Gwern Branwen - Footnote 34
The source only mentions Gwern's discussion of China as a "fast follower" in AI races. It does not mention his exploration of whether AI automation speeds capabilities over safety, or his questioning of whether automated alignment research precedes takeoff. The source also does not mention meta-alignment failures in RL where agents pursue corrupted utility functions, reducing original value.
Our claim
entire recordNo record data available.
Source evidence
1 src · 1 checkNoteThe source only mentions Gwern's discussion of China as a "fast follower" in AI races. It does not mention his exploration of whether AI automation speeds capabilities over safety, or his questioning of whether automated alignment research precedes takeoff. The source also does not mention meta-alignment failures in RL where agents pursue corrupted utility functions, reducing original value.