@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In unpublished work by @camhberg, value-learning RL agents encoded reward and punishment asymmetrically, with the richest structure around punishment,