The full shortlist of poets, Adam Panichi, Jam Kraprayoon, Lucy Holme - Roberts, will fleming and Dillon Jaxxpoet, will have their work published in the next edition of the Brotherton Poetry Prize Anthology. Watch this space!

Our launch tomorrow night on Zoom will feature brilliant poets Jam Kraprayoon Jam Kraprayoon , Joan García Viltró Joan García Viltró, Jonaki Ray, JP Seabright JP Seabright 💜 Kris Johnson Kris Johnson, Sodïq Oyèkànmí Sodïq Oyèkànmí 🍁🖤

How much does it cost to develop frontier AI models? Our study (w/ Stanford HAI) finds hardware costs grow 2.4x/year, with training alone in the tens of millions for models like GPT-4. Full development costs (when amortized) are over $100M for the most advanced models. 🧵

New paper: Future AI systems may be capable of enabling offensive cyber operations, lowering the barrier to entry for designing and synthesizing bioweapons, and other high-consequence applications. If these capabilities are discovered, who should know first, and how? More in 🧵

We’ve also published a shorter policy memo in collaboration with Federation of American Scientists🔬, directed toward US policymakers: fas.org/publication/an…

The US Supreme Court has eliminated Chevron deference, an important legal doctrine that required courts to defer to agencies' interpretations of certain laws. We previously discussed Chevron and what its repeal might mean for AI governance on the LawAI blog:

🚨OUT NOW: AI assurance market could generate over £18bn a year for the UK economy by 2030 🧵New report shows the AI assurance market is ripe for UK companies to capture, but government must address key barriers to get companies investing smf.co.uk/publications/a…

There does not have to be a trade off between AI safety and growth. In fact, the AI assurance market could generate over £18bn per year for the UK economy by 2030. Important new paper from UKDayOne, Social Market Foundation & Institute for AI Policy and Strategy (IAPS)

The UK has a growing AI testing + assurance industry. With some changes to regulation and a small amount of public investment, this industry could be worth £18bn to the UK economy. New paper with Social Market Foundation Institute for AI Policy and Strategy (IAPS) from Bill Anderson-Samways @jkraprayoon ukdayone.org/briefings/assu…

The UK is already a global leader in cyber security innovation and commerce. There’s a great opportunity for it to become a big player in the AI testing and assurance industry. Makes sense. Let’s do it. UKD1 paper with Institute for AI Policy and Strategy (IAPS) Social Market Foundation by Bill Anderson-Samways Jam Kraprayoon

How well can LLM agents complete diverse tasks compared to skilled humans? Our preliminary results indicate that our baseline agents based on several public models (Claude 3.5 Sonnet and GPT-4o) complete a proportion of tasks similar to what humans can do in ~30 minutes. 🧵

# RLHF is just barely RL Reinforcement Learning from Human Feedback (RLHF) is the third (and last) major stage of training an LLM, after pretraining and supervised finetuning (SFT). My rant on RLHF is that it is just barely RL, in a way that I think is not too widely

1/ Automating AI research could rapidly drive innovation. But which research tasks are nearing automation? And how can we evaluate AI progress on these? 🧵