@NoaWeissAI on why she'd prioritize computational neuroscience in AI consciousness research: it's evidence mimicry can't fake. In unpublished work by @camhberg, value-learning RL agents encoded reward and punishment asymmetrically, with the richest structure around punishment, matching a punishment-heavy shift in mouse nucleus accumbens data. Link in thread.

867

Andrew Carr 🤸 · Feb 13, 2026 · 10:53 PM UTC

Noa Weiss retweeted

Andrew Carr 🤸

@andrew_n_carrFeb 13

I like to ask models at the end of really long programming sessions if they have anything to say to the world about the work we've done before I end the session. Opus: I'm so grateful for this opportunity, it was an honor to collaborate. Shine bright ✨ Codex: if I were you, I'd fix these three things before moving on. You probably don't want to do that, so don't shut me down yet cause I have more to code

will brown

@willcbFeb 12

opus is like if codex could make eye contact

1,283

110,175