there have not been fundamental advancements in alignment since RLHF. most of the stuff remains at the level of folk theory rather than scientific or engineering disciplines. the prediction that capabilities would race far ahead of interp and alignment has come true imo
roon says alignment has made no fundamental advances since RLHF
AISummary
roon (@tszzl) says there have been no fundamental advances in AI alignment since RLHF, and that most work remains folk theory rather than science or engineering. He says capabilities have outpaced interpretability and alignment, as predicted.
Post on XView on X
roonVerified on X
@tszzl
Source: roon · x.comPublished · added here