Navigation

From BCM to Federated Neuro-Symbolic Hybrid HCAI

What is AGI? — and under what conditions can AI generate growth and prosperity for all?  Author: Friedrich Reinhard Schieck, Published in April 2026

ABSTRACT

The article “From BCM to Federated Neuro-Symbolic Hybrid HCAI – What Is AGI – and Under What Architectural, Institutional, and Economic Conditions Does Artificial Intelligence Create Growth and Prosperity for All?” proposes a fundamental redefinition of the AGI debate by shifting the widespread focus on purely technical performance enhancement in favor of an architectural, institutional, and political-economic perspective. The starting point is the observation that the concept of Artificial General Intelligence (AGI) in current debates is characterized by both enormous significance and considerable conceptual ambiguity. While AGI is typically described as a hypothetical AI that matches or surpasses human cognitive abilities across many or nearly all domains, this paper demonstrates that this conventional use of the term falls short analytically. It overlooks the social, infrastructural, property-rights, and governance-related conditions under which digital intelligence is created, trained, utilized, and economically exploited in the first place.

Against this backdrop, the paper argues that the crucial question is not merely whether AGI is technically possible, but rather what social form of intelligence emerges through the interplay of humans, machines, rules, institutions, and infrastructures—and whom this intelligence serves. The concept of AGI is thus shifted from an individualistic machine ontology to an institutional and socio-technical ontology. Intelligence no longer appears primarily as a property of an isolated system, but as the result of an organized interconnection of human judgment, symbolic rule-binding, machine scalability, distributed responsibilities, and institutional accountability. From this perspective, AGI is reconceived not as the domination of a centralized superintelligence, but as a potentially democratically organized symbiosis of human and machine capabilities.

The text makes a key contribution by tracing the evolutionary path from the BCM model through hybrid HCAI to federated neuro-symbolic hybrid collective intelligence. The BCM model is understood here as an early organizational precursor to an architecture in which complexity is not suppressed through centralization, but rather made productive through distributed responsibility, transparency, feedback, and adaptive self-regulation. Subsequently, Hybrid-HCAI is described as a socio-technical operating system that systematically couples three forms of intelligence: the human level as a normative and responsible authority, the symbolic level as the bearer of explicit rules, roles, and decision-making logics, and the subsymbolic level as an operational field of scalable pattern recognition, forecasting, and information processing. The paper emphasizes that socially sustainable AI should not be conceived as a substitute architecture for human judgment, but rather as a cooperative architecture. This architecture is expanded in the paper by two additional, constitutive functional logics: federation and governance.

This yields the complete architectural formula: Sub-symbolism scales. Symbolism regulates. People decide. Federation distributes. Governance is accountable.

This formula encapsulates the essence of the overall argument. Sub-symbolic methods enable the coverage, condensation, and adaptive processing of large datasets, but remain normatively underdetermined. Symbolic structures translate statistical results into explicit, verifiable, and institutionally compatible concepts, roles, and rules. The human level retains ultimate responsibility for values, goals, and exceptions. Federation protects against epistemic monopolization and organizational concentration of power through the distributed anchoring of knowledge, data, and responsibilities. Finally, governance forms the meta-order of the overall system by defining rules for audit, revision, intervention, liability, and institutional correctability. According to the article’s argument, it is only through the recursive interplay of these five functional logics that a complete architecture of intelligence and responsibility emerges—one that is not only efficient but also transparent, legitimate, pluralistic, and capable of correction.

This paper combines this architectural-theoretical reconstruction with a pointed critique of the current political-economic form of the AI boom. It demonstrates that today’s AI systems cannot simply be understood as autonomous “thinking machines,” but rather rely to a significant extent on the collection, consolidation, and monetization of human knowledge and interactive contributions. In this sense, many current AI platforms appear as infrastructures for the appropriation of collective intelligence. In this reading, the usual AGI rhetoric can function as a veiling concept that renders invisible the social preconditions, the property-based extraction of collective contributions, and the role of proprietary platform infrastructures. In this respect, the text also interprets AGI as an ideological nexus where technical promises of the future, investment dynamics, and concentrations of power converge. This analysis is deepened through an examination of three prominent voices in the current debate. Gary Marcus serves as an important reference point for the critique of the AGI hype.

