“You know, I thought in 2020, like we can scale this a…” — Dwarkesh Podcast at 18:52
“You know, I thought in 2020, like we can scale this a bunch more, but I wonder if it's more efficient to scale it more or to start adding on these other objectives like RL. I thought maybe if you do as much RL as, you know, as you've done pre-training for a, you know, 2020 style model, that that's the way to go and…”
0:00
1:58:43