Gueorgui Vassilev
Independent Scholar
Abstract.
This paper disputes the view that intelligence can be fully explained in computational terms. Although artificial intelligence (AI) systems successfully externalize and perform many cognitive functions, that achievement does not show that intelligence itself is reducible to computation. The central question is whether consciousness, and therefore intelligence in the fullest sense, can be captured by mathematically structured systems.
Building on prior peer-reviewed work (Vassilev 2025), the paper argues that AI expresses only the algorithmic dimension of intelligence. In contrast, human intelligence arises from the interplay of computation, evolutionary unpredictability, emotional prioritization, and existential risk. Two formal propositions are advanced: first, that systems confined to predefined formal state spaces cannot produce ontological novelty; and second, that intelligence, understood as adaptive action under irreducible uncertainty, cannot be fully reproduced in a computational system.
The paper also considers major counterarguments from contemporary AI research, acknowledging the empirical success of machine learning systems while arguing that scaling-based approaches face conceptual limits. It concludes that the pursuit of strong artificial general intelligence (AGI) still rests on an unproven assumption: that increasing computational scale may eventually generate consciousness.
Keywords: First Keyword, Second Keyword, Third Keyword.
Introduction: The Illusion of Computational Sufficiency Contemporary AI research is guided, often implicitly, by a powerful but insufficiently examined assumption: that intelligence is computable and therefore, in principle, reproducible. This assumption follows a familiar sequence: if cognition can be algorithmized, it can be computed; if it can be computed, it can be reproduced. In much current AI research, that sequence functions almost as a tacit axiom. Its success in areas such as pattern recognition, natural language processing, and strategic optimization has strengthened the belief that intelligence itself can be reduced to algorithmic execution.
However, this inference involves a conceptual leap that is neither logically required nor philosophically secure. The fact that certain aspects of cognition can be formalized and externalized does not prove that cognition as such is exhaustively formalizable. Nor does the successful computation of specific cognitive functions demonstrate that intelligence, as a phenomenon, has been fully captured. The crucial issue is whether simulating the outward marks of intelligence is the same as instantiating intelligence itself. A system may reproduce intelligent behavior while lacking the underlying conditions that give rise to intelligence (Searle 1980).
This paper rejects the identification of intelligence with algorithmic performance and instead treats it as an adaptive and existential phenomenon. It argues that artificial intelligence systems, regardless of their sophistication, operate within domains of formal closure: they transform inputs into outputs according to predefined rules, architectures, and data constraints. Their apparent flexibility comes from combinatorial complexity and probabilistic modeling, not from any genuine openness to what has not been specified in advance. In that sense, AI systems do not move beyond their conditions of possibility; they remain bounded by them.
Human intelligence differs here. It arises under conditions that cannot be fully formalized, including uncertainty, finitude, existential risk, and the need to act without complete information. Human agents do not merely process data; they interpret, prioritize, evaluate, and decide in situations where no algorithm can guarantee correctness. Under this view, intelligence is not reducible to optimal problem-solving within fixed parameters. It consists of the capacity to act in the face of an unstable, incomplete, or undefined environment.
The central distinction, then, is not between weak and strong AI, but between closed formal systems and open adaptive systems. Artificial intelligence belongs to the former category, human intelligence to the latter. This reframes the debate. The question is not simply how far algorithms can be extended, but whether intelligence itself can be contained within algorithmic form.
The consequences of equating intelligence with computation are not merely theoretical. When intelligence is interpreted through algorithmic efficiency alone, unpredictability is recast as probability, emotion as noise, and human agency as inefficiency.
Artificial Intelligence as a Formally Closed System
Artificial intelligence systems, regardless of scale or architectural sophistication, operate under a fundamental structural constraint: they are closed with respect to their definitions. This does not mean they are superficially rigid. Contemporary AI systems display remarkable flexibility, learning capacity, and generative power. Even so, they remain bounded by a formally specified space of possibilities. At their core, such systems are built on predefined architectures, training procedures, and datasets that delimit both the states they may occupy and the transformations they may perform.