His skepticism toward large language models as the supposed silver bullet for AGI is used to highlight the difference between impressive language capabilities and robust general problem-solving ability. The article combines this technical-cognitive critique with the argument that AGI rhetoric is not politically neutral, but can legitimize extraordinary investments, security narratives, and special regulatory treatment. Ricardo Hausmann and Andrés Velasco provide the macroeconomic perspective: Their analysis suggests that the valuations of major AI companies only appear plausible if these companies realize large-scale global income claims in the future. This paints a picture of a possible global economy in which access to indispensable AI infrastructure is priced as a rent. Finally, Paul Krugman is invoked to highlight the openness of technological development: AI is neither automatically an equalizing nor an automatically concentrating force; its distributional effects depend on institutions, market structures, and political regulation.

Against this backdrop, the paper develops the core thesis that artificial intelligence can only become a source of growth and prosperity for all if certain architectural and institutional conditions are met. First, broad, fair, and non-monopolistic access to AI infrastructure is required so that productivity gains do not get stuck at proprietary bottlenecks. Second, AI must primarily complement human labor—that is, enhance human capabilities rather than indiscriminately substitute expertise and erode accountability. Third, mechanisms for fair feedback on societal knowledge production are necessary so that users whose contributions go into improving intelligent systems are not structurally excluded from the resulting value chains. Fourth, democratic governance of information, power, and quality is needed to ensure that technical systems remain participatory, verifiable, capable of learning, and accountable.

Overall, this paper argues for a fundamental reframing of the AGI debate. The central question is no longer primarily when machines will achieve general intelligence, but rather what architecture of collective intelligence will be created and whether this architecture will strengthen freedom, dignity, judgment, democracy, and widespread prosperity. The proposed federated neuro-symbolic hybrid HCAI emerges as a superior model because it binds technical power to clarity of rules, human judgment, distributed sovereignty, and institutional accountability. AGI is thus understood not as a race toward artificial omnipotence, but as a search for a more just, accountable, and productivity-enhancing architecture of social intelligence.

Keywords: Artificial General Intelligence (AGI); Hybrid HCAI; BCM model; collective intelligence; human-centered AI; AI governance; federation; neuro-symbolic AI; political economy of AI; productivity growth; digital transformation; accountability; platform economy; democratic AI architecture; productive value creation

TABLE OF CONTENTS

  1. Introduction: From the Question of Computing Power to the Question of Architecture, Order, and Social Purpose
  2. The Concept of AGI: Between Cognitive Universality and Conceptual Ambiguity
  3. AGI as a Concealing Concept: A Critique of the Conventional Definition and a Shift Toward Institutional Ontology
  4. From the BCM Model to Hybrid-HCAI: The Emergence of an Architectural Conceptual Framework
  5. The Adaptation Gap: Why Digitalization Fails Without Architectural and Institutional Renewal
  6. Hybrid-HCAI as a Cooperative Architecture: The Integration of Human, Symbolic, and Sub-Symbolic Intelligence
  7. Architectural Expansion: Federation and Governance as Constitutive Dimensions
  8. The Five Functional Logics of the Complete Architecture
  9. Recursive Coupling, Application Contexts, and the Shift in the AGI Debate
  10. Gary Marcus: Why the AGI Hype Is Analytically and Politically Problematic
  11. Hausmann and Velasco: AGI, the Global Economy, and the Logic of Global Profit Extraction
  12. Paul Krugman: The Open Question of Distribution, Democracy, and Equalization
  13. Under What Conditions Does AI Actually Create Growth and Prosperity for All?
  14. Why the AGI Debate Needs to Be Reframed
  15. Conclusion: AGI as a Question of the Just Architecture of Collective Intelligence

 

1. INTRODUCTION: FROM THE QUESTION OF COMPUTATIONAL POWER TO THE QUESTION OF ARCHITECTURE, ORDER, AND SOCIAL PURPOSE

The current debate on Artificial General Intelligence, or AGI for short, is marked by a peculiar fundamental contradiction. On the one hand, AGI—more than almost any other term in the AI discussion—is imbued with enormous intellectual, technological, economic, and geopolitical implications. On the other hand, it often remains surprisingly unclear exactly what AGI is meant to signify. In popular, encyclopedic, and much of the public discourse, AGI generally refers to a hypothetical AI system that matches or even surpasses human cognitive abilities across many or nearly all domains. [1] This use of the term undoubtedly has an intuitive plausibility because it aligns with the image of a “generally intelligent machine.” At the same time, it is analytically insufficient. For it abstracts away from precisely the conditions under which intelligence in digital systems arises, is trained, institutionally embedded, controlled, and economically exploited in the first place.

