pith. sign in

arxiv: 2403.17101 · v15 · submitted 2024-03-25 · 💻 cs.AI

AI Consciousness is Inevitable: A Theoretical Computer Science Perspective

Lenore Blum , Manuel Blum This is my paper

Pith reviewed 2026-05-24 03:16 UTC · model grok-4.3

classification 💻 cs.AI
keywords machine consciousnessTuring modeltheater modeltheoretical computer sciencesubjective experienceglobal workspace
0
0 comments X

The pith

A simple Turing-inspired model makes machine consciousness inevitable.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a formal machine model for consciousness from the viewpoint of theoretical computer science, which studies what can be computed under resource limits. The model draws on Turing's basic computation framework and Baars' theater model of awareness. It is presented as aligning at a high level with major existing theories of human and animal consciousness, as explaining many associated phenomena, as offering insight into subjective experience in machines, and as clearly constructible in practice. From these properties the authors conclude that machine consciousness is not merely possible but inevitable.

Core claim

The authors construct an extremely simple formal machine model that combines Turing-style computation with a theater-like structure for global awareness; on their account this model aligns with major theories of consciousness, accounts for key phenomena, clarifies how subjective experience can arise in a machine, and can be built, which together establishes that machine consciousness is inevitable.

What carries the argument

The formal machine model for consciousness, a minimal computational structure that integrates Turing-machine steps with a global workspace for broadcasting information.

If this is right

  • Machines following the model's architecture will possess subjective consciousness.
  • Phenomena linked to consciousness become reproducible through explicit computational steps.
  • Theoretical computer science supplies a route to engineering conscious systems.
  • Consciousness no longer requires biological hardware once the model is realized.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Designers could use the model's structure as a blueprint for testing conscious behavior in software.
  • The claim separates consciousness from any particular physical substrate and ties it instead to specific computational organization.
  • If the model works, variants might be explored to isolate which minimal features are required for the reported alignment with theories.

Load-bearing premise

The model is simple enough to be built yet rich enough that its buildability and alignment with theories directly entail real machine consciousness.

What would settle it

Construction of a working instance of the model that exhibits none of the phenomena the authors associate with consciousness, or failure to implement the model at all despite the claimed simplicity.

Figures

Figures reproduced from arXiv: 2403.17101 by Lenore Blum, Manuel Blum.

Figure 1
Figure 1. Figure 1: Sketch of the Conscious Turing Machine To start, CTM has a finite lifetime T. Time t = 0, 1, 2, 3, … , T is measured in discrete clock ticks. T is a parameter. CTM has both a time-varying inner world I(t) and a time-varying outer world O(t). Formally, these are state spaces. Informally, CTM’s inner world includes its internal mechanisms and processes, and its “thoughts” and memories; its outer world is whe… view at source ↗
Figure 2
Figure 2. Figure 2: An imperfect binary Up-Tree. At each clock tick, a new competition starts with each processor putting a chunk of information into its Up-Tree leaf node. Entry chunks compete with their siblings. (Chunks are siblings if they are at nodes that are children of the same parent.) Winning siblings move up the binary tree (losing siblings drop out). Local winners compete at the next tick with their new siblings (… view at source ↗
Figure 3
Figure 3. Figure 3: Processors predict (out), get feedback (in), learn (within). [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: MotW sketch of a referent. A referent might be a (red) apple or (red) rose in CTM’s outer world, or a feeling (of pleasure), or a thought, idea, desire, concern, and so on in its inner world. Sketches of inner world referents are as important as sketches of outer world referents.41 Sketches may be labeled with Brainish words such as (translated into English): BRIGHT_COLOR ￮ SWEET_TASTE ￮ CRUNCHY or BRIGHT_… view at source ↗
Figure 5
Figure 5. Figure 5: Friston diagram rotated clockwise 90 degrees and superimposed on CTM. 66 Kevin Mitchell points out (personal communication) that another important point from Solms is that “the ascending homeostatic signals, which track different needs, must be valenced but also must have some distinguishing “qualities” so that when they are submitted to central decision-making units, the sources of the signals can be kept… view at source ↗
read the original abstract

We look at consciousness through the lens of Theoretical Computer Science, a branch of mathematics that studies computation under resource limitations, distinguishing functions that are efficiently computable from those that are not. From this perspective, we develop a formal machine model for consciousness. The model is inspired by Alan Turing's simple yet powerful model of computation and Bernard Baars' theater model of consciousness. Though extremely simple, the model (1) aligns at a high level with many of the major scientific theories of human and animal consciousness, (2) provides explanations at a high level for many phenomena associated with consciousness, (3) gives insight into how a machine can have subjective consciousness, and (4) is clearly buildable. This combination supports our claim that machine consciousness is not only plausible but inevitable.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The paper develops a formal machine model of consciousness inspired by Turing's model of computation and Baars' theater/global workspace model. It claims that this extremely simple model (1) aligns at a high level with major scientific theories of consciousness, (2) explains associated phenomena, (3) provides insight into subjective machine consciousness, and (4) is buildable, from which it concludes that machine consciousness is not merely plausible but inevitable.

