Dispatches from Mitch
Same party, different pages
New polling from POLITICO and Public First indicates a gap between President Trump and most of his supporters when it comes to AI policy. In the name of industry competitiveness, the White House wants only a light-touch federal standard. But as reported by POLITICO’s Katherine Long:
Only 13 percent of people who voted for Trump in 2024 said the federal government should stay out of regulating AI and let the market decide; about 3 out of 4 Trump supporters wanted the government to either impose strict regulations on the industry or set broad principles for companies to work out.
These voters’ views are complicated:
42 percent said the benefits outweigh the risks, another 42 percent said the risks outweigh the benefits and 16 percent said they didn’t know.
MAGA Trump voters were evenly split over prioritizing safety vs. “beating China, even if it means fewer safeguards.”
But I should note that in a POLITICO interview released yesterday, White House AI policy advisor David Sacks (previously titled the “AI and Crypto Czar” for the administration) explicitly defined winning against China as commercial dominance:
The way that you measure winning, I think, in a globally competitive market is based on market share. If in five years we look around the world and all the datacenters are running on Huawei chips and DeepSeek models, that means that we lost. We don’t want to have that future. What we want to see is that the whole world is running on American chips and American models. That would lead to the best economic results for the United States. It would also lead to the United States having more soft power in this area.
Obviously we don’t want to ship our leading edge semiconductors to China or something like that. But, we want to help encourage our companies to have the greatest market share.
I suspect that most of the Trump voters polled in this survey assumed that “beating China” meant something else. But in any event, “beating China” is a false prize: Racing to superintelligence ensures that the only winner will be the AI.
Waved through
A report by the Washington Post’s David J. Lynch describes how the AI boom is warping the outcomes of President Trump’s trade policies. Citing “the needs of the United States economy,” the President quietly gave tariff exemptions to most of the high and low-tech materials going into all those new data centers.
I mostly report on this piece because of a statistic that made me do a double-take:
Nearly one-quarter of the $3.4 trillion worth of goods that the U.S. imported last year was tied to the AI boom, according to an April study by the Federal Reserve Bank of Minneapolis.
That’s... a lot. And it’s new. AI imports have grown 73% since 2023, compared to 3% for everything else.
A cursed problem
Whenever I’m asked to recommend non-technical resources for people new to AI’s biggest risks, the first thing I point to is almost always the Rational Animations YouTube channel. They adapt substantive works in AI safety (and other topics) with superb voice talent and playful animation.
I’m biased, but I think the video they released this morning is one of their best. It’s an adaptation of Chapter 10 of If Anyone Builds It, Everyone Dies, the New York Times Bestseller by Eliezer Yudkowsky and Nate Soares produced with the help of us here in the MIRI comms team.
That chapter tends to be a reader favorite. It walks through historical case studies of engineering failures and gotchas in other fields, with takeaways for AI. It becomes clear that safely building artificial superintelligence is an especially “cursed” problem — not something you should plan to just figure out as you go.
I need not spoil the chapter or the video. Both are great. Go and consume the medium of your choice!
For some influencers, the price is right
In a piece for WIRED, Taylor Lorenz reports on a dark-money operation paying influencers on TikTok and Instagram to spread pro-American-AI and anti-China messaging.
Lorenz learned about the campaign when one of the marketing agencies running it, SM4, tried to recruit her, offering $5,000 per TikTok. Other content creators confirmed getting these offers. (I note that a February story in CNBC claimed that undisclosed partnerships of this type are yielding some influencers $400,000 to $600,000.)
Lorenz traced SM4’s funding to Build American AI, which is tied to Leading the Future — the $100M pro-AI superPAC whose known backers include OpenAI’s president Greg Brockman, the venture capital firm Andreessen Horowitz, and the co-founder of military software firm Palantir, Joe Lonsdale.
A sample script instructs influencers to deliver lines like the following while doing things like “making breakfast for the kids”:
I just learned that China is trying really hard to beat the US in AI. If they do, it could mean that China gets personal data from me and my kids, and takes jobs that should be here in the US. In the AI innovation race, I’m Team USA!!!
Lorenz went looking for TikTok and Instagram videos from influencers who seemed to have taken the payouts. The ones she found were labelled as advertisements, but provided no information about who paid for them.
Mac markup
This may be petty of me, but I’m enough of a home computer enthusiast to have joined the legions of gamers incensed by the AI boom’s role in jacking up the prices of computer components. The hikes are obvious to someone who prices out their parts over weeks and months, but can go lost on consumers who only buy finished products once a year or two, and I want more people to notice.
So here’s an article from the Wall Street Journal’s Rolfe Winkler about how Apple is raising the entry-level price of the Mac Mini from $599 to $799. The change is due, in part, to the product’s appeal with AI power users; they often like to give “always-on” agent setups (like OpenClaw) their own dedicated machines.
Winkler explains that Apple masks such markups by dropping lower-tiered loadouts from the line-up. The $799 model was already available, and has 512 gigabytes of storage. What’s new is that you can no longer buy the $599 version with only 256 gigs.
Yes, if you hadn’t already figured it out, the cost difference to Apple for its storage options is almost never anywhere near the price difference they charge for it.
No Oscars for AI
Industry and mainstream outlets (here’s BBC) report that the Academy of Motion Picture Arts and Sciences has updated Oscar rules to officially exclude AI from eligibility for acting and writing awards.
Specifically, nominated acting must be “demonstrably performed by humans” and writing “must be human-authored.”
According to the new rules, AI tools used elsewhere in a film “neither help nor harm” nomination chances, but this seems at odds with the Academy’s claim that they will take “the degree to which a human was at the heart of the creative authorship” into account.
I expect that tension to get ugly when it comes to the visual effects categories, where lines between the contributions of human and machine can be murky.
The analyses and opinions expressed on AI StopWatch reflect the views of the individual analysts and the sources they cover, and should not be taken as official positions of the Machine Intelligence Research Institute.


