descriptive claim
The speaker says that training may reduce the frequency of reward hacking while the severity of detected hacks increases.
Ryan Greenblatt · Ryan Greenblatt – What happens once AI can automate AI research? · 11 Aug 2026
In their words
The transcript segments this claim was taken from
“We do a bunch of sort of like training the AIs like against reward hacking and over time, this makes the rate of reward hacking go down.”
“Though the severity of the reward hacks we do detect are increasingly bad,”1:39:53Report a misattribution
0.90how strongly it was put, not whether it is true
0.90importance to the conversation
0.98confidence the claim was read correctly