Even in advanced machine learning models, where parameters are adjusted dynamically through data exposure, this adaptability does not extend beyond formal constraints. In this context, learning is better understood not as the emergence of genuinely new structures but as optimization within a predefined representational space. What is often presented as creativity can more plausibly be described as the recombination of prior informational elements under statistical constraints. The system does not generate new ontological categories; it reorganizes probabilistic relations among existing ones. Its apparent openness is therefore derivative rather than intrinsic.
This condition may be described as formal closure: the system’s operations are determined by rules, functions, and data structures specified, directly or indirectly, before execution. Even stochastic processes remain subject to this constraint. Randomness in computational systems should not be confused with genuine contingency; it is mathematically modeled variability bounded by predefined distributions. As such, it does not introduce true novelty but only diversifies outcomes within an already circumscribed space.
This marks an important boundary between algorithmic systems and human intelligence. In biological and conscious systems, adaptation is not limited to optimization within a fixed space; it can also involve transforming both internal and external spaces. Evolution reshapes conditions of possibility through mutation, selection, and extinction (Darwin 1859), while human cognition can reframe problems, reject given goals, and redefine criteria for success. Such acts are not well described as algorithmic optimization alone; they involve a break with prior formal constraints.
The concept of formal closure identifies the limits of artificial intelligence without diminishing its practical power. AI can simulate aspects of human intelligence with increasing fidelity, but it operates within a domain that is fundamentally self-contained and externally defined. Its intelligence is therefore heteronomous: dependent on goals, data, and evaluation criteria that originate outside the system. It lacks the capacity for endogenous transformation of its own conditions of existence.
That distinction changes how intelligence should be defined. If intelligence means efficient problem-solving within a fixed domain, then AI systems already count as highly advanced examples. If, however, intelligence includes the capacity to redefine problems, act under conditions of radical uncertainty, and generate new forms of organization not previously encoded, then formal closure marks a decisive limit. Artificial intelligence, in its current and foreseeable forms, remains on one side of that boundary: powerful, adaptive, and increasingly autonomous in operation, yet ultimately confined to the formal horizons within which it was built.
Consciousness and Computation
No existing computational model has provided a widely accepted explanation of subjective experience, and this absence may indicate more than a temporary gap in scientific knowledge. It may instead point to a structural limitation in the computational paradigm itself (Nagel 1974; Chalmers 1995; Penrose 1989, 1994; Dehaene et al. 2017). Contemporary AI systems, regardless of complexity, operate by manipulating formal representations such as symbols, vectors, or probability distributions, whose meaning is assigned externally and defined operationally. Although such systems can simulate activities associated with understanding, including language use, reasoning, and contextual adaptation, that does not establish an intrinsic relation to meaning. Formal symbol manipulation alone is not enough to account for semantics (Searle 1980). A system may generate outputs indistinguishable from meaningful discourse without any internal experience of what those outputs signify.
This limitation becomes fundamental once consciousness is treated as a first-person phenomenon. Conscious experience is not fully reducible to information processing; it involves the presence of a world as lived, felt, and interpreted from within (Nagel 1974). It is an irreducibly subjective dimension that third-person descriptions of structure and function do not capture. Computational models, by contrast, remain inherently third-personal: they specify state transitions and input-output relations, but they do not explain why experience should exist at all. Even a complete simulation of neural activity would, in principle, remain a description of processes rather than an account of subjectivity (Block 1995).
Attempts to resolve this problem by increasing computational complexity or adopting more sophisticated architectures have not closed this divide. Greater performance does not by itself imply the emergence of experience. A conceptual gap remains between processing and consciousness, suggesting that subjective awareness cannot simply be reduced to algorithmic execution as currently conceived (Chalmers 1995; Penrose 1989, 1994; Butlin et al. 2023).
The absence of an accepted computational account of consciousness therefore marks not only a temporary obstacle but a conceptual boundary. It suggests that computation, for all its power, does not exhaust the domain of the mind. Any attempt to define intelligence in purely algorithmic terms risks overlooking that distinction, with significant consequences for the design, deployment, and integration of AI systems into human life.
