Anil Seth: Silicon AI Almost Certainly Unconscious, Simulation Is Not Instantiation

Consciousness researcher Anil Seth delivered a clear verdict in his TED 2026 talk (over 480,000 views): silicon-based digital AI is extremely unlikely to be a conscious system. The debate over "whether AI is conscious" then spilled onto X, sparking several technically grounded exchanges and becoming a focal point in recent discussions of AI and consciousness.

Confirmed

• Seth compared people's perception of an AI's inner life to "seeing faces in the clouds": humans tend to project inner life onto machines that are good at mimicking, but this does not mean the machines actually have subjective experiences.
• In discussions with others, Seth laid out the core argument of his paper: simulation is not instantiation — matching inputs and outputs does not mean possessing consciousness.
• Responding to Roger Highfield (Science Museum), Seth proposed that the further a biological system departs from standard digital computation, the worse its substrate flexibility for computation, with two consequences: such computations are less likely to be realizable in silicon, and this weakens computational functionalism (the view that the right computation alone suffices for consciousness).
• Highfield and Seth discussed why the "digital vs. analog" perspective rarely comes up in discussions of the brain, AI, and consciousness.

Controversies and Responses

• Some challenged the "simulation ≠ instantiation" distinction by arguing that "simulations can simulate each other to arbitrary precision"; Tim Tyler likewise noted that a simulated system can emulate a digital system to arbitrary precision, adding a new angle to the consciousness debate. Seth responded to this (see m2 for details).
• gerardsans argued that the AI consciousness debate "has gone off track": the harder problem is people who cannot shake anthropomorphic bias — those who form intimate relationships with chatbots and start trusting AI over their own judgment, while the system has no self at all.

Why It Matters

• Seth's argument directly challenges the popular intuition that "sufficiently complex AI will become conscious." If it holds, it has major implications for debates about AI welfare and AI rights; meanwhile, the problem of anthropomorphic dependence suggests that regardless of whether AI is conscious, psychological and social risks at the user level already exist.

2026-09-05 ~ 2026-09-06 · 5 related posts

• [source] Neuroscientist Anil Seth: silicon-based digital AI is vanishingly unlikely to be conscious — anilkseth · 2026-09-05
• The real AI problem isn't consciousness, it's chatbot dependency, researcher argues — gerardsans · 2026-09-05
• [source] Anil Seth: departing from digital computation undermines computational functionalism — sebkrier · 2026-09-05
• Digital vs analogue: why the brain-AI consciousness debate ignores simulation equivalence — tim_tyler · 2026-09-06
• [source] Anil Seth argues simulation isn't instantiation — same I/O doesn't imply consciousness — anilkseth · 2026-09-06