Jeff Ladish: concerned people studying rogue AI failures shouldn't all leave the labs

JeffLadish · x · 2026-09-10

Jeff Ladish adds nuance to his argument on lab talent: it's weaker for pretraining team members to stay if they believe catastrophe is likely, but it seems unwise for researchers studying rogue AI failures to all leave, since some crucial research like interpretability can only be done inside labs.

Related event: AI safety researchers debate whether staying inside frontier labs is justified(6 posts)→

Original post →

More from AGI Musings

AGI Musings channel →