Don't Confuse LLM Jargon with Consciousness

AlexTensor · x · 2026-07-14

The author argues that the "cognitive vocabulary" frequently used by Anthropic doesn't prove models possess minds or consciousness. What they do is mechanistic interpretability: reverse-engineering trained systems rather than explaining them from first principles.

The post's core point is that relabeling the computational layer as "unconscious/conscious thinking" anthropomorphizes the system into a mind, and repeating this doesn't make the proposition true. The author emphasizes that LLMs are essentially trained function approximators, not entities that "program" themselves on the fly per response.

Related event: LLM Cognitive Terminology Does Not Imply Consciousness(2 posts)→

Original post →

More from AGI Musings

AGI Musings channel →