I'm not addicted to using language models, I can write and do work without them, I still remember how to search by myself, I keep repeating... as I repeatedly click refresh on Claude and ChatGPT. Thankfully, Gemini (partially) saves the day.

It's nice to see that #AAAI2026 is avoiding any religious discrimination; it's on Friday, Saturday, and Sunday, so it's equally hostile to Muslims, Jews, and Christians.

If you're among the 2000+ authors citing research evaluated by Unjournal.org, go to unjournal.pubpub.org to learn how commissioned experts rated & assessed that research. Top 'citers of unjournal-evaluated research': https://bit.ly/3WRHc8W include Esther Duflo, Julian Jamison, & Berk Özler

I'm here at #AIES2025, and still worry quite a lot about this. The deep skepticism about AI systems ever being generally capable, or even human-level in specific domains, doesn't seem to have changed over the past few years.

Excited to be here today at #AAAI #AIES2025. Looking forward to meeting more people and discussing governance and societal impacts of AI.

New RAND report on an important (and messy) question: When should we actually worry about AI being used to design a pathogen? What’s plausible now vs. near-term vs. later? (1/12) I helped convene two expert Delphi panels in AI + Bio to weigh in. Full report: www.rand.org/pubs/researc...

From the other site - not a complete explanation, but correct as far as it goes: "the current configuration of economics/ wealth distribution is pretty solidly optimized to drive the wealthiest people in society batshit insane, which - to some extent - explains a lot of things you see around you"

🤖📰 Effective YESTERDAY: China has mandated a digital watermark for all AI-generated content. www.cac.gov.cn/2025-03/14/c... Translating in 🧵.

When a measure becomes a target, it ceases to be a good measure—Goodhart’s law seems clear enough. But look closer and you see at least 4 variants, argue Manheim & Garrabrant: Regressional, Extremal, Causal and Adversarial, which offer deeper insight: buff.ly/h9ORc76

We're clearly hitting a wall - because if AI progress is really exponential, why aren't the numbers going up faster? Checkmate, AI industry!

Short post on why chip production location, and policy to influence it, matters less for AI than people believe it will. forum.effectivealtruism.org/posts/7GHbwi...

Yet another place where @garymarcus.bsky.social is right that LLMs don't have a correct world model. (The filter exerts pressure on the water level. It doesn't understand that. But then, I'd bet most humans wouldn't realize this either.)

Critics say the same thing about most current news reporting...

Are They Starting To Take Our Jobs?

Not all technology is safe. I wouldn't try to stop people from producing knives because they can be used for stabbings, but I would ask for mandated safety tests before cars can be sold, and rules about who can drive those cars, and law enforcement to enforce the laws.

LLM slop on arXiv is indeed a huge challenge for us. The trouble is that we don't think reviewers in conferences and journals are handling it well either, so we can't rely on them. We are trying many things to detect and remove slop.

Last year, the crypto industry set up a network of super PACs. Its strategy? Throw huge amounts of money at politicians to discourage crypto regulation. Now many of those same people have launched a $100m campaign to do the same thing for AI. Read my full piece on Transformer:

text microblogging sites tend to punish behaviors people say they value (nuance, humility, vulnerability, accountability) and value behaviors people say they dislike (nastiness, irony-poisoning, scatalogical humor, etc)

New blog post, with Ram Rahum, "A Conservative Vision For AI Alignment," presenting our ideas for what a (intellectually, politically, and socially) conservative view of AI alignment could look like.

The damage to the rule of law caused by public figures not being called to account for obvious corruption is not easy to measure, but it is huge. Poisoning important social norms isn't a reversible action.

How can you know? It's like saying there must be some smallest item in a set - but sometimes there isn't! It might be that the earliest use of the term is infinitely divisible because ther are infinite mentions in finite time, or the term goes back in time forever, I guess if there was no big bang?

As long as academia allows and encourages narrow disciplines to be self-reinforcing and obscurantist, thus largely immune to outside scholarly input, I expect that the primary pushback against these narrow and often very ideologically-driven faculties will be funding cuts, not any academic reform.

Wow look at all these anti-Semitic Israelis wow

The success of superforecasters, compared to both the general public and domain experts, is mostly not about overperformance by a special group. The skills and practices are unusual, not especially hard, nor intrinsically rare - and getting those basics right is a huge advantage.

If done well, regulation and liability for current medical AI could reduce mistakes and mitigate ethical concerns. But if your concern reliably leads to more people being dead because doctors aren't using new technology, you're doing ethics wrong!

The collapse of liberal Zionism needs to be studied. As one of the main bulwarks against Israel's most horrific tendencies, the inability and refusal of so many in both America and Israel to stop Netanyahu's fantasies has helped this genocide to go unchecked. Few will say it, but it's the reality.

Amazing work by @kyletokens.bsky.social and @scasper.bsky.social et al. Congrats!

Sycophancy isn't really addressed by current OpenAI approaches. This is seen, for example, contrasting types of errors found - not phrasing, but actually substantively different analysis, discussing different issues - when asking the two substantively identical below questions.

