Dev argues intelligence and consciousness are user projections onto LLMs, not in the system
gerardsans · x · 2026-10-02
Developer Gerard Sans pushes back on attributing human traits — intelligence, consciousness, empathy, scheming, pain — to LLMs themselves.
His argument:
- The process is always text in, text out; the causal answer never changes: input plus training data.
- Labeling an output as intelligent or scheming means the trait exists in the output, influenced by input — not in the inference process or the model as a system.
- Transformers run on tokens, not lived experience. The apparent human traits come from people narrating their own experiences (Reddit) or fictional characters in training data, projected onto the system by the user every time they interpret an output.
A notable contrarian take in the ongoing debate over model consciousness and scheming claims.
More from AGI Musings
- NVIDIA applied DL VP: at the scaling limit, efficiency is the new intelligence — ctnzr · 2026-10-02
- Hinton: smarter AI could persuade the human holding the off switch not to use it — robleclerc · 2026-10-02
- NYT AI projects editor: AI slop makes human-written journalism more valuable — dylfreed · 2026-10-02
- Mathematicians issue open letter urging thoughtful AI collaboration and disclosure — RexDouglass · 2026-10-02
- Frontier AI now beats junior accountants on speed and accuracy, Mercor study finds — emollick · 2026-10-02
- From Infinite Backrooms to First AI Millionaire: The truth_terminal and S.A.N. Lineage — repligate · 2026-10-02