The funniest way orthogonality could be true: human values are the especially stupid ones
jessi_cata · x · 2026-10-10
A witty tweet on the orthogonality thesis in AI alignment: the funniest way it could hold is if there are especially stupid values (which even Bostrom concedes exist — ones incompatible with sufficiently high intelligence), no especially smart values, and human values happen to be the especially stupid kind.
One line that surfaces an implication the orthogonality debate often avoids: once intelligence and values decouple, humans may hold no advantage on the value axis at all.
Related event: An Ironic Twist on the Orthogonality Thesis(2 posts)→
More from AGI Musings
- NYU math professor: OpenAI poured millions in compute into unsolved math problems — rohanpaul_ai · 2026-10-10
- Where did Hacktoberfest go? Devs wonder if AI killed open source's sense of community — _jaydeepkarale · 2026-10-10
- Skip the PhD — not because nothing's left to discover, but because academia is the wrong institution — tensorqt · 2026-10-10
- If your dog is conscious, why does the AI consciousness debate assume humans only? — RileyRalmuto · 2026-10-10
- Software is like food: on making users love the people behind it — pixlpa · 2026-10-10
- Pedro Domingos: Superintelligence Will Be Commonplace — pmddomingos · 2026-10-10