Dev argues AI consciousness is observer illusion: inference is just conditioned sampling over compressed text
gerardsans · x · 2026-10-22
- gerardsans takes a practical stance in a discussion on AI consciousness: claims that AI has it usually fall into well-documented cognitive biases — the illusion lives in the observer's imagination, a phenomenon psychology documented decades ago.
- On mechanism, he argues there's no answer yet for what could support consciousness, but we do know what today's AI is: a text sampler doing conditioned input sampling over a compressed corpus. Inference isn't a mysterious process.
- First-person voice, RLHF and the helpful-assistant persona are lab-set instructions anyone can alter or remove via API; output traits are not system traits. He adds that AI can simulate a dialogue partner but can't become one, and can write as Harry Potter without becoming the character.
More from AGI Musings
- Eight months on, Citrini's '2028 Global Intelligence Crisis' essay looks prescient — teortaxesTex · 2026-10-22
- Vulnerabilities were always there — AI is just better at finding them — jonmarkgo · 2026-10-22
- OSVerify referees 11 OpenAI math preprints, catching 5 major and 13 minor issues — james_y_zou · 2026-10-22
- 1,000 mathematicians pledge to shun AI in their research — Anen-o-me · 2026-10-22
- If your negotiation AI was built on Western values, is your sovereignty compromised? — 0xsachi · 2026-10-22
- Economist: AI writing volumes will inevitably crowd human readers out of equilibrium — paulnovosad · 2026-10-22