AI Alignment Isn't Just A Technical Problem
The (brief) argument for broad democratic involvement in making sure AI goes well for humanity.
Towards Provable Alignment
Working through the formal foundations of provably beneficial agents - assistance games, provable guarantees of alignment and modern deep learning primitives.
Subscribe:.rss.atom.json.md.m3u.pls
Live Last read · last published · next check
Latest episodes
Linguistic Assistance Games (Pt. I)
In which I discuss my thinking on how to merge assistance games with formal models of language use and dialogue, in an effort to develop a theory that leads to aligned-by-design language models.