Resolution: Scale and Automation for Higher Confidence in Alignment | Jesse Hoogland (Resolution)

Jesse Hoogland (Resolution) introduces a new AI alignment organization betting on theory plus research automation, backed by $160M. Hoogland announces Resolution, an organization built around a portfolio of higher-confidence "theory bets" combined with research automation, on the view that theory offers a source of ground truth that lets research move faster than benchmark-chasing. Resolution joins the UK AISI alignment team, with Geoffrey Irving as chief scientist, and Timaeus, the singular learning theory group Hoogland co-founded, now its learning-theory division. Its opening bets span learning theory, scalable oversight, a science of personas, complexity theory, and agent foundations. With $160M from Coefficient Giving, the aim is to give safety research the talent density and compute that frontier labs have. He argues the timing works because research automation is starting to pay off and safety funding is growing, and closes with open roles. Chapters 0:00 Introducing Resolution: resolving alignment 0:22 A federated portfolio of theory bets 0:35 Betting on theory plus research automation 0:59 $160M from Coefficient Giving 1:25 Who they are: UK AISI and Timaeus join forces 2:02 The research bets: learning theory, scalable oversight, and more 3:00 Why this structure? Closing the gap with frontier labs 4:15 Why now? Research automation is starting to work 4:49 Safety funding is scaling up 5:14 Hiring: get involved More AI safety research: https://far.ai Alignment Workshop playlist: https://youtube.com/playlist?list=PLBY5kyt_LfFg&si=8Nmp0_f_QlSJkQGi FAR.AI is a research nonprofit working to ensure the safe development of advanced AI. We host the Alignment Workshop series and publish frontier alignment research.