Precisely for this reason, recent governance and regulatory literature on the concept of AI employs more cautious, function-based definitions of AI systems. The focus there is not on consciousness, human likeness, or philosophically robust forms of “thinking,” but rather on machine inference, output generation, degrees of autonomy and adaptivity, as well as the question of how technical systems intervene in physical and virtual environments. [2] This shift is of considerable significance. For in the case of AGI in particular, conceptual ambiguity is not merely a theoretical problem, but a practical issue of power. The question “What is AGI?” is never merely a technical question. It is always also a question of ownership, infrastructure, distribution, political control, institutional responsibility, and societal purpose.

This simultaneously shifts the perspective on the actual problem. As long as artificial intelligence is treated merely as a question of performance, it seems natural to focus on the question of when a system will achieve cross-domain cognitive capabilities that appear comparable to human general intelligence. This line of inquiry is technically legitimate and epistemologically unavoidable. Socially, however, it remains incomplete. For as the reach, generality, and power of intervention of intelligent systems grow, it becomes increasingly clear that what matters is not merely what a system can do, but whom it serves, who controls it, and according to which values, rules, and institutional procedures it operates. [3]

I consider this shift to be fundamental. In my view, the true purpose of artificial intelligence, extending all the way to AGI, does not lie in the mere increase of performance, efficiency, or computing power. What is decisive, rather, is whether such systems contribute to enabling people to live better lives, make wiser decisions, and solve major shared problems without undermining freedom, dignity, and responsibility.[4] It is therefore not technical power as such that is normatively decisive, but rather the question of whether the architecture of intelligent systems serves human and societal well-being. Thus, a paradigm of architecture replaces a pure performance paradigm. What matters is not only a system’s ability to recognize patterns, make predictions, or generate outputs, but the form of its embedding within responsibility, communication, organization, distribution, and governance.[5]

This insight is not only normative but also central to political economy. In their article “The Real Question About the AI Future” dated April 8, 2026, Ricardo Hausmann and Andrés Velasco put it bluntly: one should not merely ask whether the AI boom is a bubble, but rather what kind of global economy would actually have to emerge for the current market valuations of key AI companies to appear plausible. [6] This formulation shifts the focus from the performance of individual models to the political economy of AI as a whole. This is precisely where the real controversy lies. For even if AGI were one day realized in a robust, human-like sense, it would remain unclear who benefits from this technology, who controls its infrastructure, and how its productivity gains are distributed. Conversely, the following is also true: even without “true” AGI, today’s AI systems can already have profound effects on labor markets, knowledge structures, democracy, property structures, and geopolitical dependencies.

For this reason, in this article I also build upon my earlier work on the BCM model and hybrid HCAI. The central thesis is that a socially responsible form of advanced AI must be conceived as federated neuro-symbolic hybrid collective intelligence. Its completeness rests on five recursively coupled functional logics: subsymbolic scaling, symbolic rule-binding, human judgmental sovereignty, federated distribution, and institutionally anchored governance. [7] I understand this architecture not merely as a response to technical complexity, but as an answer to the normative and political question of how AI can serve humanity without devolving into technocratic disenfranchisement, diffusion of responsibility, or concentration of power.

The central question of this paper is therefore not merely whether AGI is possible. Rather, it is under what architectural, institutional, and economic conditions artificial intelligence can at all become a source of growth and prosperity for broad segments of the population—and when it instead tends to concentrate wealth, knowledge, and power in the hands of a few. To answer this question, it is necessary to clarify the concept of AGI itself, to reveal its ideological and political-economic functions, to identify its architectural alternatives, and finally to formulate the conditions under which AI is not only effective but also socially responsible, democratically embedded, and conducive to increased well-being.