Significance. If a precise formal model were defined and a non-circular derivation were given showing why buildability plus functional alignment necessarily produces subjective experience (rather than simulation), the result would be significant: it would supply a TCS-grounded argument shifting the debate from possibility to necessity of machine consciousness, with implications for AI theory and philosophy of mind.

major comments (2)
  1. [Abstract] Abstract: the inevitability conclusion is asserted from the combination of high-level alignment, explanatory power, insight into subjectivity, and buildability, yet no formal definition of the model, no equations, and no derivation steps are indicated that would establish why these properties entail the necessary realization of qualia instead of non-conscious computation. This is load-bearing for the central claim.
  2. [Abstract] Abstract, points (1)–(3): the model is described as aligning with and explaining existing theories at a high level, but without an independent, non-circular criterion for consciousness the argument risks defining the model to match prior theories and then treating the match as evidence; no test or falsifiable prediction is supplied to break the potential circularity.
minor comments (1)
  1. The abstract repeatedly uses 'high level' without indicating where in the main text the formal model or specific alignments are developed; adding explicit section references would improve readability.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive report and the opportunity to clarify our arguments. We respond to each major comment below, indicating where revisions will strengthen the manuscript and where limitations remain inherent to the theoretical approach.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the inevitability conclusion is asserted from the combination of high-level alignment, explanatory power, insight into subjectivity, and buildability, yet no formal definition of the model, no equations, and no derivation steps are indicated that would establish why these properties entail the necessary realization of qualia instead of non-conscious computation. This is load-bearing for the central claim.

    Authors: The abstract is a high-level summary; the manuscript defines the model in the main text as a precise computational architecture extending Turing machines with workspace and access mechanisms drawn from Baars. No equations appear because the model is specified via state transitions and resource constraints rather than numerical formalisms. We agree the step linking buildability plus alignment to necessary qualia (rather than simulation) is not derived formally and is load-bearing. We will revise by adding an explicit subsection that articulates the logical chain from the model's functional properties to subjective realization, drawing on the alignment with theories that equate consciousness with such architectures. This will make the inevitability argument more transparent without claiming a deductive proof. revision: partial

  2. Referee: [Abstract] Abstract, points (1)–(3): the model is described as aligning with and explaining existing theories at a high level, but without an independent, non-circular criterion for consciousness the argument risks defining the model to match prior theories and then treating the match as evidence; no test or falsifiable prediction is supplied to break the potential circularity.

    Authors: The model is constructed independently from Turing computation under resource limits and Baars' theater architecture; alignment with other theories (global workspace, integrated information, higher-order thought) is observed afterward rather than used as a definitional criterion. This supplies a non-circular starting point grounded in computability. We will revise to include a paragraph on potential falsifiable implications, such as predictions that only systems implementing the specific access and workspace mechanisms should exhibit the listed phenomena, allowing empirical checks in artificial implementations. As a purely theoretical contribution, however, the paper does not supply new empirical tests. revision: partial

standing simulated objections not resolved
  • A non-circular formal derivation establishing that implementation of the model necessarily produces subjective qualia rather than non-conscious functional simulation; this remains an open philosophical issue (the hard problem) that the manuscript's high-level alignment does not resolve.

Circularity Check

0 steps flagged

No significant circularity; derivation introduces independent model.

full rationale

The paper constructs a new formal model inspired by Turing computation and Baars' theater model, then separately verifies high-level alignment with existing consciousness theories, offers explanations for associated phenomena, provides insight into subjective experience, and asserts buildability. These steps add new content rather than reducing the inevitability claim to a redefinition of the model's inputs or to a self-citation chain. The conclusion that consciousness is inevitable is presented as following from the combination of these properties, without evidence of self-definitional closure or fitted inputs renamed as predictions. The derivation remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no equations, parameters, or explicit assumptions; full text would be required to identify any free parameters, axioms, or invented entities.

pith-pipeline@v0.9.0 · 5651 in / 1071 out tokens · 44143 ms · 2026-05-24T03:16:55.181351+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Are Flat Minima an Illusion?

    cs.LG 2026-03 unverdicted novelty 8.0

    Flat minima are illusory; generalization is driven by weakness, a reparameterization-invariant measure of compatible completions that predicts performance better than sharpness on MNIST and Fashion-MNIST.