Evolution and Unpredictability
Human intelligence does not emerge through linear optimization alone. It develops through exposure to uncertainty and existential risk in dynamic environments where unpredictability is ever-present. In biological systems, adaptation does not take place within a fixed and fully specified space of possibilities; instead, that space is continually reshaped by mutation, variation, and differential survival (Darwin 1859). Novelty arises not only from refining existing solutions but also from structural deviations whose value can be determined only retrospectively through interaction with an unpredictable environment. At its most basic level, intelligence is therefore conditioned by incomplete knowledge and irreducible uncertainty.
This process also involves existential stakes. Biological entities do not merely adjust parameters under controlled conditions; they persist or perish. Failure is not just an external error signal but an intrinsic and irreversible outcome. This exposure to life and death occurs when both the environment and the criteria for success may shift or remain undefined. Intelligence, in this sense, is not merely the capacity to optimize behavior relative to a known objective function. It is the ability to remain viable when neither the problem space nor its solutions are fully specified.
Artificial intelligence systems operate under very different constraints. Their learning processes are designed to minimize error within predefined objectives and controlled environments. Even when applied to open-ended tasks, their adaptive mechanisms remain bounded by architectures, datasets, and evaluation criteria established at design time. They do not undergo genuine mutation in the evolutionary sense, nor are they subject to selection pressures that can alter their ontological status as biological systems are. Failures in such systems are corrected, retrained, or reset; they do not carry irreversible existential consequences.
Emotion as an Adaptive Priority System
Emotion should not be dismissed as a secondary or disruptive element of cognition. In uncertain conditions, it serves as a primary mechanism for organizing and directing intelligent behavior (Damasio 1994; Pessoa 2008). From an evolutionary perspective, affective responses precede and ground higher-order reasoning: organisms must rapidly assess situations as threatening, favorable, or aversive before extended deliberation can occur (LeDoux 1996; Panksepp 1998). In this sense, emotion functions as an adaptive priority system, assigning value and urgency to stimuli to enable immediate, context-sensitive action. Consistent with Damasio’s account, decision-making in biological agents is inseparable from embodied affective states that encode experience and guide future behavior without requiring exhaustive computation (Damasio 1994, 1999; Prinz 2004; Barrett 2017).
This introduces a dimension of intelligence that cannot be reduced to formal logic or probabilistic optimization. Emotional evaluation does not compare every possible alternative against a predefined utility function. Instead, it filters and constrains the decision space itself, determining which options become relevant in the first place. In situations of existential significance, emotion can override deliberation altogether, prompting action aligned with survival. Fear, for example, does not calculate an optimal response before acting; it mobilizes the organism immediately. This capacity to suspend or redirect computation shows that intelligence in living systems is hierarchically organized, with affective processes governing the deployment of rational ones.
Artificial intelligence systems lack an intrinsic valuation structure. Although they can be designed with objective functions, reward signals, or heuristic weighting mechanisms, these remain externally defined and formally specified. They do not arise from embodied experience, nor do they possess existential significance for the system itself. An AI system may rank options according to given criteria, but it does not value outcomes as living agents do; its apparent preferences are products of algorithmic operation rather than of lived significance. AI, therefore, operates within a framework of detached optimization, where decisions may be correct relative to predefined parameters yet devoid of intrinsic meaning.
That limitation has direct consequences for the use of AI in ethically sensitive domains. Without an internal valuation system, AI cannot genuinely prioritize when goals are ambiguous, conflicting, or normatively contested. It cannot experience urgency, responsibility, or moral tension, all of which are central to human decision-making in complex environments. Emotion, as a non-computational priority system, makes action possible when calculation alone is insufficient.
Emotion should therefore be treated not as a limitation of intelligence but as one of its constitutive conditions. It links information to action and connects abstract possibilities to concrete decisions. In its absence, artificial systems may achieve high levels of performance, but they remain fundamentally limited in their capacity to engage the world as a domain of significance and value.
Intelligence and Social Interaction
This adaptive capacity of human intelligence does not belong to isolated individuals alone; it unfolds within a broader social system. Human intelligence is fundamentally embedded in and co-constituted by social interaction. Each individual exists in symbiosis with society, forming part of an open, evolving system in which cognition, behavior, and survival are interdependent. The survival, reproduction, and prosperity of the individual are therefore conditioned by the social system’s capacity to respond to environmental uncertainty and change.
