In the current consciousness literature, there is no clear answer to the question of whether consciousness can be multiply realized, such that Artificial Intelligence (AI) potentially could be, or even already is, conscious (Bayne et al., Reference Bayne, Seth, Massimini, Shephard, Cleeremans, Fleming, Raphael, Mattingley, Menon, Owen, Peters, Razi and Mudrik2024; Overgaard & Kirkeby-Hinrup, Reference Overgaard and Kirkeby-Hinrup2024; Kirkeby-Hinrup, Stenseke, & Overgaard, Reference Kirkeby-Hinrup, Stenseke and Overgaard2024).

Seth strongly argues that the question of conscious AI is both pressing in light of recent technological advancements and influenced by logical fallacies in our reasoning. He proposes that consciousness is fundamentally biological, best understood in terms of organisms that self-regulate and maintain their own existence. On this view, consciousness is “substrate sensitive” rather than multiply realized – meaning that biological matter and its specific organization are not incidental to consciousness but rather necessary for it.

However, the argument intended to make “biological naturalism” more plausible than the theory of multiple realization appears underdeveloped.

While most would agree with Seth that cognitive fallacies may bias us toward ascribing consciousness in certain situations (e.g., to AI) as a form of anthropomorphism, such bias does not in itself constitute an argument that the attribution is necessarily mistaken in any given case. The very same type of bias Seth refers to is also what leads us to ascribe consciousness to other people and animals – a judgment we must assume is not fallacious. Accordingly, even if cognitive biases influence our judgments about AI, this fact alone is neutral with regard to whether AI is actually conscious.

Seth argues that the functional/structural relations of the brain are fundamentally different from those in artificial systems. However, such a difference only serves as an argument against consciousness in AI if we presuppose that consciousness cannot be realized in a system with a different functional/structural organization – which is precisely the assumption Seth sets out to investigate. The argument for multiple realization is exactly that such differences do not preclude consciousness. Indeed, a proponent of multiple realization would use the same observation – that AI and human systems are functionally and structurally different – as the starting point for their case. This difference is not a counterargument, but the very premise upon which the theory is built.

Seth cites Chalmers’ thought experiment “neural replacement,” after which he refers to several authors who have criticized the idea. He concludes that while this critique does not disprove the possibility of silicon-based consciousness (i.e., consciousness realized in non-biological matter), it does challenge the assumption that such substrate flexibility must be true. He then asserts: “Brains seem to be the kind of thing for which it is hard and perhaps impossible to separate what they do from what they are.” However, this conclusion does not clearly follow from the preceding discussion. The claim remains vague and appears more as an implicit appeal to identity theory than a substantiated argument.

In the end, Seth leaves us with a new name for an old idea – that consciousness depends on biology. He presents several reasons why one might consider this view and why multiple realization is not necessarily true. However, what is needed is an argument why multiple realization cannot be true, or why we should believe biological naturalism over multiple realization. As none of the arguments succeed in any of these goals, the scientific field of consciousness research is in the very same spot after Seth’s presentation of his theoretical position as before.