prediction claim
Optimizing with reinforcement learning against an LLM judge will almost certainly produce adversarial examples that exploit the judge.
Andrej Karpathy · Andrej Karpathy — AGI is still a decade away · 17 Oct 2025
In their words
The transcript segments this claim was taken from
“And if you reinforce my learning with respect to them, you will find adversarial”
“examples for your LM judges,”
“almost guaranteed.”Play from 46:21Report a misattribution
95%how strongly it was put, not whether it is true
95%importance to the conversation
97%confidence the claim was read correctly