Most AI companies have made an active decision not to watermark AI-generated videos in any way. Sure, it would not be a perfect solution - but it makes it obvious they aren't interested in trying to be at all responsible about technology they are developing and deploying.

Sounds like a co-founder of OpenAI is effectively publicly admitting they have no expectation that there will be meaningful human oversight and/or control of the systems they will deploy. (Contra their recent commitment under the EU AI act; "AI-powered tools" are not meaningful *human* oversight.)

Hot take: Most AI-bioterrorism risk "debate" is people shadow-boxing over definitions, not different predictions about concrete points. I've talked to many other experts. We often don’t agree on what “enable” means, so people think the risk is uncertain in ways it is not. 🧵

Luke Kemp's new book, Goliath's Curse, is an absolute must-read for anyone thinking about humanity's future - I've certainly learned a ton. Alas, it's not coming out in the US until September - but you can order the e-book from the UK already.

Does Trump's AI Action Plan have what it takes to win? There are sharp policies for US AI but also contradictions around China strategy, AGI preparedness, permitting reform, and execution speed. Here are my ten takes on what it gets right, what it misses, and what it means for America's AI future.

Friends! Today I'm launching the Realtime Techpocalypse Newsletter. It's motto is: "Stop doomscrolling and just read this!" I'll provide a weekly mix of news that isn't covered elsewhere, and commentary drawing from my unique background in eschatology. www.realtimetechpocalypse.com/p/introducin...

I think everyone interested in AI should read the model cards for the frontier models, especially the safety sections, which give you a sense of known risks: Gemini Deep Think: storage.googleapis.com/deepmind-med... Claude 4: www-cdn.anthropic.com/07b2a3f9902e... o3: cdn.openai.com/pdf/2221c875...

LLMs are advancing fast in areas where self-supervised learning and verification are easy - as the IMO Gold results show. But I don’t see strong reasons why research enabling AI self-improvement must be hard to verify.

Very happy to announce that my org has helped get commitment from the Israeli Ministry of Health to finalize plans to mandate salt iodization in Israel. Following the meeting this past Wednesday, they have one month to present the final plan to the health committee. Wish us luck!

People think of respiratory disease as a fact of life, especially for kids - and assume that it doesn't really matter. In fact, @meltemdaysal.bsky.social et al showed children's minor illnesses has significant impact for decades. Reducing "routine" disease is a big deal! cepr.org/voxeu/column...

AI safety standards often rely on a comforting assumption: that human oversight can always be applied. But in many cases, that’s simply false. New paper (with @adnhw.bsky.social ) explains oversight, control, and the need to admit both are limited. arxiv.org/abs/2507.03525

AI companies are suddenly very concerned about job loss. Not because that’s the biggest risk— but because “we might kill everyone” isn’t a great look for their quarterly revenue planning.

AI developers often say “don’t worry, the systems have human oversight.” But for many current AI systems, meaningful oversight is impossible! New paper (with @adnhw.bsky.social): “Limits of Safe AI Deployment: Differentiating Oversight and Control” 🧵

The EU AI act regulates the "intended purpose" of an AI system - that is, what the provider intended. Frontier AI firms say their models will soon be able to do all human work - which seems to be saying that there are no unintended purposes.

Kakistokleptommunism: A political-economic system that fuses kakistocracy (rule by the worst), kleptocracy (rule for theft), and central control. It systematically destroys coordination, misprices everything, and imposes tariffs to enrich politically connected authoritarians.

My newest post on Exploring Cooperation, responding to @hlntnr.bsky.social's recent essay. I discuss Virginia Postrel's dynamism as being too fragile in the face of technological acceleration, but argue that material progress is still critical for fighting stasism. www.lesswrong.com/posts/upKqoH...

120 years later, and the @nytimes.com has learned nothing about technology predictions. (via @peterwildeford.bsky.social, @theojaffee.bsky.social)

Clearly Trump is corrupt - but I think most of the world is currently greatly overestimating the extent to which he's going to stay 'honestly' corrupt.

Our new paper is out in AI & Society: "The Necessity of AI Audit Standards Boards," By myself, Sammy Martin, Mark Bailey, @contact.ms, & Ross Greutzemacher. We argue that without a board creating independent audit standards, AI auditing is broken by design. link.springer.com/article/10.1...

I have been using an LLM to help me with tracking / coaching myself on health and nutrition. It's an area where I'm comfortable asking an LLM instead of an expert - not because I trust the LLM, but because the median "expert" I'd consult is at least as bad.

Our new paper is out in AI & Society: "The Necessity of AI Audit Standards Boards," By myself, Sammy Martin, Mark Bailey, @contact.ms, & Ross Greutzemacher. We argue that without a board creating independent audit standards, AI auditing is broken by design. link.springer.com/article/10.1...

"Gemini 2.5 Pro system card has now been updated with frontier safety evaluations results, testing for critical capabilities in CBRN..." "...biological, radiological and nuclear domains." So there's a C that everyone keeps omitting in CBRN, which seems bad.