Is meta-RL in PFC just a big RNN or can we figure out its mechanism? Here we show that RNNs learn a particular, but general, method for meta-RL. They place options in 'ranked’ subspaces (slots), and dynamically re-rank them. Value doesn’t pick the action; it reorders the slots 🧵 (Reposted w/ vids)

Show Post

New blogpost: my lab's AI policies and norms. I'm trying to figure out how researchers should make use of AI to accelerate discovery without letting it replace human understanding or falling off the "slop cliff." Here are some ideas. babieslearninglanguage.blogspot.com/2026/10/my-...

Show Post

How I talk to my kids during zoom meetings

Show Post

Lots of debate this week on whether journals should ban AI in peer review. We actually tested this in our new preprint, and the results were pretty clear to me. TLDR: From now on I won't do any peer review without AI. For clinical trials, I'd even call human-only review dangerous 🧵

Show Post

Sound interesting. But, respectfully, that’s a crazy use of a diverging colourmap.

Show Post