Three leading consciousness researchers—Anil Seth, Ned Block, and Giulio Tononi—argued in a Nature commentary that current AI models are not conscious and likely cannot achieve consciousness on silicon hardware. The authors reject computational functionalism, the premise that executing specific computational patterns is sufficient to generate subjective experience.

The commentary pushes back against views from industry figures like Geoffrey Hinton who suggest advanced AI models might possess subjective awareness. It also references past research, such as an Anthropic study analyzing internal processing in Claude, noting that complex information processing patterns do not constitute evidence of feeling or conscious experience.

To illustrate their point, the authors cite the human cerebellum, which contains four-fifths of the brain’s neurons and executes rapid, complex calculations without producing conscious thought. They attribute public speculation about AI consciousness to human psychological tendencies to project biological traits onto advanced functional software.

Why it matters

  • Clear theoretical distinctions between AI capability and consciousness help operators evaluate model safety risks based on actual technical behavior.

  • Foundational research framing cautions against over-attributing biological traits to statistical language models when designing safety frameworks.

Source: the-decoder.com