If interpretability research goes well, it may get dangerous

By So8res @ 2023-04-03T21:48 (+33)

This is a crosspost, probably from LessWrong. Try viewing it there.