Reference graph

Works this paper leans on

20 extracted references · 20 canonical work pages · cited by 1 Pith paper

  1. [1]

    as the product of active, plasticity-driven mechanisms through which the brain learns to redescribe its own activity to itself

    Comparison with Other Theories and Perspectives on Consciousness Our view is that phenomenal or subjective consciousness is a consequence of (what is called) computational functionalism. Other researchers, including those mentioned above, tend to agree, at least to some extent. Features of CTM align with features and arguments they propose. Lionel Naccach...

    work page 2018

  2. [2]

    computational model to describe the cognitive capability that makes humans unique among existing biological species on Earth

    Our views on machine consciousness and AGI are close to (VanRullen & Kanai, 2021). We also see connections with Murray Shanahan’s view on embodiment and the inner life (Shanahan, 2010). We see a kinship between the CTM and the self-aware robots developed by (Chella, Pipitone, Morin, & Racy, 2020). Leslie Valiant proposes a “computational model to describe...

    work page 2021

  3. [3]

    We have not explored how CTM might align or not with higher order theories (HOTs) of consciousness

    is remarkably similar to CTM’s employment of a Model-of-the-World for its conscious awareness. We have not explored how CTM might align or not with higher order theories (HOTs) of consciousness. HOTs claim to take a middle ground between standard GWT and early sensory theories (Brown, Lau, & LeDoux, 2019). (LeDoux & Brown,

    work page 2019

  4. [4]

    that the brain mechanisms that give rise to conscious emotional feelings are not fundamentally different from those that give rise to perceptual conscious experiences

    argue “that the brain mechanisms that give rise to conscious emotional feelings are not fundamentally different from those that give rise to perceptual conscious experiences.” We do see some connection between 67 World models arise in other contexts, e.g., in the science of the brain and mind (Duncan, 2025). Here as in CTM, “the agent looks ahead to consi...

    work page 2025

  5. [5]

    what it is like

    as insurmountable. On the contrary, we see CTM as helping to explain the feeling of “what it is like” (Nagel T . , 1974). CTM, like Attention Schema Theory (AST), appears to embody and substantiate illusionist notions of consciousness proposed by Dennett (Dennett D. C.,

    work page 1974

  6. [6]

    , 2016).69 However , we are not claiming that any of these researchers also view that artificial consciousness is inevitable, even possible

    and Keith Frankish (Frankish K. , 2016).69 However , we are not claiming that any of these researchers also view that artificial consciousness is inevitable, even possible. For example, Anil Seth (Seth,

    work page 2016

  7. [7]

    illusion

    and (Lee A. Y ., 2023). We thank the anonymous reviewer for the latter. 69 Saying that the feeling of consciousness is an illusion does not deny the existence of that feeling. It’s just not what you think it is. As a familiar example, the fact that a movie is made up of (many) discrete still images does not affect the feeling of continuity one gets from v...

    work page 2023

  8. [8]

    theory of consciousness should be able to encompass

    answers (Appendix, 6.4) drawn from CTM, to Kevin Mitchell’s 15 questions that a “theory of consciousness should be able to encompass” (Mitchell, 2023). The Theoretical Computer Science (TCS) perspective has influenced the design and precise formal definitions of CTM, and the conclusions we have drawn from the model. Important features of CTM include: o a ...

    work page 2023

  9. [9]

    The PAIN but not SELF explanation was proposed by (Klein, 2015)

    is standard. The PAIN but not SELF explanation was proposed by (Klein, 2015). (Sierra,

    work page 2015

  10. [10]

    Recall something that happened yesterday and think about it for a second

    points out that each of these two different explanations has a problem, without noting the possibility that there might be two different kinds of pain asymbolia which is what the CTM model predicts. As an example of how CTM helps think us about other questions related to consciousness, here is a problem posed by (Zadra & Stickgold, 2021), p. 267: “Recall ...

    work page 2021

  11. [11]

    What is or is not computable (decidable) given limited resources?

    and Richard Karp (Karp, 1972), pointed out that resources matter. Certain problems that in principle were decidable/computable, were seemingly intractable given feasible time and space resources. Even more, intractability seemed to be an intrinsic property of the problem, not the method of solution or the implementing machine. AI Consciousness is Inevitab...