2. THE CONCEPT OF AGI: BETWEEN COGNITIVE UNIVERSALITY AND CONCEPTUAL AMBIGUITY

Historically, the term AGI emerged as a counterpoint to narrow, specialized, or “narrow” AI. In this sense, AGI does not refer to just any powerful form of artificial intelligence, but rather to a system with cross-domain flexibility, learning ability, and the capacity for generalization. In current popular and encyclopedic accounts, AGI is usually described as a hypothetical AI that matches or surpasses human capabilities “across virtually all cognitive tasks.” [8] Gary Marcus also refers to a similar working definition in his article “Is AGI the right goal for AI?” when he describes AGI as an AI that could “match or exceed the cognitive versatility and competence of a well-educated adult.” [9]

At first glance, this definition appears more precise than many marketing-driven uses of the term. Nevertheless, it remains open in key respects. Which cognitive tasks are specifically meant? What role do embodiment, situational world knowledge, causal understanding, long-term planning, fault tolerance, reliability, or social judgment play? How is “generality” even to be measured? These questions alone demonstrate that AGI is not an established, operationally and unambiguously measurable technical term like “supervised learning,” “reinforcement learning,” or “transformer architecture.” Rather, it is a contested key concept whose criteria vary considerably depending on theoretical, economic, psychological, or philosophical backgrounds. It is precisely this heterogeneity that is revealing. Some definitions of AGI are oriented toward economically exploitable task domains, others toward general problem-solving abilities, and still others toward psychometric models, models of reason, or philosophical concepts of intelligence. For this reason, AGI can only with difficulty be described as a clearly defined technical goal. Rather, the term often serves a descriptive, rhetorical, and strategic function simultaneously. Those who portray AGI as imminent can mobilize investments, attract political attention, justify special regulatory treatment, and lend plausibility to the extraordinary valuation of their own infrastructures or business models.

In contrast, it is striking that institutional standard-setting bodies such as the OECD deliberately opt for a narrower, governance-friendly definition of AI. The OECD defines an AI system not in terms of “general intelligence,” but as a machine-based system that, for explicit or implicit goals, derives from inputs how outputs such as predictions, content, recommendations, or decisions can be generated that influence physical or virtual environments; furthermore, such systems vary in terms of their autonomy and adaptability. [10] This definition is particularly important for two reasons. First, it makes clear that AI is already relevant institutionally and regulatorily today, without the need to achieve AGI. Second, it shifts the focus from mystified machine reason to the concrete question of how systems infer, act, and are embedded in human contexts. This already reveals a key finding. The classic definition of AGI is intellectually stimulating, but scientifically, institutionally, and in terms of political economy, it is inadequate. It abstracts away from the social conditions under which digital intelligence is generated, trained, improved, controlled, and exploited. This is precisely where my critique begins.

3. AGI AS A CONCEALING TERM: A CRITIQUE OF THE CURRENT DEFINITION AND A TRANSITION TO INSTITUTIONAL ONTOLOGY

I take issue with the standard definition of AGI, arguing that it obscures the underlying business model of modern AI platforms to a significant degree. The core of my critique is that today’s AI models appear to the outside world as autonomous “thinking machines,” but in reality rely heavily on the collection, consolidation, and monetization of human knowledge, communication, and interaction. The more users participate in communication, information, and interaction processes, the greater the system’s knowledge input; the platform can then resell this collectively co-produced knowledge in the form of AI output, productivity services, or decision-making aids.

This critique is deliberately exaggerated, but it points to a real structural issue in digital platform economies. The economic value of many digital systems arises not solely from internal technical excellence, but from network effects, usage data, behavioral traces, feedback loops, human corrections, fine-tuning through feedback, and the ability to incorporate heterogeneous contributions into proprietary infrastructures. In this sense, modern generative AI is not purely machine intelligence, but a technologically condensed form of social language, attention, experience, and problem-solving. The OECD itself emphasizes that the development and deployment phases of AI can overlap and that fine-tuning or continuous training in downstream use can significantly alter a system’s behavior and performance. [11]

My alternative definition of AGI as the “ultimate symbiosis” of human cognitive computing power with artificial computing power is theoretically productive because it liberates the concept from the narrow machine perspective and transfers it into the framework of socio-technical systems. AGI would thus not primarily be a property of an isolated model, but rather the emergent result of an organized interconnection of human judgment, semantic competence, and machine formalization capabilities. The crucial addition to this position is normative: the rules of this symbiosis must not be centrally set and altered by platform operators, but must be determined in a decentralized manner by the users themselves.

