If you're wondering why AI safety people are so worried: there are currently no approaches we know won't break as you increase capabilities, too few people are working on core problems, and we're…

Posters: anthrupad. anthrupad

3 likes recorded by the source corpus.