DeepMind gives up on mechanistic interpretability researchlesswrong.com3 points·cubefox··0 commentsOpen articleSaveView on HN