A critical context is the emerging field of Diverse Intelligence (Lyon, Reference Lyon2006; Reber & Baluska, Reference Reber and Baluska2021), seeking to understand what all agents, regardless of their origin story or composition, have in common. The neural replacement argument is only the tip of the iceberg, given the interoperability of life: cyborgs, hybrots, biobots, AI-mediated interfaces to collective intelligence of somatic cells, and all manner of products of synthetic bioengineering are already sharing our world. Should a being who is a coherent, functional integration of both natural, engineered, and software components be subjected to some sort of purity test seeking to neatly classify them (with all its legal and ethical implications) as “real human” vs. “machine”? When there is no one to speak out for beings embodied in unconventional materials, there will eventually be no one left to speak out for all of us, who are embodied in “proteins that are just doing what proteins do.”

Advances in bioengineering mean that colloquial linguistic categories of “life” vs. “machine” are doomed (Clawson & Levin, Reference Clawson and Levin2023; Levin, Reference Levin2022), and “life” as unique substrate for consciousness is questionable given that we do not fundamentally understand what the secret sauce is that is supposed to separate the living from the non-living (Bender et al., Reference Bender, Kofman, Agüera y Arcas and Levin2025). It’s not just about evolutionary transitions: each of us has made the gradual, continuous journey from a quiescent bag of chemicals (the unfertilized oocyte) to a conscious adult who will make statements about not being a machine. Developmental biology offers no bright line between being fully describable by chemistry and being fodder for behavioral science, psychoanalysis, and love. Synthetic, cell-free systems, and active matter (Cejkova et al., Reference Cejkova, Banno, Hanczyc and Stepanek2017; Kaygisiz & Ulijn, Reference Kaygisiz and Ulijn2025) systems will surely merge with efforts to make AI. The gradualist lessons of modern developmental biology support not magical binary categories, but the need to understand metamorphosis, scaling, and self-assembly of systems with significant inner perspective from minimal components.

We are unprepared to recognize intelligence, never mind consciousness, in systems of unfamiliar spatial and temporal scales or in unconventional substrates (Fields & Levin, Reference Fields and Levin2022). We can begin to address our native mind-blindness with data on basal cognition in cells, tissues, plants, etc. (Baluška & Levin, Reference Baluška and Levin2016; Reber & Baluska, Reference Reber and Baluska2021), as the use of cognitive science tools to understand and manipulate their function is already leading to empirical advances in regenerative medicine and engineering. Groups of cells implement a collective intelligence that forms memories, pursues goals, makes decisions, and solves novel problems in creative and ingenious ways (Levin, Reference Levin2025b; McMillen & Levin, Reference McMillen and Levin2024) - a stepping-stone to more alien cases. The electrophysiological mechanisms underlying cell intelligence are the same (indeed, evolutionarily homologous) to those that seem to underpin consciousness in human bodies) (Fields, Bischof, & Levin, Reference Fields, Bischof and Levin2020). But cells perform their sensing-deciding-acting loop in the high-dimensional spaces of the following: gene expression, physiological states, and anatomical morphospace. They pursue goals, avoid obstacles, and deal with unexpected perturbations with various degrees of ingenuity and success (Levin, Reference Levin2023). They sense, strive, and perhaps suffer in these other spaces largely invisible to our obsession with 3D space (though it is as constructed by our cognitive autopoiesis as the others).

We are just beginning to adjust our intuitions (and vocabulary) to accommodate the discovery that intelligence, in its essence, is basal and far older than neurons. “Embodiment” does not mean a motile robotic body: organoids and software AI’s, looking to us as un-embodied (“not engaging with the real world, and thus not binding symbols” etc.) may well be navigating spaces we don’t recognize. Embodiment is critical for minds, but that does not mean only in 3D space. Finally, all of the criteria commonly used to assume consciousness with each other (problem of other minds) – similar composition, common evolutionary origin, behavioral competencies that suggest valence, and causal architectures – now reveal that we absolutely must take seriously the likelihood that our body structures have their own (non-verbal) consciousness (far beyond the right hemisphere in split brain patients). Their consciousness is inaccessible to the main linguistic Self as that of other humans. If we cannot expand our detection of consciousness to this close test case, any discussion of consciousness in AI is hopeless.

I agree with Seth that computation does not suffice for consciousness. While many find it plausible that biochemistry does not tell the full story of the mind, it is universally believed that algorithms and materials tell the full story of “machines.” But, in synthetic beings, as well as in evolved ones, our formal models never tell the full story of what something is or what it can and cannot do. Even humble Bubble Sort has competencies (not just complexity or unpredictability, but properties recognizable by behavioral scientists) and pursues unexpected side-quests not mentioned in its short, transparent, deterministic algorithm (Zhang, Goldstein, & Levin, Reference Zhang, Goldstein and Levin2025). Simple sets of coupled differential equations can do several different kinds of learning (Biswas et al., Reference Biswas, Manicka, Hoel and Levin2021). All of these were not recognized despite decades of study, until they were looked at through the diverse intelligence lens (Manicka & Levin, Reference Manicka and Levin2019). If even these truly minimal systems can do that, how many competencies, in what spaces, perhaps having absolutely nothing to do with language, could be present in current and future AIs? These intrinsic motivations (things they do that are specifically not prescribed by their algorithm) are likely to have much to do with being substrates for consciousness, free will, etc. Perhaps there is no dead matter, no “machines” that are not somewhere on the spectrum of life and mind. Perhaps what we make as offspring, whether biological or engineered, are interfaces to patterns of form and behavior from a non-physical latent space that contains the familiar low-agency truths of mathematics, but also a very wide range of higher-agency forms we recognize as kinds of minds.

In summary, I think Seth raises key questions that are not about AI at all (Levin, Reference Levin2025a). How can conscious, morally important minds exist in a world in which their embodiments are made of parts following rules (whether of biochemistry, electronics, algorithms, etc.), and whether anything in our world is only what our computational or other models say it is? I would complement Seth’s powerful conclusion as follows. If we seek to buy our exceptionalism too cheaply, we underestimate not only “machines” but also ourselves.

Figure 1.

The space of possible embodied minds is truly vast, breaking ancient linguistic categories of “living being” vs. “machine.” Because of the interoperability of life, almost any combination of evolved, engineered, and hybrid materials can support forms of embodiment and behavior that must be taken seriously as possible conscious kin. Our pre-scientific heuristics for recognizing other minds must be significantly expanded to enable ethical synthbiosis with beings who are not like us, as humanity matures. Image by Jeremy Guay of Peregrine Creative.