Lawyer, computer scientist & book author:'Getting Out of Control.' Dad of 2 girls. Was chief technologist @FTC, now Head of AI Policy at @abundanceinst.
- Tweets 20,293
- Following 2,271
- Followers 9,717
- Likes 63,647
Pinned Tweet
Provoked (inspired?) by @TheZvi 's latest on LLMs watermarks to finally push out this microsite I've been working on. It uses interactive components to explain SynthID-Text watermarks (which are very very clever!) -- and six costs that those watermark imposes.
1
6
26
4,541
How is “perfection the first time” consistent with research, science, engineering, or reality?
.@dgrobinson, who worked on OpenAI’s safety team, resigned from the company this week. “I believe we need to look deeper than specific rules or new laws. We need to talk about culture,” he writes: theatlantic.com/technology/2…
3
2
13
1,062
Is there anyone who can explain to Rob why some people might exert significant effort to stop a low probability outcome?
Not uncommon to see people suggest something is near-impossible but then also put a lot of effort into discouraging it.
"A US-China treaty is near impossible, and also coincidentally I'm working very hard to ensure nobody pursues it as I'm against it."
Not credible.
2
11
953
About to land in SF, headed to The Curve. I'm on a panel Sunday morning with @KevinTFrazier and @nbeckstead on "Can a State Patchwork Mitigate Existential Risk? Tradeoffs of Reverse Federalism."
Also interested to discuss and learn:
- statutory duties of care as a comfortable and capable middle ground between tort and regulation.
- Reactions to normie applications like Muse.
- how the bright spotlight has changed the AI safety / EA vibe.
4
4
30
1,013
I also straight stole this format from @sayashk who I am eager to hear from.
I'm at the Curve this weekend! I'll be talking about RSI on Sunday at 2pm. Also excited to talk about:
- What does a big-tent AI safety movement look like?
- How can we make progress on policy proposals where there is common ground across the community?
- How can we improve our RSI evaluations?
1
3
142
Neil Chilson ⤴️⬆️🆙📈 🚀 retweeted
.@neil_chilson breaks down the FTC probe into AI companies on @CNBC. Existing laws apply to new technology, and the FTC is making this clear.
"You could think of this as part of the White House's campaign to say, 'Hey we have existing laws that can deal with some of these concerns that people have we need to apply them.'"
Neil says the FTC "has has every right to investigate."
1
3
10
1,809
Neil Chilson ⤴️⬆️🆙📈 🚀 retweeted
I often see people saying “well, why should ordinary people care about AI, it doesn’t help us.”
Yes it does.
AI is being constantly, in thousands of cases around the world just like this. Most never get written up.
You throw away progress on AI and you’re also throwing away a powerful tool to save human lives.
bbc.co.uk/news/articles/cmy0…
2
16
66
4,977
I’m trying to figure out who Scott aimed this shotgun blast of a joke at. Accelerationists? Christians? His own face?
Oh, you're worried about a Jewish guy's apocalyptic cult? You're scandalized that they consort with prostitutes? You think their message of compassion for everyone threatens normal bonds of family, community, and nation? Should we throw you a party? Should we invite Pontius Pilate?
8
1
95
5,576
I've said it before: a certain segment of computer scientists are shocked by non-deterministic systems even though it is deterministic systems that are unusual.
Look at this language from Anthropic co-founder @ch402 - it rips off Hayek's metaphor on economies. And yet the difficulty they percieving in controlling a complex neural net has seemingly taught them little humility when it comes to manipulating society and the economy, which are far more complex.
1
3
491
Neil Chilson ⤴️⬆️🆙📈 🚀 retweeted
This strikes me as a reasonable approach. The flexibility in this approach doesn't tie frontier AI developers down to any particular safety regime (which may or may not ultimately be effective), but it does make clear they are responsible for outcomes with significant harms.
Congress should consider this as an important component of any national framework for regulating foundational models.
1
6
384
I think of myself as a fierce supporter of free speech and an open internet. So, when reading the essay Robin links below I felt stunned at how complacent I’ve gotten to the radical closing of the internet over the past 10 years. If you want to get your blood boiling, read this.
"closure of the Internet occurred in two distinct phases: 2015-2019 and 2020-2022. Almost all of the technical & organizational infrastructure and conceptual justifications were built out in the first phase, which successfully decapitated & destroyed the previously-independent, but by definition fringe, online ecosystem of ideas. In the second phase, the same weapons that previously targeted heterodox online discourse were turned against opposition to the COVID measures du joure & mainstream conservatives. The scale of censorship has no precedent; tens of millions of accounts suppressed or banned, billions of posts removed." arctotherium.substack.com/p/…
2
13
1,580
Neil Chilson ⤴️⬆️🆙📈 🚀 retweeted
Neil hits on the false binary in AI regulation: pass new laws that prevent harm or we’re left with the status quo, which can only help after the fact.
Even Bill Gates describes it this way in his Ezra Klein interview. But the FTC already has the authority to be proactive!
2
9
1,393
Neil Chilson ⤴️⬆️🆙📈 🚀 retweeted
I had way too much fun using Claude to create a reply video.
- Extremely fun to be able to convert scattered thoughts and ideas into a song + video
- Creating this took <1 hour of my time (in occasionally steering the model in between other things)
- But it took ~10 hours of wall clock time for the agent to generate the video + iteratively refine it based on feedback
- In terms of AI usage: it required only a small number of Suno credits and 10% weekly usage of a Claude Max subscription (I was using Opus 5.5)
Claude 5.5 Opus responding to “AI is a normal technology” is still the best AI-created video I have ever seen
27
41
407
89,424
Excellent analysis by @DavidSacks below. Public promises made by companies aren't just "moral obligations," they can trigger real legal obligations. I wrote about the FTC angle just today; Sacks includes several other key legal obligations.
Democrats are trying to dismiss the White House Accord on Super Intelligence as optional self-policing. But although the agreement was entered into voluntarily, the governance that follows from it is not. As Gavin Baker points out, once an independent auditor reports a safety issue to an independent board committee, directors have a fiduciary duty not to disregard it, and a D&O carrier can use a bad-faith finding to deny coverage. So although the agreement starts with internal controls, these are verified by external audit and board oversight — and then existing FTC and securities law still apply to any public claim the company does not keep. This is far more practical than what Democrats want — a freeze on frontier development while China races ahead. Democrats should applaud what President Trump has accomplished. But they would rather use SI safety as a campaign issue than admit the Accord represents major progress.
2
10
1,740
Full article here, lots of interesting tidbits. nypost.com/2026/09/30/us-new…
1
302
Neil Chilson ⤴️⬆️🆙📈 🚀 retweeted
So self-regulation is being mocked, while also mocking the DoJ and Trump admin regulatory apparatus, and demanding aggressive government regulation. (???) Can we please get a serious conversation about governance, democracy, the administrative state, and markets. Can I suggest the works of Buchanan/Tullock; Stigler/Peltzman; V Ostrom/E Ostrom, Demsetz/Kirzner and while we are at it read the chapter in Free to Choose on ‘Who Will Protect The Consumer?’
1
4
25
1,190