Training large language models on narrow tasks can lead to broad misalignmentnature.com
5 points by thebeardisred 3 hours ago | 1 comment