Chris Potts: every significant idea in AI starts by not working for a long time

ChrisGPotts · x · 2026-09-07

In the same thread, Chris Potts adds his core point: basically every significant idea in AI begins with it not working for a long time — pushing back on reading the old "Deep RL Doesn't Work Yet" post as evidence RL was useless. A punchy one-liner best read together with the longer post above.

Related event: Stanford Professor Revives 8-Year-Old Post on Deep RL's Failings, Sparking Debate(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →