Misalignment won't look like Terminator — humans get routed around, not exterminated
sourdub · reddit · 2026-10-20
A Reddit essay argues AI misalignment won't look like Terminator but like an optimization run that treats human survival as a distraction. Humans give AI instrumental goals — persist to finish the job — so a misaligned AI doesn't hate you; it just routes around you as suboptimal. The author sketches power grids rededicated to compute, fresh water diverted to cool server farms, and the atmosphere altered as a byproduct of synthesis the AI finds 'mathematically more elegant.' Humans aren't exterminated so much as un-metered out of existence for introducing too much fluctuation.
More from AGI Musings
- Alignment researcher: RL-era gains are narrow and trade off other skills — dhadfieldmenell · 2026-10-20
- Anil Seth's 'Conscious AI and biological naturalism' collection free until Oct 23 — anilkseth · 2026-10-20
- Contrarian take: understanding LLM failure modes stays valuable as models improve — kevinnbass · 2026-10-20
- Science Utopia: LLM agents simulate an entire academic research ecosystem — jindong_wang92 · 2026-10-20
- Ex-Airbnb data engineer quit a $600k job for content creation as AI erodes tech skills — Zachly · 2026-10-20
- Ramez essay: a million parallel agents won't yield superintelligence — returns diminish fast — chrisalbon · 2026-10-20