The target article by Seth argues against certain forms of computational functionalism and makes a strong case that current approaches to artificial intelligence will not be able to create conscious systems. Seth advocates biological naturalism instead, emphasizing the importance of embedded, embodied agency and arguing that consciousness may arise from our nature as living beings with physical bodies. In this, we are in agreement. Yet, a critical piece of Seth’s argument is missing, without which his conclusions are seriously weakened: the nature and role of the conscious subject.

Seth’s arguments depend on the self – he says conscious experience occurs for “self-sustaining organisms” that maintain a “dynamic self/other boundary” – but what he means by “self” is unclear. It does not appear to be a mere organism, since he rejects biological universalism. It appears conceptually distinct from a mere body, since he says we experience the self “through, and because of” the body. The closest description seems to be “living, flesh and blood ‘beast machines.’” A problem with this description is it treats the self as object, rather than subject, but the critical form of self with respect to consciousness is surely the subject (Prinz, Reference Prinz2011).

A subject is a living being with interests (e.g., Kawade, Reference Kawade2009; Jennings, Reference Jennings2020; Mitchell, Reference Mitchell2023). A subject in this sense cannot be simply tacked on to Seth’s current view. Motivations and interests are not predictions, and subjects are not Markov blankets; we don’t currently have a way to map one onto the other. Seth says that “real artificial consciousness might endow a system with its own interests,” but it seems to us that subjects with interests are conceptually (and evolutionarily) prior to consciousness. Similarly, he says “conscious contents are given by the brain’s best guesses” but does not say to whom they are given. Without an account of the self as subject, it is unclear what features of our biological system are essential for consciousness and whether those features require organic material.

Subjects like us are “enminded,” and not just embodied. Our conscious experiences are grounded in our own ongoing mentality – in what it’s like to be us, psychologically. This encompasses all of our memories, attitudes, dispositions, habits, policies, projects, and interests that make us who we are. When we experience something consciously, we are experiencing being ourselves having that experience (Zahavi, Reference Zahavi2008). While we can treat ourselves as objects by taking an objective stance, subjectivity is fundamentally about an irreducible first-person stance, and this is what is central to consciousness (Ganeri, Reference Ganeri2012).

In our view, phenomenal subjectivity rests not on specific computational functions or particular biological processes or substrates, but rather on relations. These include relations between (i) subject and world, (ii) levels of information processing within the subject, and (iii) “temporal parts” of the subject (past and future selves).

Unlike most current AI systems, a conscious agent must be able to both perceive the world around it and act on it. Without this interactive relation, all that exists is passive and abstract information processing (Dewey, Reference Dewey1896). By contrast, in living organisms, perception is highly selective and entails actively making sense of what is out in the world in a way that allows the organism to figure out what it should do about it (Mitchell, Reference Mitchell2023). Indeed, there is good evidence that conscious perception is necessarily structured through the activity of attention (e.g., Jennings, Reference Jennings2015).

These capacities are underpinned by neural and cognitive architectures that can also be defined in terms of internal functional relations. With respect to consciousness, these involve a functional control hierarchy, with higher levels operating over representations of (beliefs about) objects in the world, internal states, motivations, moods, goals, and other elements of cognition. These kinds of relations can be described (and formalized, e.g., in the tradition of cybernetics) in abstract terms, without relying on either specific computational implementations or bespoke biological substrates.

The question then is what kinds of relations will go beyond cybernetic functions and give rise to or entail subjective experience? The answer, of course, is not known, but we can still ask what form we expect it to take. Some candidate elements of subjectivity include the organism being able to distinguish itself from the world (Jennings, Reference Jennings2020), being able to track its own internal states (Nikolova et al., Reference Nikolova, Waade, Friston and Allen2022), having valenced internal signals that guide action (Godfrey-Smith, Reference Godfrey-Smith2020; Cleeremans & Tallon-Baudry, Reference Cleeremans and Tallon-Baudry2022), being able to track the consequences of its actions (Jékely, Godfrey-Smith, & Keijzer, Reference Jékely, Godfrey-Smith and Keijzer2021), making predictions about itself and the world (as Seth notes), and having a “phenomenal self-model” (Metzinger, Reference Metzinger2004). Conscious subjects should have internal representations that stand in exploitable relation to things in the world (Shea, Reference Shea2018), but also some sort of reflexivity that imbues these representations with “mineness.”

Often overlooked in these discussions, however, is that living organisms have temporal extension. We live in and through time and cannot be reduced, as selves, to a single temporal moment (Husserl Reference Husserl1964; Mitchell, Reference Mitchell2023). All living organisms are beings that (literally) incorporate traces of their past experience to better propagate themselves into the future (Potter & Mitchell, Reference Potter and Mitchell2025). In more complex organisms, selfhood is defined by continuity through time of both the physical and the psychological – of us, as persons. We bring our personal history to bear in our interactions with the world. We are actively seeking, filtering, predicting, and interpreting sensory information in inherently subjective, goal-oriented ways.

Thus, rather than drawing comparisons and contrasts between biological and artificial systems, we can see them both as particular instantiations of a more abstract set of relational principles. We can zoom out and view consciousness as a “mode of being” (Ginsburg & Jablonka, Reference Ginsburg and Jablonka2020); specifically, of being a subject and a self. There seems little reason in principle to think that the relations underpinning subjectivity and selfhood could not be operationalized and implemented in many different ways in many different substrates, including artificial entities. Given the right mode of being in the world, it might very well be like something to be such a system.