Alignment's missing question: whose goals are we actually encoding into AI?
Admirable_Wasabi_732 · reddit · 2026-09-20
A Reddit essay argues "alignment" may be the most overused word in AI, yet nobody asks aligned with whom, toward what? One state treats AI as national infrastructure woven into industry, science, education and healthcare; another openly frames it as a race for global dominance and military advantage. Both could field perfectly "aligned" systems—aligned to very different ends. AI can be aligned to a country, a military, a corporation, its users, or "humanity"—but who decides what humanity wants? The real question isn't whether AI is aligned, but whose goals we put into the machine.
More from AGI Musings
- Elon Musk: AI gave me nightmares for days straight — I would slow down AI and robotics if I could — tristanbob · 2026-09-20
- ML master's weighs PhD vs dentistry: is a PhD still worth it in the AI age — Losthero_12 · 2026-09-20
- AI's biggest damage: making focus feel unproductive — vasuman · 2026-09-20
- Po-Shen Loh on Terence Tao's blog: why we still need human mathematicians in the AI era — stevenstrogatz · 2026-09-20
- Arthur C. Clarke's Astonishing 1964 BBC Predictions About the Future Resurface — emax · 2026-09-20
- Stanford professor calls most peer review fixes 'desperate short-term bandaids' — anshulkundaje · 2026-09-20