Dewi Gould

🚨 New blog post 🚨 Collab between the new scalable oversight team at @ArcadiaImpact, @jacob_pfau and Simon Marshall.

@redwood_ai @ConstellOrg @MATSprogram @AetherAIS LW post: lesswrong.com/posts/SieLowPg…

Thanks to the following orgs for supporting this word: @redwood_ai @ConstellOrg @MATSprogram @AetherAIS

New paper! Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models @METR_Evals showed that models' time horizons have doubled every few months. We ask: what length of tasks can models complete without any CoT?

*NEW* AI alignment research team! We're announcing the new alignment team @ArcadiaImpact. A London-based team, working closely with @AISecurityInst to tackle 3 ambitious agendas in AI alignment! 👇 🧵

🚀 Applications are now open: Constellation's Astra Fellowship 🚀 Fully funded, 5-month fellowship at our Berkeley research institute. Pair with mentors across empirical AI safety research, strategy, and governance at @ConstellOrg! 📅 Apply by May 3rd (begins Sep 2026) 🔗 constellation.org/programs/astra…

New paper. A Positive Case for Faithfulness. When asked to explain their decisions, LLMs can give highly plausible self-explanations. But are these explanations actually faithful, or are they just post-hoc rationalizations? We measure faithfulness via simulatability.

Big thank you to amazing collaborators @HarryMayne5 @Justinkangs and SPAR mentor @noahysiegel !

Paper: arxiv.org/pdf/2602.02639 LessWrong: lesswrong.com/posts/Y4MJRniZ…

New paper: A Positive Case for Faithfulness. When asked to explain their decisions, LLMs can give highly plausible self-explanations. But are these explanations actually faithful, or are they just post-hoc rationalizations? We measure faithfulness via simulatability.