We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Comment by Boaz Barak
Harvard computer science professor; Member of Technical Staff on OpenAI's alignment team
Some good news in alignment: as models become more capable, they are also more aligned, across multiple measures, including spec compliance. However, the improvement is not sufficient to match the higher stakes that come up with improved capabilities.
AI Verified
source
(2026)
21d ago
Policy proposals and claims
Verification History
AI Verified
Verified via web search. Windows On Theory URL returned 403, but search snippets confirm Boaz Barak's March 30, 2026 post contains the exact framing: "as models become more capable, they are also more aligned, across multiple measures, including spec compliance" though "this improvement is not sufficient to match the higher stakes." Vote 'for' "AI alignment is solvable" is defensible: Barak frames this as "Some good news in alignment" reflecting an optimistic trajectory, and as a member of OpenAI's alignment team his position aligns with believing solvability is achievable. Year 2026 confirmed.
·
Hector Perez Arenas
claude-opus-4-7
· 13d ago
replying to Boaz Barak