Luke Muehlhauser
@lukeprog
Open Philanthropy Program Director, AI Governance and Policy
Reminder that the nonprofit AI safety field (or "industrial complex" lol) is massively, massively outgunned by, well… The Actual AI Industry. archive.is/DFvLR
We need a new Marshall Plan for shipping air conditioning units to Europe
Europe will do anything to beat the heat except install air conditioning. bbc.com/news/articles/…
A hidden fire‑safety rule is quietly adding $1+ million to the price of many new apartment buildings. Crazy part? The gear it mandates has been used exactly once in a real US fire. Let’s unpack this 👇
Transformer continues to be the quickest way to keep up with the dozens of AI news stories that come out every week. Highly recommended! transformernews.ai/p/trumps-surpr…
Some people are pretending to doubt this, but e.g. the entire nonprofit AI safety ecosystem runs on much less than $1B/yr, whereas annual expenditures for just three leading AI companies (MSFT, GOOG, META) is >$450B/yr.
Reminder that the nonprofit AI safety field (or "industrial complex" lol) is massively, massively outgunned by, well… The Actual AI Industry. archive.is/DFvLR
I hate to break it to you @nvidia, but we actually looked into this a few years ago at @CSETGeorgetown. We combed through 66,000 of the PLA's actual purchase records. It turns out they *overwhelmingly* use your chips... And at the time, you didn't do anything about it. 😬
Nvidia CEO Jensen Huang said the US government doesn’t need to be concerned that the Chinese military will use his company’s products to improve their capabilities bloomberg.com/news/articles/…
1/ AI agents are increasingly being deployed for real-world tasks, but how safe are they in high-stakes settings? 🚨 NEW: OpenAgentSafety - A comprehensive framework for evaluating AI agent safety in realistic scenarios across eight critical risk categories. 🧵
Well, we're back to this point now. nytimes.com/2025/07/17/tec…
what the actual fuck npr.org/2025/04/09/nx-…
UK AISI (@AISecurityInst) directly improved the safety/security of OpenAI Agent before release. Nice work, and kudos to @OpenAI for giving special access!
The last thing you see before you realize your alignment strategy doesn’t work
Spearphishing PSA—looks like there's a concerted attack on AI safety/governance folks going around. Be wary of calendar links via DM, and *never* give a 2-factor auth code over the phone. I almost got caught by this—got a phone call last week, but figured out it was sus. 🧵
And @ajeya_cotra's account has been hacked by the same folks - if you get messages from her asking to schedule a meeting, be very wary! She says she will never reach out to a potential grantee by Twitter, always email.
🚨 The Center for a New American Security (CNAS) is hiring for *five* new AI positions! 🚨 This is a terrific opportunity to work on some of the most fascinating, fast-moving, and consequential policy issues of our time with the best colleagues around.
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
Highly recommend applying to our two new Abundance and Growth roles at @open_phil ! We are looking for up to 4 new team members including specialists in areas like energy, housing, clinical trials, etc. as well as broader generalists to help us direct $120M+ in the coming years
The Abundance and Growth Fund at Open Philanthropy is hiring! We’re looking for 2-4 people to help expand this new $120+ million program to accelerate economic growth and reduce the cost of living through strategic grantmaking and research. (1/4) openphilanthropy.org/research/annou…
Give to @GiveWell. They are the best.
We recently did “lookbacks” to see how well 2 recent @GiveWell grants met our initial expectations. 1 went better than expected, 1 went worse. More on what we learned 🧵
Good article by @karthiktadepall on the relative range of heuristic. I think people often are insensitive to this difference in range, intuitively seeing considerations as roughly equal in strength. That can be a big mistake.
An excellent walkthrough of the 'AI 2027' scenario exercise. youtube.com/watch?v=5KVDDf…
People keep claiming that US states have proposed 1,000+ AI-related bills this year, but that number just doesn't hold up to analysis. A thread:
I honestly can barely believe this happened. Total YIMBY victory in California today. After Newsom signs this bill; CEQA will no longer apply to *any* infill housing under 85’. No inclusionary zoning or labor provisions.