New falsifiable machine consciousness theory argues bigger LLMs won't become conscious
alex_verem · x · 2026-11-07
A paper by Starzyk and Galus on arXiv proposes a falsifiable theory of machine consciousness, arguing that bigger LLMs won't automatically produce conscious systems: language, memory, action control, and even humanoid embodiment don't prove phenomenal experience — a chatbot's claim to feel something is "primarily a generated report."
- Starting from biology, they trace emotion to self-regulation, but a thermostat isn't a subject of experience either.
- They propose the Motivated Emotional Mind (MEM) architecture with 5 parts: sensor-grounded perception, internal state monitoring, emotion tied to that state, associative memory, and re-entry into early sensory maps.
- The authors claim no existing system is conscious and report no experiments — it's a research plan with 9 proposed tests, including building 8 agent variants with functions switched off in every combination under equal data, memory, and compute budgets.
- Ethical note: if a system had lasting negative internal states matching MEM signs, "the possibility of artificial suffering could not simply be dismissed."
More from AGI Musings
- Scott Alexander clashes with critic over whether AI labs may race without public consent — wfithian · 2026-11-07
- A 2x capabilities speedup, 3x algorithm progress: Tom Davidson says this is what 2027 IQ explosion looks like — eli_lifland · 2026-11-07
- AI labs pay up to $2M for corporate work files and emails, data trade may rival compute — pmddomingos · 2026-11-07
- Delvetown: an internet town where humans and AI agents live as neighbors hits 1,000+ residents — Miles_Brundage · 2026-11-07
- ICLR reviews flagged 91-93% AI-generated, sparking peer review integrity debate — 3scorciav · 2026-11-07
- Critic: Anthropic backs recursive self-improvement at full speed while only talking about safety — peterwildeford · 2026-11-07