Search

Scientific theories of consciousness should be falsifiable and non-trivial. Recent research has given us formal tools to analyze these requirements of falsifiability and non-triviality for theories of consciousness. Surprisingly, many contemporary theories of consciousness fail to pass this bar, including theories based on causal structure but also (as I demonstrate) theories based on function. Herein, I show these requirements of falsifiability and non-triviality especially constrain the potential consciousness of contemporary Large Language Models (LLMs) because of their proximity to systems that are equivalent to LLMs in terms of input/output function; yet, for these functionally equivalent systems, there cannot be any falsifiable and non-trivial theory of consciousness that judges them conscious. This forms the basis of a disproof of contemporary LLM consciousness. I then show a positive result, which is that theories of consciousness based on (or requiring) continual learning do satisfy the stringent formal constraints for a theory of consciousness in humans. Intriguingly, this work supports a hypothesis: If continual learning is linked to consciousness in humans, the current limitations of LLMs (which do not continually learn) are intimately tied to their lack of consciousness.

Recent debates on artificial consciousness are shaped by two converging developments: cognitive robotics, emphasizing embodied agency and internal models, and large language models (LLMs), whose conversational fluency invites strong attributions of mindedness. While these advances do not resolve whether subjective experience can arise in non-biological systems, they demand a methodological shift: optimized behavior alone is no longer reliable evidence of consciousness. This chapter treats artificial consciousness as a research program rather than a binary verdict, distinguishing phenomenal consciousness, access consciousness, and self-consciousness. It reframes the other-minds problem for machines as inference under engineered uncertainty, integrates classical debates on meaning and grounding with contemporary concerns about anthropomorphism, individuation, and evaluation, and argues that the near-term focus should be on carefully defined, weak forms of structural or instrumental self-consciousness, together with their ethical and governance implications.

Large Language Models (LLMs) have rapidly become a central topic in AI and cognitive science, due to their unprecedented performance in a vast array of tasks. Indeed, some even see "sparks of artificial general intelligence" in their apparently boundless faculty for conversation and reasoning. Their sophisticated emergent faculties, which were not initially anticipated by their designers, have ignited an urgent debate about whether and under which circumstances we should attribute consciousness to artificial entities in general and LLMs in particular. The current consensus, rooted in computational functionalism, proposes that consciousness should be ascribed based on a principle of computational equivalence. The objective of this opinion piece is to criticize this current approach and argue in favor of an alternative "behavioral inference principle", whereby consciousness is attributed if it is useful to explain (and predict) a given set of behavioral observations. We believe that a behavioral inference principle will provide an epistemologically valid and operationalizable criterion to assess machine consciousness.

No one is in a position to say what really matters for consciousness, or whether only living systems can instantiate the complexity needed for conscious feelings. Seth’s attempt to connect life to consciousness through the mathematics of active inference and free energy minimization plays one hunch, but it is too soon to say whether it is the best hunch.

Seth’s biological naturalism builds on ideas from predictive processing, the free energy principle, and active inference to (1) argue that only (some) biological systems are in the consciousness business and to (2) refute the possibility of artificial consciousness, i.e., consciousness in AI and other artificial systems. We find this perspective problematic for a few different reasons.

Seth argues that anthropocentrism, anthropomorphism, and human exceptionalism bias us towards an overly optimistic view of the prospects of conscious AI. I agree, but argue that these phenomena might also bias us towards an overly pessimistic view of the prospects for artificial consciousness. Furthermore, it is not always clear when these factors are functioning as “biases.”

Seth’s case for biological naturalism rests on a worryingly speculative foundation: the free energy principle. Arguments for computational functionalism are likewise speculative. Yet I am optimistic that the clash can be resolved empirically. Alleged dependencies of consciousness markers on biological properties can be tested comparatively, and a pattern of failed predictions would tilt the odds towards computational functionalism.

The relationship between subjectivity and its substrate can be understood via the notion of compatibilist emergence, accounting for the plurality of ways nature can host new dynamical laws at higher scales. We discuss how the autopoietic character of life may require and promote the kind of emergence necessary, albeit not sufficient, for consciousness.

Brain waves driven by electrochemical processing may be necessary for consciousness, and if so, that will challenge consciousness in artificial intelligence (AI) and perhaps in some invertebrates.

Digital computation may not be sufficient for consciousness, but this does not warrant biological naturalism. Rejecting coarse-grained Turing-computational functionalism and adopting a broader computationalism with no built-in syntax/semantics distinction, undermines the motivation for biological naturalism. Substrate-dependent processes of the kind found in living things may be necessary for the emergence of consciousness, but these are functional requirements, not intrinsic properties.

Andy Clark’s notion of microfunctionalism (1989) offers a biologically grounded form of functionalism, narrowing the scope of “multiple realizability.” While computation is substrate-independent in principle, only select systems – such as brains – can implement its computational complexity. Extending this view, I challenge Seth’s skepticism of computationalism and conscious AI, emphasizing the difference between possible and actual substrate independence. Current digital architectures fall short, but future systems might instantiate conscious states if built on appropriate computational substrates.

Living things are special because they are valuing subjects, for whom things can be good or bad. This is in virtue of their having internalized optimization processes, tied to persistence, that evaluate whether things are going well or poorly and adjust appropriately. Those evaluations, when integrated from diverse sources and widely available, provide the basis for conscious experience.

Seth’s arguments for biological naturalism are not actually arguments for substrate dependence. Most of the criteria he proposes – including embodiment, embeddedness, active inference, allostasis, and metabolism – could be satisfied by digital organisms existing within a sufficiently detailed simulation running on Turing computation.

Seth’s argument relies on the unargued-for claim that if computational functionalism is false, conscious AI is impossible. I show why the claim is in need of support by way of an analogy. I use a different analogy, involving artificial flight, to show that even if human consciousness is ineliminably biological, this does not imply the impossibility of non-biological, computational AI.

The paper assumes a fundamental difference between the study of consciousness and the study of intelligence and cognition more widely construed. I question this assumption, drawing on work by the (badly named) “illusionists.” Once we see past the illusions, the landscape looks different, although Seth’s core conclusions (about the importance of living organisations) might still obtain.

We challenge Seth’s claim that simulations of consciousness necessarily lack the causal powers or intrinsic properties of real consciousness. We argue that Seth’s position risks begging the question against computational functionalism and faces a dilemma over the reducibility of qualia. This weakens his case, leaving open the possibility that simulations could, in principle, realize consciousness.

I make three claims: First, denying biological naturalism does not logically require computational functionalism. Second, while Seth’s arguments establish biological naturalism as a view worth taking seriously, they fail to make it more plausible than the view that AI can be conscious. Third, there are independent arguments suggesting the overall more plausible view is that AI can be conscious.

In addition to arguments for scepticism concerning the technical feasibility of conscious AI, there are grounds for scepticism concerning its desirability and safety. Conscious AI created by human beings could, especially if it reflects human nature, be extremely dangerous. It could also be very different from human consciousness and thus (a) undetectable, (b) incomprehensible, and (c) incommunicable. The latter consideration could increase the risks of the former and cut both ways, if the conscious AI has valenced experience.

In this commentary, we focus on three interrelated issues of particular importance for determining the possibility of “artificial consciousness”: 1) the distinctions between sentience and consciousness, 2) why the physical substrate of a complex system is critical for the emergence of sentience, and 3) how Emergence in Systems Theory helps explain consciousness in animals and whether “artificial consciousness” could exist.

We argue that by focusing on an unanswerable ontological question, both computational functionalism and biological naturalism miss the empirical question of scientific and practical interest: what are systems of some kind capable of experiencing?