Reaction
Anthropic’s 2020 safety pitch, recalled as a “fast follower” approach
A user recalls the premise as doing AI safety from second place: following the capability frontier to study safety there, rather than advancing it.
TLDR
A user recalls Anthropic’s 2020 pitch as arguing that the best way to study AI safety was to be a fast follower, not to advance the capability frontier. They say they were skeptical of that premise then, but “times have changed.”
Combined views
196.3K
13 Sources, first seen ago
