📢 Are you fed up with the binary discussions between AI existential risks and AI social/ethical risks? Then read my new manuscript 👇🏻. Feedback very welcome. arxiv.org/pdf/2401.07836…

(1/5) Very excited to announce the publication of Bayesian Models of Cognition: Reverse Engineering the Mind. More than a decade in the making, it's a big (600+ pages) beautiful book covering both the basics and recent work: mitpress.mit.edu/9780262049412/…

📢📢Introducing the 1st workshop on Sociotechnical AI Governance at CHI’25 (STAIG@CHI'25)! Join us to build a community to tackle AI governance through a sociotechnical lens and drive actionable strategies. 🌐Website: chi-staig.github.io 🗓️Submit your work by: Feb 17, 2025

I'm very grateful for all the advocates fighting for factory-farmed animals -- and winning. And nothing says gratitude like money. So, this year-end, here are a few animal groups I'm donating to. (Not a complete list: there are many other great groups working for animals.)

a ton of thought went into the UI of Community Notes

My talk from International Association for Safe & Ethical AI on "Dynamic Preferences in AI Alignment: A Deliberative Democracy Lens" is up here video.oecd.org/embed-or-en-ff… First pass sharing these ideas, more fleshed-out write-up to come~

Perhaps AI conferences (e.g., ICML) need to stop people from submitting low-effort course project reports or anything similar to the main conference. These submissions are not a good practice and don’t end well. Surprisingly, I found that half of the papers I reviewed fall into

Check out our #ICLR2025 paper on Context Steering, where we present an inference-time method that modulates the level of contextual inference for controllable generation ✍️ Joint work with Jerry Zhi-Yang He Mariah Schrum Anca Dragan

I am also thankful that, rather than distraction, many researchers are interested in building bridges between the two communities & seeing where, despite differences, we can all make progress on shared concerns! Such as Atoosa Kasirzadeh Seth Lazar David Krueger and many more

I won't be at ICLR this year, but I'm sharing my papers here, including an oral in the main conference and several orals in the workshops. Topics include: distillation, data selection for pre-training, benchmark + provable alg for unlearning, pref learning, and overtraining.

🚨 NEW PAPER: “Democratic AI is Possible: The Democracy Levels Framework Shows How It Might Work” #icml2025 AI is reshaping our world. How should we steer its development? We introduce the Democracy Levels Framework to define concrete milestones toward meaningfully