causal claim
A model distilled only on easily verifiable tasks performs worse on a broader, more realistic task distribution.
John Schulman · AI researchers debate how close we are to recursive self-improvement · 11 Sept 2026
In their words
The transcript segments this claim was taken from
“But you do worse on this broader distribution.”Play from 27:05Report a misattribution
85%how strongly it was put, not whether it is true
75%importance to the conversation
94%confidence the claim was read correctly