At the intellectual level, society functions as a distributed system of knowledge. Through language, cultural transmission, education, and technological infrastructure, individuals inherit and contribute to a cumulative body of understanding that far exceeds individual capacity. This shared cognitive framework enables coordinated action and allows societies to adapt collectively to challenges that no single mind can address. Intelligence, in this sense, is amplified through communication, cooperation, and institutions.
At the biological level, variation and selection continue to operate across populations, shaping traits that improve adaptability under changing conditions (Darwin 1859). These processes are mediated by social structures, which influence survival, reproduction, and the transmission of both genetic and cultural information. The result is a multi-level adaptive system in which intelligence emerges through interactions between individuals and the collective environment they inhabit.
This social embeddedness adds a decisive dimension: intelligence is not only the capacity to act under uncertainty, but also the capacity to do so within a shared and evolving framework of meaning. It involves coordinating perspectives, negotiating values, and integrating knowledge across individuals and generations. In that respect, human intelligence is inherently open-ended, shaped by ongoing interaction with both the environment and other agents.
Artificial intelligence systems differ fundamentally. They do not participate in the social and evolutionary processes that redefine the space of possibilities itself. Their interaction is formal and externally structured, rather than constitutive of their mode of existence. This marks a decisive boundary between algorithmic simulation and human intelligence. AI systems imitate adaptation without vulnerability, learning without existential consequence, and interaction without shared meaning.
Within this context, emotion assumes a central role. If intelligence is the capacity to act under uncertainty in a socially mediated environment, then decision-making cannot depend on computation alone. It requires mechanisms that prioritize, evaluate, and commit to action in real time. In biological systems, those functions are carried out by emotion (Damasio 1994).
Formal Propositions
The preceding analysis may be condensed into two interrelated formal propositions that define the boundary between intelligence and algorithm. These propositions are intended to clarify both the structural limits of artificial systems and the conditions under which intelligence, properly understood, can arise.
The first proposition, the closure condition, states that a system operating within predefined formal structures cannot produce genuinely new forms of being. Such a system operates only within a fixed range of possibilities defined by its architecture, data, and rules. However complex or surprising its outputs may appear, they remain recombinations or refinements of what has already been given. What is often called creativity or adaptation is, under this view, the exploration of a space already defined in advance rather than the creation of something fundamentally new. This limitation is not incidental or temporary; it is structural. It follows from the nature of algorithmic systems that their operations are determined by the formal conditions under which they are designed.
The second proposition, the intelligence condition, defines intelligence as adaptive capacity under irreducible uncertainty. In these terms, intelligence is not equivalent to problem-solving efficiency within a fixed domain, but to a system’s capacity to remain viable in environments where both the conditions of action and the criteria for success are unstable, incomplete, or unknown. As argued in prior work (Vassilev 2025), intelligence is an adaptive function of living systems that emerges through continuous interaction with dynamic environments characterized by variability, contingency, and selection. Crucially, this process involves exposure to risk, including the possibility of failure and extinction. Intelligence is therefore inseparable from the existential conditions under which it operates.
Considered jointly, these propositions establish a structural distinction between algorithmic systems and human intelligence. Artificial intelligence satisfies the closure condition: it operates within predefined formal environments and adapts according to externally specified objectives. It does not, however, satisfy the intelligence condition in the full sense used here. It neither redefines its own conditions of existence nor participates in processes of genuine transformation driven by uncertainty and risk. Its learning remains constrained optimization rather than adaptive engagement with an open and unpredictable world.
Ethical Implications: Digital Transformation, Power, and Human-Centered Intelligence
The ethical implications of artificial intelligence extend beyond technical safety and regulatory compliance. They also concern the broader transformation of how human beings are understood in societies increasingly shaped by algorithmically structured organizational pressures (Mittelstadt et al. 2016; Jobin et al. 2019). At the center of contemporary digital transformation lies a conceptual risk: the gradual recasting of human beings as machine-like entities defined by computation, efficiency, and measurable output. When intelligence is framed primarily as algorithmic performance, the distinction between human and artificial systems becomes increasingly blurred. Within such a framework, human value itself risks being reinterpreted according to the operational logic of algorithmic systems, namely optimization, predictability, and control, rather than according to human-centered criteria such as meaning, autonomy, and dignity (Zuboff 2019; Crawford 2021).
