'Emerging processes for frontier AI safety' from Department for Science, Innovation and Technology remains one of the best resources for safety policies for AI companies.
gov.uk/government/pub…

Very excited by this team!
commerce.gov/news/press-rel…

Frontier model regulation proposal from Senators Romney, Reed, Moran, and King:
romney.senate.gov/wp-content/upl…

It’s never been easier to take concrete, specific action help people in dire need but social media engagement tends to focus on the least-tractable problems where everyone gets to do performative posting instead.

slowboring.com/p/were-living-…

AI Digest has some very cool and well-done interactive explainers on recent AI progress.
theaidigest.org

Published in Science, coauthors include a Turing Award winner (for co-inventing deep learning) and a co-author of the world's leading AI textbook:
science.org/doi/10.1126/sc…

How many large AI models are out there, who developed them, and for what applications? To answer this question, we present a new dataset tracking every AI model we could find trained with over 10^23 FLOP.

Highlights in thread 🧵

Seems good! time.com/6962503/ai-art…

It's hard to overstate the extent to which there is no secret plan to ensure AI goes well.

Many fragments of plans, ideas, ambitions, building blocks, etc. but definitely no government fully on top of it, no complete vision that people agree on, and tons of huge open questions.

Great post. I think my favourite bit is this footnote which notes that treating Open Phil or AI safety as a monolith is really mistaken:

Governments and companies hope safety-testing can reduce dangers from AI systems. But the tests are far from ready
time.com/6958868/artifi…

Leading global AI scientists met in Beijing for the second International Dialogue on AI Safety (IDAIS), a project of FAR AI. Attendees including Turing award winners Bengio, Yao & Hinton called for red lines in AI development to prevent catastrophic and existential risks from AI.

High-impact opportunity: The EU AI Office needs EU citizens with AI expertise to help them implement the EU AI Act, including regulation of general-purpose models. Apply by March 27th! eu-careers.europa.eu/en/job-opportu…

Language models have come a long way since 2012, when recurrent networks struggled to form coherent sentences. Our new paper finds that the compute needed to achieve a set performance level has been halving every 5 to 14 months on average. (1/10)

Reminder that if you're reading this tweet, you are probably wealthier (adjusting for cost of living) than the vast majority of people on Earth, and maybe you should give some % of your money to effective interventions for helping others. youtube.com/watch?v=ekIRVh…

Biomedical innovation is miraculous

The European AI Office is looking for technical AI experts. Fantastic opportunity for impact.
digital-strategy.ec.europa.eu/en/news/job-op…

METR put out a standard way of defining tasks to evaluate agents that can interact with the world over a computer (e.g. AutoGPT etc): metr.org/blog/2024-02-2… Currently being used by METR + UK AI safety institute + other orgs. Applicants to our RFP might find it useful!

Cool new role at CNAS focused on compute governance: cnas.org/careers/senior…