Jacy Reese Anthis

jacyanthis.bsky.social

15 September 2025 17:54

LLM agents are optimized for thumbs-up instant gratification. RLHF -> sycophancy We propose human agency as a new alignment target in HumanAgencyBench, made possible by AI simulation/evals. We find e.g., Claude most supports agency but also most tries to steer user values 👇 arxiv.org/abs/2509.08494

Jacob Haimes

jacobhaimes.bsky.social

15 September 2025 17:54

Super happy to share HumanAgencyBench, which takes steps towards understanding the impact of chatbot interactions on huma agency. Working with @jacyanthis.bsky.social (and the team) has been fantastic, and I'd happily do it again. If you have the chance to work with him, don't pass it up!

Into AI Safety

intoaisafety.bsky.social

24 June 2025 00:36

👀 New Into AI Safety episode is live! Will Petillo from PauseAI joins to discuss the grassroots movement for pausing frontier AI development, balancing diverse perspectives in activism, and why meaningful AI governance requires both political engagement and public support kairos.fm/intoaisafety...

thegermanpole.bsky.social

27 January 2025 22:13

Probably our best episode yet, Chris was an awesome guest and is a true Mensch