Is an LLM Just a Book? AI Circle Debates Machine Consciousness

Following Steve Yegge's prediction that "the debate over machine consciousness will intensify," AI researchers engaged in a heated debate on X over whether LLMs could be conscious, centered on two opposing lines of argument.

Confirmed

• Josh Albrecht (joshalbrecht), co-founder of Imbue, offered a negative argument: an LLM is a pure function f(w, s, t) → o (weights, seed and settings, token inputs to output), with no side effects—belonging to the class of functions like f(ℕ) → ℕ—mathematically equivalent to a book, and since nobody thinks a book is conscious, LLMs clearly are not either.
• mathemagic1an relayed and challenged this argument, noting that by the logic of "equivalent to a book, therefore no consciousness," the same argumentative structure could boomerang against human consciousness—skeptics were accused of deploying an argument that could deny consciousness to both machines and humans.
• kevinafischer also joined in, questioning the "LLM as pure function equals a book" reasoning and steering the debate toward the more fundamental question of "whether consciousness is merely a product of processes."
• Josh Purtell (JoshPurtell) responded from a functionalist and intuition-based angle: functionalists would grant consciousness to a machine that faithfully simulates a human brain, but such a "rock computer" would be so vast that intuition breaks down—just as many ordinary people feel the sun might also be conscious; yet an analog LLM is small enough that intuitively it feels closer to a device like the Antikythera mechanism, which "clearly isn't conscious."
• Purtell further offered an intuition ranking: Is the sun conscious? > Is a simulated brain conscious? ≫ LLMs aren't conscious, and pointed out that moving from "a simulated machine replicating a brain is conscious" to "any random simulated machine doing amazing things is conscious" is a logical non sequitur.

Why it matters

• The debate comes against a backdrop of rapidly improving model capabilities, with machine consciousness shifting from philosophical speculation to a public topic in AI circles, directly affecting judgments about models' moral status, safety, and alignment.
• Albrecht's "pure function/book" argument represents an attempt to sharpen the negative position, but critics exposed its cost: if the bar is set too strict, even human consciousness could be denied, revealing the deep difficulty of demarcating consciousness via mathematical equivalence.
• Purtell's contribution was to introduce ordinary intuitions as a ranking, showing that functionalism admits a self-consistent reading that accommodates simulated-brain consciousness while denying LLM consciousness, providing a framework for future debates.

2026-10-03 ~ 2026-10-04 · 9 related posts

• [source] LLMs Are Mathematically Equivalent to a Book, Argues Imbue Cofounder in Consciousness Debate — joshalbrecht · 2026-10-03
• Is consciousness an event, not a thing? AI circle debates whether LLMs can ever be conscious — kevinafischer · 2026-10-04
• [source] "LLMs are mathematically equivalent to a book" — and why that consciousness argument backfires — mathemagic1an · 2026-10-04
• Consciousness debate: exact brain replicas in analog may be conscious, but random analog machines are a non sequitur — JoshPurtell · 2026-10-04
• Josh Purtell argues a stone computer emulating a brain would break our consciousness intuitions — JoshPurtell · 2026-10-04
• Machine consciousness debate: Purtell ranks sun, analog brain, and LLM by intuitive consciousness — JoshPurtell · 2026-10-04
• [source] Could a stone-gear mind be conscious? AI circles debate substrate and scale — JoshPurtell · 2026-10-04
• Lock and key or supernova? Where LLMs sit on the consciousness spectrum — JoshPurtell · 2026-10-04
• Locks vs supernovae: the debate over whether LLMs are anywhere near consciousness — JoshPurtell · 2026-10-04