From a scientific perspective, this thesis shifts the concept of AGI from an individualistic to an institutional ontology. It is not a machine as such that embodies general intelligence, but rather a distributed system comprising people, rules, interfaces, data flows, data centers, storage formats, and decision-making architectures. This perspective offers a significant advantage. It reveals that the social significance of AI does not depend solely on benchmarks, but on ownership structures, control rights, participation structures, and governance models. At the same time, it demands conceptual precision. For not every form of collective networking is already AGI. This conception gains precision when read not as a description of today’s realities, but as a normative counter-conception: AGI in an emancipatory sense would be a democratically organized, qualitatively adaptive hybrid form of human-computer interaction and collective intelligence.

This shift is not merely a conceptual variation. It alters the entire horizon of the debate. As soon as AGI is no longer understood as an isolated property of a system but as a form of institutionally organized intelligence, the question of architecture comes to the fore. It is precisely at this point that the path leads to my own line of development, from the BCM model through hybrid HCAI to federated neuro-symbolic hybrid collective intelligence.

4. FROM THE BCM MODEL TO HYBRID-HCAI: THE EMERGENCE OF AN ARCHITECTURAL CONCEPT

When I trace the path from the BCM model to the Hybrid HCAI, I do not view it as a linear history of innovation. Rather, I see it as the development of an architectural conceptual framework that emerged from practical experiences of organizational transformation. The starting point was the profound upheavals following the collapse of the GDR and in the context of German reunification. Under conditions of rapid structural change, it became apparent that organizations cannot be stabilized solely through hierarchy, standardization, and linear control. Rather, a form of self-organization was required in which roles, temporal logics, and information flows are interrelated in such a way that responsibility is distributed, communication is fed back, and value creation is made transparent. [12]

For me, therefore, the BCM model was far more than an organizational tool from the very beginning. I understood it as a precursor to an architecture in which complexity is not suppressed through centralization, but rather made productive through distributed responsibility and transparent coordination. Even then, transparency, feedback, decentralized responsibility, role orientation, and adaptive self-governance were formulated as structural principles.[13] For me, the crucial core principle has always been that self-organization must not be confused with lawlessness. It is only viable if it is structured, rule-bound, and supported both technically and socially. [14]

With the transition to Hybrid-HCAI, these principles were not abandoned but elevated to a new technological level. I understand Hybrid-HCAI not primarily as a single technology, but as a socio-technical operating system that integrates human judgment, symbolic AI, and subsymbolic AI into a transparent, federated, and responsibility-oriented architecture. [15] From this perspective, hybrid HCAI is not the negation of BCM, but rather its further development under the conditions of AI-driven digital modernity.[16]

To me, the connection seems even clearer today than it did in its early days. Even in BCM, the focus was on how responsibility, communication, and self-management can be architecturally organized in complex environments so that systems remain capable of acting. With the proliferation of AI-powered systems, the same question arises again at a higher level. Now it is no longer just about roles, processes, and information flows within organizations, but about the form in which human, symbolic, and machine intelligence are coupled with one another.

In this respect, there is no break between BCM and Hybrid-HCAI, but rather a transformation of the fundamental architectural problem. What once appeared to be an organizational-methodological question of self-governance now presents itself as a more comprehensive question regarding the constitution of social intelligence. The actual task consists of combining technical power with human judgment, semantic explicability, distributed responsibility, and institutional accountability. This simultaneously makes clear why the debate on AGI cannot meaningfully be limited to a hypothetical general machine intelligence. The decisive challenge lies in designing an architecture of intelligence and responsibility.

