"As countries and companies race to uncertain frontiers, we need an equally urgent effort to ensure AI safety, security, and trustworthiness." !

“AI also brings profound risks… As countries and companies race to uncertain frontiers, we need an equally urgent effort to ensure AI’s safety, security, and trustworthiness… In the years ahead, there may well be no greater test of our leadership, than how we deal with AI.”

This is wonderful, but 2 years later than it should have been. We need to invest in immediate response, not respond to outbreaks after they are established - unlike the way we're mishandling H5N1, for example. Or even better, we can and should be preventing their emergence!

Two years after first suggesting this simple test, AI systems still don't seem to get this joke. Maybe once o1 can do images it will work?

The #AI Pact signatories list is out. Even with significantly diluted pledges, Apple & Meta decided not to join, continuing their collision course with the EU. Mistral & Anthropic are also out, but that makes sense since they are only in the model segment. digital-strategy.ec.europa.eu/en/policies/ai…

Found something I could do with fifteen minutes of research that Claude and GPT-4o don't seem to be able to do in unlimited time: identify all the cereals in this image from Nathan W Pyle art Human superiority restored?

RAND's Meselson Center is seeking candidates of all seniority levels who would like to use their technical and domain expertise to directly impact AI, biosecurity, and cybersecurity policy in government (and beyond). More: rand.org/global-and-eme… Job ads: rand.wd5.myworkdayjobs.com/External_Caree…

👀!

I feel a bit more justified in this after seeing the Ivanka tweet

𝔊𝔴𝔢𝔯𝔫 vindicated, comment from 2024-03-07: lesswrong.com/posts/KXHMCH7w…

Remember when OpenAI's nonprofit board was like "this Altman guy is constantly lying to us and doesn't seem like he takes the nonprofit mission at all seriously" and people called them "clods" and mocked them? It's fun that they were completely right reuters.com/technology/art…

Insane week for permitting reform (really feels like we've reached a tipping point): 1. NEPA exemption for chip fabs passed Congress 2. NEPA reform for wildfire prevention passed House 3. NEPA catex for geothermal passed House 4. And now Harris calls for permitting reform

So we now know that Mira only informed SamA this morning, the same morning we learn OpenAI is to become a B-Corp and abandon its mission, and they're going to lose too more top people, Bob and Barret. Huh.

AI safety has recently gone SUPER mainstream. In just the past two days, we’ve seen: • >100 big names in Hollywood have endorsed SB1047, California’s AI safety bill • Biden spoke at the UN about AI risk • Ivanka Trump told her audience to read Leopold’s Situational Awareness

gwern strikes again

To state the obvious, another example of how advances in technology are enabling more malicious targeting by state (and nonstate) actors. Very concerning overall but especially in targeting politicians on sensitive issues punchbowl.news/article/washin…

Some companies are looking to beat OpenAI and Google to the agents game