The framing of AI alignment as a single problem to solve
An author shares a chapter they describe as a structured way to think about AI alignment’s many facets, challenging the framing of alignment as either “solved” or “unsolved.”
TLDR
Debating whether AI alignment is “solved” or “unsolved” misses the point that there is no single alignment problem, an author argues. They credit other researchers with making that distinction before and share a chapter offering another structured way to think about alignment’s many facets.
Combined views
—
1 Source, first seen 17d ago