AIAI reward circuits and parallels with biology
One post calls biological parallels overhyped and argues that reward circuitry is unsurprising in a large neural network trained for general tasks.
TLDR
A post describes AI reward circuitry as “incredibly obvious convergent evolution,” arguing that some reward circuit is to be expected in a giant, dense neural network trained for general tasks. It calls the comparison with biology overhyped, but stresses that this does not imply the pace of AI development is safe.
Combined views
1.2K
2 Sources, first seen 5h ago
19 likes
AI reward circuits and parallels with biology
One post calls biological parallels overhyped and argues that reward circuitry is unsurprising in a large neural network trained for general tasks.