Comment by Howard Stark (howard-stark) on The Natural Abstraction Hypothesis: Implications and Evidence · 2022-06-15T23:00:03.879Z · LW · GW

There are definitely interesting "convergences" that provides some evidence for a weak form of this claim. See https://arxiv.org/abs/2202.12299 :    Capturing Failures of Large Language Models via Human Cognitive Biases https://www.biorxiv.org/content/biorxiv/early/2020/03/05/860759.full.pdf : Do deep neural networks see the way we do?

https://www.youtube.com/playlist?list=PLyGKBDfnk-iDWTxSdeGkSA2DYNc7enyhv  some of the work in this conference