Anil K. Seth (2024), Behavioral and Brain Sciences 49, e315 (target article; online 2025-04-21, volume 2026). First posted as a PsyArXiv preprint. — DOI-10.1017/S0140525X25000032
Key takeaways
- Conscious AI needs both computational functionalism and enough substrate flexibility to run consciousness-relevant computations in silicon, and both are doubtful. Seth argues that positive, non-vitalist reasons from predictive processing, the free energy principle and autopoiesis tie consciousness to being alive, so real AI consciousness is unlikely on current trajectories, while conscious-seeming AI is nearly inevitable and ethically pressing.
Seeded from the abstract and a skim, before the reading in NOTE-169. What follows is what the work says about itself.
Seth asks whether advancing AI might become conscious as well as intelligent. He diagnoses the biases that make people expect it (anthropocentrism, human exceptionalism, anthropomorphism, and the resulting conflation of intelligence with consciousness). He questions whether computation suffices for consciousness and argues instead for biological naturalism, on which consciousness depends on our being living organisms. After surveying scenarios for artificial consciousness, he concludes that it is unlikely on current trajectories but grows more plausible as AI becomes more brain-like or life-like. He ends with the ethics of AI that is, or convincingly seems to be, conscious: “If we sell our minds too cheaply to our machine creations, we not only overestimate them – we underestimate ourselves.”
Standing in the record
Filed from the 2026-09-26 sweep of the papers-feed tracker for philosophy: 5 seconds of active reading over 1 session. It was filed Deferred, unread. NOTE-169 is the close reading of 2026-09-26, and it placed the work: Active — worth reading as the leading statement of biological naturalism about AI, with a useful five-scenario map (Table 1) of what conscious AI would require. Its positive case rests on the free energy principle, which the paper itself calls “much debated” (§4.2), and it claims plausibility rather than necessity. The commentaries and the Author’s Response (BBS 49, 2026) are the next thing to read, and the response was not reached here.
When seeded, before the reading in NOTE-169, the entry gave a priority for a deeper reading: high — t=5 s, but the paper is load-bearing for any AI-consciousness note, and the BBS package (commentaries and response) adds substantially beyond what the skim captures.
What the seeding asked a deeper reading to check (NOTE-169 is that reading):
- It is the leading statement of the biological-naturalist side that Birch’s LIT-111 sets against computational functionalism, and it connects to LIT-120 (autopoiesis), LIT-195 (life–mind continuity), LIT-103 (Frankish’s functionalism) and LIT-185 (IIT).
- A deeper reading should check whether the free-energy/autopoiesis argument gives necessity or only an explanatory bridge. Seth’s own epistemic/ontological split suggests the strong claim is not established.
- The BBS commentaries and Seth’s response (2026 volume) are the natural next reading. They were not reached here.
Access when seeded: The Cambridge Core page gave only the abstract (paywalled target article; “Published online by Cambridge University Press: 21 April 2025”). Read the accepted-manuscript PDF (42 pp., headed “Behavioral and Brain Sciences (forthcoming) … accepted for publication and has not yet been copyedited”) from the PsyArXiv/OSF download tz6an (v2, 22 Apr 2025). Read the abstract, keywords, all section heads, §1, the opening of §2, §3.9 Summary, §4.5 Summary, §5.8 Summary, §6.2 and §7 Conclusions. published: is PsyArXiv v1 (OSF API date_published 2024-06-30; Seth’s blog announced the preprint on 13 Jul 2024). OSF’s “original publication date” field says 2024-12-10, a discrepancy that is unexplained. The commentaries and author’s response in BBS were not read.