The "alignment won't generalize out of distribution" argument should obviously apply to humans as well and it's bizarre to me that Soares et all pretend like it doesn't.

Posters: jd_pressman. jd_pressman

2 likes recorded by the source corpus.