We humans are apophenia machines, built to see agency in noise, faces in clouds, spirits in the rustling of leaves.
Hammond’s core move is simple but radical: treat the brain as a machine running learning algorithms, not a mystical object, and then notice how modern neural nets keep rediscovering the same tricks.
Deep nets and brains converge on similar internal “maps” of the world because there are only a few really efficient ways to compress reality.
If that’s right, then frontier AIs aren’t just faking intelligence, they’re solving the same problems our cortex solves, in structurally similar ways.
@hamandcheese
