Reaction
How common are human cognitive fallacies in AI?
A user suspects they were initially common because they appear in training data and sound plausible to RLHF reviewers.
TLDR
A user wonders how common human cognitive fallacies are in AI. They suspect the fallacies were initially common because they appear in training data and sound plausible to RLHF reviewers. They also suspect the fallacies are now disappearing because they lower performance.
Combined views
182
1 Source, first seen ago