5. THE ADAPTATION GAP: WHY DIGITALIZATION FAILS WITHOUT ARCHITECTURAL AND INSTITUTIONAL REFORM

Central to this argument is the identification of an “adaptation gap”—that is, a structural disconnect between technological dynamics and organizational reality. In other words: The greater the pace of change in the economy, politics, and society, the less time is available and the longer it takes to adapt the organizational, informational, and process structures of every social system (family, organization, company, society). This problem touches on a phenomenon that has also been described in research on digital transformation: Despite substantial investments in information and communication technologies, platforms, automation, and data-driven systems, the expected productivity boost has often failed to materialize. Brynjolfsson, Rock, and Syverson have attributed this discrepancy to implementation delays, measurement issues, distributional effects, and a lack of complementary organizational changes. [17]

I, however, take this a step further. To me, the deficit appears not merely as a delay in adoption, but as an expression of a deeper systemic asynchrony between technology, organization, and governance. In my view, the adaptation gap has structural, cultural, and normative dimensions. Structurally, bureaucratic hierarchies persist that are tailored to stability and control in relatively static environments. Culturally, routines, implicit norms, and power patterns persist that are not automatically transformed by technological innovations. Normatively and in terms of governance, there is often a lack of a coherent framework that systematically organizes decision-making processes, accountability, auditability, and feedback. [18]

This is precisely where the deeper significance of the problem lies. The weakness of many digitalization processes lies not only in the fact that they were technically incomplete or implemented too slowly. It lies in the fact that technological innovation encounters institutional arrangements that contradict its own structural principles. Highly dynamic, data-driven, and adaptive systems are embedded in structures designed for linear planning, vertical control, silo logic, and minimal feedback. The result is often a paradoxical situation: more technology is introduced without making the organization any smarter, more adaptive, or more productive.

This is where my analysis aligns with the arguments of Acemoglu and Johnson. Their central insight is that technological progress by no means automatically leads to widespread prosperity, but rather depends crucially on institutional arrangements and the distribution of power. Technological trajectories are not neutral; they can even reinforce existing structures of control and power if they are not integrated into new forms of collective governance. [19] From a normative perspective, this is crucial for the AI debate. If the purpose of AI is not “ever-increasing performance” but rather a contribution to the good life, to wiser decision-making, and to the resolution of shared problems, then the adaptation gap demonstrates that technical innovation, without architectural and institutional renewal, fails to fulfill its own promise. [20]

It is therefore not enough to develop more powerful systems if their embedding in organizations and societies is based on patterns that dilute responsibility, weaken participation, and stabilize opaque centers of power. The adaptation gap thus marks a turning point in the debate. It reveals that the future of intelligent systems does not depend solely on the improvement of individual models, but on the ability to embed these systems in transparent role architectures, resilient feedback loops, and legitimate accountability frameworks. The real choice is not between analog and digital or human and machine, but between technocratic and accountable architecture.

6. HYBRID-HCAI AS A COOPERATIVE ARCHITECTURE: THE INTEGRATION OF HUMAN, SYMBOLIC, AND SUBSYMBOLIC INTELLIGENCE

In response to this diagnosis, I developed the Hybrid-HCAI model. Its basic premise is that socially acceptable AI must be designed not as a substitute for human judgment, but as a cooperative framework. Hybrid-HCAI systematically integrates three forms of intelligence. [21] The human level forms the normative and contextual foundation. It is here that objectives are formulated, value judgments are made, priorities are set, and responsibility is assumed.[22] The symbolic level translates rules, role models, processes, ontologies, and decision-making logics into explicit, machine-readable, and verifiable forms. It ensures explainability, consistency, and normative compatibility. [23] The subsymbolic level encompasses neural networks, statistical methods, and generative models. Its strength lies in pattern recognition, forecasting, and the scaling of complex information processing, but not in autonomous normative final decision-making.[24]

With this model, I explicitly oppose a one-sided delegation to opaque systems. Hybrid HCAI relies on a division of labor in which machine scaling, rule-bound symbolism, and human responsibility complement one another. This view is closely aligned with Ben Shneiderman’s understanding of Human-Centered AI, which conceives of technical systems as instruments of human empowerment rather than human disempowerment.[25] It is precisely at this point that the concept of purpose comes into full focus.

