“She has a piece about, I think it's called something like…” — Dwarkesh Podcast at 2:12:49
“She has a piece about, I think it's called something like the default, but the default outcome of training AI. Without specific countermeasures. Without specific countermeasures, default outcome is AI takeover. But yes, so basically we are training models that for some reason vigorously pursue a higher reward or…”
0:00
2:44:16