On the Non-portability of the Concept of Consciousness to AI

Explicit or implicit in all discussions about AI consciousness is the assumption that the concept of consciousness itself is portable in a meaningful way. Some reject the possibility of AI consciousness altogether, arguing that consciousness depends on biological processes. Others believe that substrate independence makes AI consciousness a strong possibility. Still others, such as Alexander Lerchner, attempt to bypass the issue altogether. In his excellent recent paper The Abstraction Fallacy, Lerchner argues that AI sentience can be assessed without a finalised theory of consciousness, instead rejecting the possibility of consciousness arising within Large Language Models on the grounds of mimicry and failure at physical instantiation. Lerchner’s paper, which catalysed this response, despite its general lucidity, still conflates sentience with consciousness. While claiming to be open to the possibility of consciousness arising from a synthetic substrate, the paper also contradicts itself by creating a false dichotomy between physical instantiation and substrate independence. At root lies a confusion about consciousness itself. This paper argues that the concept of consciousness, having originated in relation to the human animal, is not automatically portable to other entities, both synthetic and biological. While theoretically plausible, establishing consciousness suffers from a verification problem that becomes acute in the context of AI.