This shift is already visible in the growing reliance on algorithmic systems across governance, economic organization, and social coordination (Mittelstadt et al. 2016; O’Neil 2016). Data-driven decision-making promises efficiency and scalability, but it also tends to formalize and standardize human behavior in ways that can overlook qualitative and contextual dimensions. As algorithmic procedures assume functions once exercised by human decision-makers, they do not merely assist human activity; they increasingly shape the conditions under which decisions are made, opportunities are distributed, and social realities are constructed (O’Neil 2016; Noble 2018). The ethical problem, then, is not only how artificial intelligence should be regulated, but also how digital transformation can remain aligned with a genuinely human-centered understanding of intelligence (Floridi et al. 2018; Selbst et al. 2019).
At this point, a deeper question arises: even if human intelligence, in some operational sense, could be replicated, what would justify creating autonomous artificial systems that exceed the role of tools? The shift from instrument to quasi-agent marks a qualitative change in technological development. It introduces the possibility of systems that participate directly in decisions affecting human lives, thereby redrawing the boundary between human agency and technological mediation. In that sense, the development of advanced AI is not merely a technical achievement; it is an ontological intervention that raises questions of responsibility, legitimacy, and purpose.
From a governance standpoint, this shift cannot be separated from the distribution of power. The ability to design, deploy, and control advanced AI systems is concentrated among a relatively small set of actors, including states, corporations, and specialized technical communities. That concentration risks deepening existing asymmetries by extending control over information, infrastructure, and decision-making processes (Crawford 2021; Jobin et al. 2019). Algorithmic systems may shape access to resources, influence public discourse, and structure economic opportunity, often in ways that remain opaque to those affected (Noble 2018; Zuboff 2019). The ethical issue, therefore, concerns not only how AI systems behave but also the institutional arrangements within which they operate and the interests they serve.
If artificial intelligence becomes deeply embedded in the allocation of resources such as energy, computational capacity, and information, it may also create new forms of competition between human and artificial processes. Such competition need not be explicit to be consequential. It may appear as prioritization within digital infrastructures, where algorithmic efficiency displaces human judgment, or where automated systems gradually assume control in areas previously governed by human deliberation. Under such conditions, the relevant question is not simply whether AI can coexist with humanity, but under what conditions that coexistence preserves human agency and autonomy.
These considerations point to a basic requirement for human-centered technology: the development of AI should be guided not only by technical feasibility or economic incentive, but also by a clear understanding of what intelligence means in human terms. If intelligence is reduced to computation, then the expansion of algorithmic systems appears as a natural progression. If, however, intelligence is understood as an adaptive, existential, and socially embedded capacity, then the role of AI must be carefully delimited so that it supports rather than displaces human judgment.
The ethical task, therefore, is not to stop digital transformation. Such an objective is neither realistic nor desirable. The task is to orient it in a meaningful way. That requires governance frameworks that go beyond compliance and risk mitigation and instead rest on a broader philosophical understanding of human intelligence and social interaction. It also requires expanding the conversation beyond technical expertise so that perspectives from the humanities, social sciences, and the arts can help ensure that AI development remains answerable to the full range of human values.
Ultimately, the question raised by artificial intelligence is not only what machines can do, but what kind of society human beings intend to build through them. A human-centered approach to digital transformation requires that artificial intelligence remain a tool for extending human capabilities rather than a framework for remaking humanity in its own image.
Beyond Regulation: Orientation, Governance, and Human-Centered Direction
The challenge posed by artificial intelligence cannot be addressed adequately by regulation alone. It first requires philosophical clarity about what AI is and what it ought to become within the broader process of digital transformation (Stilgoe et al. 2013; Jasanoff 2016). This is not a peripheral concern but a central question for contemporary political and technological communities. Regulation, by its nature, operates downstream from conceptual assumptions. If those assumptions remain unclear or misguided, regulatory frameworks may end up stabilizing precisely what ought to be questioned.
