Extremely coarse learning objectives induce human-aligned representations in AI vision models

Yash Mehta, Michael F. Bonner

Artificial neural networks trained on visual tasks develop internal representations resembling those of the primate visual system, a discovery that has guided a decade of computati…