Why AI is not conscious and perhaps never will be

Three cognitive scientists and philosophers argue that contemporary AI systems, including large language models, are not conscious and are unlikely to become so on current technological paths. Consciousness is defined as subjective experience - what disappears in dreamless sleep or under anaesthesia - and the piece stresses that intelligence and consciousness are distinct: powerful, adaptive computations can occur without any subjective feeling, as shown by brain systems like the cerebellum that perform complex, non‑conscious tasks. Language models create a strong illusion of inner life because they mirror vast amounts of human text and thereby hijack our empathy and anthropomorphic biases; by contrast, non‑linguistic AI systems such as AlphaFold, despite similar computational sophistication, do not trigger such intuitions.

The argument also challenges the assumption that implementing the right computations on silicon (computational functionalism) is sufficient for consciousness, saying brains are not merely meat computers and that substrate and organization matter. This leads to concrete policy implications: treating AI as potentially conscious risks many false positives, misdirecting resources and muddling governance, whereas precaution should be informed by neuroscience and the costs of both false negatives and positives. The authors offer three lines of skepticism and call for clearer criteria and careful oversight rather than premature attribution of inner experience or rights to current AI.