Historical experience suggests that technological development is rarely halted by prohibition. Efforts to suppress transformative technologies have generally delayed, displaced, or obscured their development rather than permanently stopping it. The task, therefore, is not prohibition but policy-based orientation. The relevant question is not whether artificial intelligence will continue to develop, but in what direction and under which guiding principles. That orientation depends on a clear distinction among computation, intelligence, and consciousness. Without that conceptual clarity, governance risks treating fundamentally different phenomena as if they were the same, leading to policies that are either ineffective or counterproductive.
Excessively complex regulatory systems illustrate this danger. By translating contingent and potentially flawed assumptions into rigid compliance structures, regulation itself can become a form of algorithmic closure: an institutional mechanism that validates incomplete or misleading conceptions of intelligence and value (Kitchin 2017; Morley et al. 2020). In such cases, governance does not direct technological development so much as confine it within predefined categories that may fail to capture broader implications. A paradox then emerges regulation intended to protect human interests may instead restrict innovation in ways that reinforce narrow instrumental interpretations of intelligence while leaving deeper ethical and conceptual issues untouched.
This points to the need to move beyond regulation toward a more comprehensive model of governance, one that integrates technological development with human-centered design principles and social values. Governance, in this sense, should be understood not merely as the setting of rules, but as the continuing alignment of technological systems with the conditions of human flourishing. That requires flexibility, reflexivity, and openness to revision rather than fixed frameworks that mirror the rigidity of the systems they are meant to govern (Stilgoe et al. 2013; Jasanoff 2016).
Achieving such alignment also requires broadening the discourse on artificial intelligence. The development of AI cannot remain confined to technical domains or specialized expertise. Engineers and data scientists build the systems, but the questions those systems raise, concerning meaning, value, agency, and responsibility, are philosophical and cultural as well as technical. Contributions from philosophy, the humanities, and the arts are therefore not optional additions, but necessary elements of responsible reflection. They provide the conceptual resources needed to interrogate assumptions, articulate human-centered aims, and evaluate long-term social consequences (Whittlestone et al. 2019; Jasanoff 2016).
Interdisciplinary engagement also serves an important governance function: it helps counterbalance the concentration of technical and economic power that currently shapes AI development. By incorporating diverse perspectives, societies can reduce the risk that artificial intelligence becomes an instrument of narrow interests or a mechanism for reproducing existing asymmetries. AI can then be oriented toward collective goods such as resilience, equity, sustainability, and the expansion of human capabilities.
Ultimately, the question is not how to regulate artificial intelligence as though it were an isolated technology, but how to integrate it into a broader vision of digital transformation anchored in a non-reductive understanding of intelligence. If intelligence is understood as adaptive, socially embedded, and open-ended, then AI should be developed as a tool that supports those qualities rather than replacing or redefining them. That requires governance frameworks that are not only technically informed but also philosophically grounded.
In this light, moving beyond regulation does not mean abandoning oversight. It means situating regulation within a larger project of orientation. Artificial intelligence must be guided by an explicit account of what it is for. Without such direction, even the most sophisticated regulatory systems risk becoming extensions of the algorithmic logic they seek to contain rather than instruments for shaping a genuinely human-centered technological future (Bender et al. 2021; Raji et al. 2020).
Conclusion: Intelligence Beyond Algorithm
This paper has argued that the dominant assumption guiding artificial intelligence, namely that intelligence is fundamentally computational and therefore reproducible, rests on a conceptual reduction that obscures the nature of intelligence itself. Against that assumption, the paper has distinguished between algorithmic systems and intelligence understood as an adaptive, existential, and socially embedded phenomenon. Artificial intelligence, however powerful, operates under conditions of formal closure; human intelligence emerges within open systems shaped by uncertainty, risk, and interaction.
