Laura Ruis

This is a great paper not only because of the clever control tasks, but also because it relates current evals of LLMs to decades old discussions on psychologism vs behaviorism

Was just reading this awesome paper that's very relevant to this (but you probably already know it ;)) Some syntactic structures and semantic categories can be learned from a part of one child's experience

Great read of a nuanced take that does LLMs current striking capabilities proper justice

Much of this scepticism transfers from epigenetics to other fields - great read