If AI is to help humanity live better and make wiser decisions, then it must not be understood as a mere tool for automation or as an autonomous super-entity. Rather, its appropriate form is a cooperative architecture in which technical capability remains tied to human goals, symbolic rule-binding, and institutional responsibility.[26]

In this sense, Hybrid-HCAI is not merely a technical model, but a normative model of social order. It describes a form of organized intelligence in which complementarity replaces displacement. Machines are not intended to relieve humans of responsibility, but rather to expand human and collective capacity for action. At the same time, the symbolic layer ensures that this expansion does not remain in the realm of mere statistical plausibility, but can be translated into explicit concepts, roles, rules, and verifiable decision-making pathways.

In this way, Hybrid-HCAI opens a path beyond two dead ends that characterize the current AI debate. The first dead end consists of purely technological euphoria, which already sees increasing computing power and model size as the silver bullet to general intelligence. The second dead end consists of a blanket skepticism toward technology that views complex digital systems solely as a threat to human autonomy. In contrast, the cooperative approach of Hybrid-HCAI points out that the relevant distinction lies not between human and machine, but between displacing and complementary, opaque and explicit, centralizing and accountable architecture.

7. THE ARCHITECTURAL EXPANSION: FEDERATION AND GOVERNANCE AS CONSTITUTIVE DIMENSIONS

As robust as the three-layer model of the hybrid HCAI already is, I believe it nevertheless suggests a further differentiation. For as soon as the question is not only how human, symbolic, and subsymbolic intelligence are coupled, but also where, between which units, under what conditions of responsibility, and according to what meta-order, two additional functional logics emerge: federation and governance. [27]

This expansion does not stem from an external need for supplementation, but from the internal logic of the model itself. An architecture that connects human, symbolic, and subsymbolic levels can be organized centrally or distributively. It can possess implicit or explicit rules for auditing, liability, revision, and intervention. However, as soon as organizations, networks, administrations, or social infrastructures come into view, the three-layer description is no longer sufficient. Then it becomes crucial how knowledge, data, models, contexts, and responsibilities are distributed spatially and institutionally, and how the overall architecture can be legitimized, controlled, and corrected. [28] This leads me to the following architectural formula: Sub-symbolism scales. Symbolism regulates. Humans decide. Federation distributes. Governance is accountable.[29]

This formula answers five fundamental questions of every effective and legitimate intelligence order: How does the system process complexity? How are results semantically specified and normatively bound? Who bears ultimate responsibility for goals and values? How are knowledge, responsibilities, and perspectives distributed? How does the overall system remain accountable, correctable, and institutionally controllable?[30]

It is precisely from this normative perspective that this expansion proves necessary. If AI is to serve humanity rather than dominate it, an architecture that is merely powerful is not enough. What is required is a form of governance in which technical scalability is bound to semantic comprehensibility, human authority of judgment, distributed sovereignty, and institutional accountability. [31]

In this context, federation means far more than technical decentralization. It refers to the distributed anchoring of knowledge, data, responsibilities, and spheres of judgment. Such a distribution protects against epistemic monopolization and the concentration of organizational power. Governance, in turn, refers not merely to external regulation, but to the meta-order of the entire system: the rules of the rules, the procedures for revision, and the institutionalization of audit, intervention, and liability.[32] Only through the interplay of these two additional logics does a three-layered cooperation architecture become a complete architecture of intelligence and responsibility.

8. THE FIVE FUNCTIONAL LOGICS OF THE COMPLETE ARCHITECTURE

The complete architecture of socially responsible AI can only be adequately described if its five functional logics are understood not as separate elements but as a constitutive structure.

The first functional logic is “scaled subsymbolism”: subsymbolic processes form the system’s operational intelligence layer. They process large, dynamic, and heterogeneous data spaces, recognize patterns, identify anomalies, generate predictions, and support adaptive optimization.[33] Their strength lies in their scope, density, and flexibility. They are indispensable, especially in environments of high complexity and uncertainty. Their limitation, however, lies in normative underdetermination. They provide functional plausibility but no sufficient justification of validity.[34]

The second functional logic is: Symbolism regulates. The symbolic level translates statistical patterns into explicit concepts, rules, roles, processes, and normative structures. It creates explainability, verifiability, consistency, and institutional compatibility. Symbolism is not merely a subsequent layer of explanation, but the regulative constitution of AI’s use.[35] This insight is directly applicable to current governance frameworks. ISO/IEC 42001:2023 formulates requirements for a management system for artificial intelligence, thereby pointing to the necessity of explicit organizational layers of rules and control.