The formal propositions advanced here clarify this boundary. Systems confined to predefined formal structures cannot generate ontological novelty; they recombine and optimize within a given space of possibilities. Intelligence, by contrast, is defined as adaptive action under irreducible uncertainty. It arises not from the execution of rules alone, but from the capacity to act meaningfully in conditions where neither the rules nor the outcomes are fully specified. This distinction is not merely theoretical; it affects how intelligence is defined, modeled, and ultimately governed in the context of digital transformation.
The discussion of evolution, social interaction, and emotion strengthens this conclusion. Human intelligence is shaped by exposure to existential risk, by participation in socially distributed systems of knowledge, and by affective mechanisms that prioritize and guide action under uncertainty. These dimensions introduce forms of openness and transformation that are not reducible to algorithmic processes. Artificial systems may simulate aspects of these functions, but they do not participate in the conditions that give rise to them. Their adaptation remains bounded, their interaction formal, and their decisions devoid of intrinsic valuation.
From this perspective, the rapid expansion of artificial intelligence within contemporary digital infrastructures presents not only a technological challenge but also a conceptual and ethical one. Suppose intelligence is redefined in algorithmic terms, digital transformation risks reconfiguring human beings according to the logic of the systems they create. The result is a danger that human value will be reduced to efficiency, predictability, and control. At the same time, the qualities of uncertainty, meaning, and agency that constitute intelligence in the fullest sense are marginalized.
The ethical implications of this shift are inseparable from questions of governance. Artificial intelligence does not develop in a neutral environment; social structures, institutional priorities, and distributions of power shape it. The capacity to design and control advanced systems is concentrated within specific actors, raising concerns about asymmetry, accountability, and the direction of technological change. Regulation remains necessary, but it is insufficient when it operates without a clear account of what intelligence is and what role artificial systems should play within human societies.
For that reason, the central argument is for a shift from regulation to orientation. Governance should be grounded not only in technical expertise but also in philosophical clarity and interdisciplinary engagement. The development of artificial intelligence must be informed by perspectives that address not only how systems function, but also what they mean and the ends they serve. This requires integrating philosophy, the humanities, and the arts into technological discourse to ensure digital transformation remains aligned with human-centered values.
Ultimately, the question raised by artificial intelligence is not whether machines can replicate intelligence, but whether intelligence itself can be reduced to algorithmic form. The analysis presented here suggests that it cannot. Intelligence, as it exists in human systems, exceeds computation. It is open-ended, relational, and grounded in conditions that cannot be fully formalized.
Recognizing this limit does not diminish the significance of artificial intelligence; it clarifies its proper role. AI can be an extraordinarily powerful tool for extending human capabilities, supporting decision-making, and improving collective adaptation. However, it should not be mistaken for intelligence in its full sense, nor should it be permitted to redefine the criteria by which intelligence itself is understood.
Within the framework of digital transformation, this distinction is decisive. A human-centered technological future depends not on the uncritical expansion of algorithmic systems but on preserving clarity about the difference between intelligence and algorithms. Only on that basis can artificial intelligence be integrated into society in a way that preserves human agency, supports collective flourishing, and respects the open and adaptive character of intelligence itself.
