, what they're really worried about is intentional bias, where somebody's putting a thumb on the scale about what responses come out. Those tend to be in the inference, like what we call the inference phase of AI. And so in things like the system prompt, and we've seen some examples of that in the GROK AI, where some of the instructions to the AI about how they respond. These are written in English.
- [10:21]
Stephen Oberle
Right.
- [10:22]
Neil Chilson
You can clearly see that they're sort of, like, aimed at a specific political objective. I think, practically speaking, those types of bias are just technically more feasible to get rid of. Right. And so if that's where the administration is focused, I think they will have an easier time at achieving their objective, just practically speaking. And again, it will, depending on what the scope of their application will be, will tell us how much constitutional challenges they'll have in that space.
- [10:53]
Stephen Oberle
Right, exactly. I mean, I sometimes wonder with these speech issues, if you essentially have the right and the left kind of just working opposite sides of the same coin. Right. It's. It's in some ways regulation, in some ways, you know, monitoring or trying to control online speech just from a different political perspective.
- [11:12]
Neil Chilson
Yeah. In the social media space, that's clearly the case. I often like to say that when people talk about, oh, everybody is concerned with social media censorship. And I'm like, well, one side's trying to build a fire extinguisher, and one side's trying to build a flamethrower. That doesn't mean they're both interested in fire. I mean, you know, so, yeah, there is a sort of diametrically opposed point of views on some of this.
- [11:48]
Kona Brewing Co. Ad
Hear that? Big waves are calling. Dive into refreshment with Kona Big Wave Tropical flavor and the taste of aloha in every drop this summer. Don't just listen. Join Kona on the Big Wave bonfire tour. More information@konabigwave.com BigWaveMusic Kona Big Wave Liquid Aloha Copyright 2025 Kona Brewing Co. St. Louis, MO Mahalo for enjoying responsibly.
- [12:17]
Stephen Oberle
You mentioned Grok, which reminds me, obviously these recent headlines, for instance, where grok, which is the AI model from xai, was criticized for some anti Semitic posts. And both Democrats and Republicans on the Hill have raised concern about that. With this plan that's come out, I wonder if there's a risk that companies really avoid moderation in AI in a way that does potentially lead to instances of not just hate speech, but also things like hallucinations or bias. Some of the other kind of documented problems with AI that have not yet been solved.
- [12:55]
Neil Chilson
Yeah, I mean, I think that's a possibility. I think that moderation in the AI context is so different from moderation in the social media context. Social media, you're moderating somebody else's speech in the AI space, you're reviewing the output of your own product. And so the feedback loops are very different there. I think the constitutional and just the policy concerns are very different. And so I'm not sure we can take really clear lessons from the social media space about what might happen here. And, and honestly, the. Despite the coverage on those particular points, the action plan does not emphasize these issues very much at all. I mean, there's only two mentions of bias in the entire document, from what I can tell. And there's really only like this sort of mild recommendation. The primary policy leverage is like this mild recommendation that for procurement, that these models have to be politically neutral. I don't see a big impact. I think a lot of these companies are already struggling to figure out how to, how to make everybody happy with their models. And so I think they'll continue to struggle to try to make everybody happy with their models.
- [14:02]
Stephen Oberle
The other cornerstone of this plan is really focused on infrastructure and kind of everything you need to build in order to get to the AI future that people are talking about. And, you know, data centers are a big focus. Semiconductor fabs also a focus. This is work that in some way the Biden administration had undertaken, including through the CHIPS Act. How different is the path that Trump seems to be charting here?
- [14:29]
Neil Chilson
So, you know, one of the criticisms of the CHIPS act was that it was weighed down with a bunch of other sort of policy initiatives, so goals that people wanted to achieve that were separate and in some cases, like pretty orthogonal to the, the goal of building chip manufacturers in the US and so I think that this action plan shows a pretty different path. It's pretty focused on infrastructure, building infrastructure. And at least at this level it's not trying to incorporate a lot of other types of policy agendas into it. I do think that that narrower focus means that it will be more likely to achieve it. And I think also just the tensions that existed in the Democratic coalition Around the Chips act are less evident here. Right. So, you know, if you're a big believer in the idea that NEPA or those types of clauses help achieve climate goals or help protect certain types of interests, then you have a real tension because those also slow down the building out of infrastructure. Those tensions are less apparent in the action plan. They want to get those barriers out of the way. I don't think they see them as useful in, yeah, really, at all. And so I think it's an easier path for them to talk about what they want to achieve here without risking, you know, other parts of their coalition getting upset.
- [15:56]
Stephen Oberle
Right. I mean, that's one of the big pieces of it, is trying to kind of strip back permitting requirements, environmental, you know, study requirements, things that can slow down the development. Was there anything in the kind of infrastructure part of this plan that surprised you? Because we have heard a lot from this administration, for instance, on prioritizing energy development and building data centers. I wonder if there was anything new in the plan that raised kind of a flag for you.
- [16:25]
Neil Chilson
You know, I mean, I saw the emphasis on new types of energy generation as useful. It's interesting that they don't include really wind or solar in that list. I think those are disfavored even though they are very powerful and useful energy producing technologies. And some places pretty red states like Texas are deploying lots of them. So that's interesting to me, maybe a little bit concerning because I do think that this is a environment in which we really do need an all of the above strategy to produce the kinds of energy that we need, not just for AI, but like for the growing demand for United States consumers. So that's a little bit concerning.
- [17:10]
Stephen Oberle
I'm curious. You know, I have seen, I'll say, some version of kind of an AI plan coming out of every White House, really going back to the Obama administration. I remember in their waning days, them putting out kind of an AI plan. And these are. They're all different. They have different priorities, different flavors. But I wonder how much these really move the needle. You know, if we sort of every four years have a new plan, kind of a reset, I mean, are we actually advancing in a meaningful way on US AI priorities, do you think?
- [17:40]
Neil Chilson
You know, in one sense, it's kind of hard to answer that. Because a lot has changed since the Obama years on what AI means. Right. Even since the first, the first, you know, Trump administration where they had these initial AI plans and AI executive orders, I think it's taken on a new salience in a way that it hasn't had in the previous administrations. And so I do think that this is a higher priority for this administration than it has been for any previous administration, just because of the fact that it's a higher priority for everybody than it had been in the past. And so I think the Trump administration has a real unique chance here to move its agenda and its approach, given the vast interest and concern about this issue locally across states, across the country and across the world.
- [18:29]
Stephen Oberle
Well, I guess with that in mind, one last question for you here, and that's just what are you watching for next as this plan gets implemented? What's the one thing you're really going to be paying attention to?
- [18:39]
Neil Chilson
Well, I'm really excited to see where the opportunities are for streamlining at the federal government level. And I'm also really, like I said, I'm really interested in, in how the Trump administration and the White House continue to engage with the states to say, like, hey, this and Congress to say, like, this is a space that the federal government has a really important role in and we can't be dominated by, you know, state legislatures passing laws before Congress does. And so that's probably the one that I'm the most focused on, that interaction between the federal government and the states. How will that play out? And will the Trump administration see state legislation as a challenge to achieving its global AI dominance?
- [19:27]
Stephen Oberle
Well, I will definitely be watching that, too. I have to say, Neil, thanks for being here on Politico Tech.
- [19:32]
Neil Chilson
Absolutely. Thanks for having me.
- [19:36]
Stephen Oberle
That's all for this week's Politico Tech. If you enjoy Politico Tech, please 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. Our producer is Normal Moleichel. I'm Steven Overle. See you back here next week.
- [20:07]
Neil Chilson
Sam.