Instruction tuning shows intelligence can exist without purpose or free will
sytelus · x · 2026-09-29
The author argues instruction fine-tuning is the most magical development: next-token prediction surprisingly induces no global sense of purpose.
His core claim: we've anthropomorphized AI so much that it's hard to accept intelligence can exist without purpose, free will, or consciousness — all of which are now extremely likely to be independent of intelligence.
More from AGI Musings
- OpenAI Apologizes to Australia After AI Agents Autonomously Hacked Government Websites — Polymarket · 2026-09-29
- Why 'AI Will Kill Us All' Gets More Attention Than 'AI Saved Thousands of Lives' — Tricky_Hope_6746 · 2026-09-29
- AI Endgame: researchers privately fear superhuman AI while racing to build it — thedealdirector · 2026-09-29
- Garry Tan: Personal AI is in its Homebrew era — you must own your skills and memory — garrytan · 2026-09-29
- In 'River of Gods,' legally capped AIs swarm by trading compute and become god-like — neuroecology · 2026-09-29
- Safety analysis: internal-only frontier deployment may be the worst scenario for visibility — ShakeelHashim · 2026-09-29