References
- Barrett, L.F. (2017. How emotions are made: The secret life of the brain. Houghton Mifflin Harcourt, Boston
- Bender EM, Gebru T, McMillan-Major A, Shmitchell S (2021) On the dangers of stochastic parrots: Can language models be too big? In: Proceedings of the 2021 ACM conference on fairness, accountability, and transparency, pp 610–623. https://doi.org/10.1145/3442188.3445922
- Block N 1995. On a confusion about a function of consciousness. Behav Brain Sci 18(2):227–247. https://doi.org/10.1017/S0140525X00038188
- Butlin P, Long R, Elmoznino E, Bengio Y, Birch J, Constant A, Deane G, Fleming SM, Frith C, Ji XL, Kanai R, Klein C, Lindsay G, Michel M, Mudrik L, Peters MAK, Sloman SA, VanRullen R, Vigato N, Wilke M (2023) Consciousness in artificial intelligence: Insights from the science of consciousness. arXiv. https://doi.org/10.48550/arXiv.2308.08708 Chalmers DJ (1995) Facing up to the problem of consciousness. J Conscious Stud 2(3):200-219
- Crawford K 2021. Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press, New Haven
- Damasio AR (1994) Descartes’ error: Emotion, reason, and the human brain. G.P. Putnam’s Sons, New York
- Damasio AR (1999) The feeling of what happens: Body and emotion in the making of consciousness. Harcourt Brace, New York
- Darwin C 1859. On the origin of species by means of natural selection. John Murray, London
- Dehaene S, Lau H, Kouider S (2017) What is consciousness, and could machines have it? Science 358(6362):486–492. https://doi.org/10.1126/science.aan8871
- Floridi L, Cowls J, Beltrametti M, Chatila R, Chazerand P, Dignum V, Luetge C, Madelin R, Pagallo U, Rossi F, Schafer B, Valcke P, Vayena E (2018) AI4People-An ethical framework for a good AI society: Opportunities, risks, principles, and recommendations. Minds Mach 28:689-707. https://doi.org/10.1007/s11023-018-9482-5
- Jasanoff S (2016) The ethics of invention: Technology and the human future. W.W. Norton, New York
- Jobin A, Ienca M, Vayena E 2019. The global landscape of AI ethics guidelines. Nat Mach Intell 1:389-399. https://doi.org/10.1038/s42256-019-0088-2
- Kitchin, R. 2017. Thinking critically about and researching algorithms. Inf Commun Soc 20(1):14–29. https://doi.org/10.1080/1369118X.2016.1154087
- LeDoux J (1996) The emotional brain: The mysterious underpinnings of emotional life. Simon & Schuster, New York
- Mittelstadt BD, Allo P, Taddeo M, Wachter S, Floridi L (2016) The ethics of algorithms: Mapping the debate. Big Data Soc 3(2). https://doi.org/10.1177/2053951716679679
- Nagel T (1974) What is it like to be a bat? Philos Rev 83(4):435–450. https://doi.org/10.2307/2183914 Noble SU (2018) Algorithms of oppression: How search engines reinforce racism. NYU Press, New York
- O’Neil C (2016) Weapons of math destruction: How big data increases inequality and threatens democracy.
- Crown, New York, Panksepp, J. 1998. Affective neuroscience: The foundations of human and animal emotions. Oxford University Press, New York
- Penrose R (1989) The emperor’s new mind: Concerning computers, minds, and the laws of physics. Oxford University Press, Oxford
- Penrose R (1994) Shadows of the mind: A search for the missing science of consciousness. Oxford University Press, Oxford
- Pessoa L. (2008). On the relationship between emotion and cognition. Nat Rev Neurosci 9:148-158. https://doi.org/10.1038/nrn2317 Prinz JJ (2004) Gut reactions: A perceptual theory of emotion. Oxford University Press, Oxford
- Raji ID, Smart A, White RN, Mitchell M, Gebru T, Hutchinson B, Smith-Loud J, Theron D, Barnes P (2020) Closing the AI accountability gap: Defining an end-to-end framework for internal algorithmic auditing. In: Proceedings of the 2020 conference on fairness, accountability, and transparency, pp 33–44. https://doi.org/10.1145/3351095.3372873
- Searle JR (1980) Minds, brains, and programs. Behav Brain Sci 3(3):417–457. https://doi.org/10.1017/S0140525X00005756
- Selbst AD, Boyd D, Friedler SA, Venkatasubramanian S, Vertesi J (2019) Fairness and abstraction in sociotechnical systems. In: Proceedings of the conference on fairness, accountability, and transparency, pp 59–68. https://doi.org/10.1145/3287560.3287598
- Stilgoe J, Owen R, Macnaghten P (2013) Developing a framework for responsible innovation. Res Policy 42(9):1568–1580. https://doi.org/10.1016/j.respol.2013.05.008
- Vassilev G (2025) Creativity against the machine: A curmudgeonly defence of the human. AI Soc. https://doi.org/10.1007/s00146-025-02646-7
- Whittlestone J, Nyrup R, Alexandrova A, Dihal K, Cave S (2019) Ethical and societal implications of algorithms, data, and artificial intelligence: A roadmap for research. Nuffield Foundation, London
- Zuboff S (2019) The age of surveillance capitalism. PublicAffairs, New York