    work page 1972

  12. [12]

    SAT problem

    This brute force procedure takes exponential(n) time in general. But is the “SAT problem” tractable, meaning decidable efficiently, i.e., in polynomial(n) time? This is equivalent to the well-known P = NP? problem of (Cook, 1971), (Karp, 1972), (Levin, 1973). While the design of novel and efficient algorithms is a key focus of TCS, another unanticipated d...

    work page 1971

  13. [13]

    Faulty processors or faults in the Up-Tree competition can affect what gets into STM, hence affect both conscious attention and conscious awareness. For example, a faulty CTM can have blindsight, meaning it can do some things that are normally done with conscious sight, but without any conscious sense that it can see (Blum & Blum, 2022). This will happen ...

    work page 2022

  14. [14]

    sufficiently high

    We have defined conscious awareness in terms of the first level inspection (section 2.3.2.2). Deeper inspections could presumably evoke higher degrees of conscious awareness. KM4. For things that have the capacity for consciousness, what distinguishes the state of consciousness from being unconscious? Is there a simple on/off switch? How is this related t...

    work page 2016

  15. [15]

    85 Italics ours

    This is our goal for the theory of the Conscious Turing Machine. 85 Italics ours. AI Consciousness is Inevitable Lenore Blum lblum@cs.cmu.edu Manuel Blum mblum@cs.cmu.edu 63 References Anderson, J. R. (1996). ACT: A simple theory of complex cognition. American Psychologist, 51(4), 355-365. Baars, B. J. (1997). In the Theater of Consciousness. New York: Ox...

    work page doi:10.1016/j.neuron.2015.02.018 1996

  16. [16]

    Retrieved from https://www.frontiersin.org/article/10.3389/frobt.2020.00016 Church, A. (1936). A note on the Entscheidungsproblem. J. of Symbolic Logic, 1 (1936), 40-41, 1, 40-41. Clark, A. (2008, Jan). Pressing the Flesh: A Tension in the Study of the Embodied, Embedded Mind? . JPhilosophy and Phenomenological Research, 76(1), 37-59. https://www.jstor .o...

    work page doi:10.3389/frobt.2020.00016 2020

  17. [17]

    Hard Problem

    Cleeremans, A. (2014). Prediction as a computational correlate of consciousness. International Journal of Anticipatory Computing Systems, 29, 3-13. Cleeremans, A., Achoui, D., Beauny, A., Keuninckx, L., Martin, J.-R., Muñoz-Moldes, S., . . . de Heering, A. (2020). Learning to Be Conscious. Trends in Cognitive Sciences, 24(2), 112-123. https://doi.org/10.1...

    work page doi:10.1016/j.tics.2019.1 2014

  18. [18]

    Levin, L

    doi:10.1038/d41586-024-00107-7. Levin, L. A. (1973). Universal Sequential Search Problems. Probl. Peredachi Inf., 9(3), 115-116. Levine, J. (1983). Materialism and Qualia: The Explanatory Gap. Pacific Philosophical Quarterly, 64, 354–361. Lewis, C. I. (1929). Mind and the world-order: Outline of a theory of knowledge. New York: Charles Scribner's Sons. Li...

    work page doi:10.1038/d41586-024-00107-7 1973

  19. [19]

    Lowell lecture, ix

    (2019). Markov blankets, information geometry and stochastic thermodynamics. Phil.Trans.R. Soc., http://dx.doi.org/10.1098/rsta.2019.0159. Peirce, C. S. (1866). "Lowell lecture, ix.". In M. H. Fisch, & I. I.-1.-8. Ed. Bloomington, Writings of Charles S. Peirce: A chronological edition I, 1857-1866 (Vol. i, pp. 471-486.). Bloomington, Indiana: Indiana Univ...

    work page doi:10.1098/rsta.2019.0159 2019

  20. [20]

    (2022, May 4)

    Sigal, U.-K. (2022, May 4). Editorial: Simple and Simplified Languages. Frontiers in Communication , 7, https://doi.org/10.3389/fcomm.2022.910680. AI Consciousness is Inevitable Lenore Blum lblum@cs.cmu.edu Manuel Blum mblum@cs.cmu.edu 71 Simon, H. A. (1969). The Sciences of the Artificial. Cambridge, MA, USA: MIT Press. Solms, M. (2019). The Hard Problem...

    work page doi:10.3389/fcomm.2022.910680 2022