"LLMs are mathematically equivalent to a book" — and why that consciousness argument backfires
mathemagic1an · x · 2026-10-04
Josh Albrecht argued that LLMs are pretty obviously not conscious and spelled out why: an LLM is a pure function f(w, s, t) -> o with fixed-size inputs and outputs, placing it in the class f(ℕ) -> ℕ—mathematically equivalent to a book.
Mathemagic1an countered that the same argument applies to any bounded physical system at finite resolution: at the Planck length, humans can also be represented in a book, so by this logic humans aren't conscious either.
The exchange highlights a structural flaw in the "equivalent to a static object" style of argument against machine consciousness.
Related event: AI Researchers Clash Over Whether LLMs Can Be Conscious(12 posts)→
More from AGI Musings
- AI tutors could break education's one-teacher limit, argues viral thread — r0ck3t23 · 2026-10-04
- Anthropic study: robots cost-competitive for just 0.3% of US work hours — VraserX · 2026-10-04
- Chamath: our token bill doubles every 45 days, output up just 5% — bittingthembits · 2026-10-04
- Gary Marcus: 87.5% of popular AI tweets use made-up numbers — a statistic he just made up — GaryMarcus · 2026-10-04
- Whatever Your Take on AI Risks, Someone Will Think You're a Lunatic — burny_tech · 2026-10-04
- Secular moral status for AI: the consciousness-equals-moral-patienthood heuristic, debated — JoshPurtell · 2026-10-04