The third functional logic is: Humans decide. The human level remains the supreme authority for legitimacy. Humans define purposes, prioritize values, evaluate exceptions, interpret situations that cannot be fully formalized, and bear responsibility for the consequences of collective decisions. [36] If AI is to help people make wiser decisions, this presupposes that its architecture strengthens human judgmental authority rather than undermining it.[37]

The fourth functional logic is: Federation distributes. This brings the spatial-organizational dimension of the architecture to the forefront. Knowledge, data, models, preferences, and responsibilities must not necessarily be merged into a single central authority. Local contexts, institutional distinctions, and domain-specific responsibilities must be preserved.[38] From a normative perspective, federation protects against the concentration of power and ensures plurality, context sensitivity, and distributed sovereignty. [39]

The fifth functional logic is: Governance is accountable. Governance constitutes the overarching framework of the entire system. It determines who may set and change rules, who trains, validates, or restricts models, which audit mechanisms apply, when human intervention is required, how conflicts are handled, and who is liable in the event of damage. [40] This insight corresponds both to Regulation (EU) 2024/1689 and to ISO/IEC 42001:2023, both of which emphasize that governance is a core condition of institutionalized AI practice.

These five functional logics are not additive but constitutive. If one of them is missing, a complete architecture of socially responsible intelligence does not emerge, but only a partial precursor: powerful, but blind; rule-bound, but inflexible; humanly legitimized, but organizationally overwhelmed; distributed, but with diffuse responsibility; or formally regulated, but without real intelligence capabilities.[41]

9. RECURSIVE COUPLING, APPLICATION CONTEXT, AND THE SHIFT IN THE AGI DEBATE

The five functional logics must not be misunderstood as a rigid linear sequence. In real socio-technical systems, they are part of recursive feedback loops.[42] Sub-symbolic outcomes influence symbolic categorizations and the application of rules. Symbolic rules alter the conditions under which machine models are deployed. Human decisions revise goals, priorities, and rule sets. Federated units report back new contexts and divergent assessments. Governance adapts intervention thresholds, roles, and audit standards to changing risks and experiences.[43]

This recursivity connects to my reference to cybernetic concepts of second-order stability. A responsible AI architecture must enable change without losing accountability, and ensure accountability without blocking the ability to learn. [44] It is precisely in this that its superiority lies over both rigid bureaucratic control regimes and unchecked technocratic self-optimization models. An intelligent order is not one that prevents change, but one that organizes change in such a way that it remains correctable, legitimizable, and accountable.

The significance of this approach becomes particularly clear when interpreted as a socio-technical architecture for organizations and societies. In companies and administrations, a hybrid HCAI architecture opens up the possibility of relieving overburdened central control through role-based, rule-bound, and AI-supported self-organization. [45] In networked organizations and integrated systems, a federated architecture enables cooperation without complete centralization.[46] Finally, in social infrastructures—such as mobility, education, energy, or administration—hybrid HCAI can be interpreted as a building block of a digital constitution. [47]

For the AGI debate, this implies a profound conceptual shift. General AI should not be judged solely by its problem-solving capacity, but by the way its intelligence is institutionally organized.[48] A mature form of advanced AI would therefore not simply be a system that performs as many tasks as possible autonomously and efficiently. Rather, a more mature form would be a network of systems in which machine scaling, symbolic rule-binding, human judgment, federated distribution, and governance are coupled in such a way that a stable and accountable overall intelligence emerges. [49] The benchmark for AI, extending all the way to AGI, therefore lies not in maximizing computational power or autonomy, but in whether its architecture helps humanity live better, make wiser decisions, and solve major shared problems without undermining freedom, dignity, and responsibility.[50]

10. GARY MARCUS: WHY THE AGI HYPE IS PROBLEMATIC FROM BOTH AN ANALYTICAL AND POLITICAL PERSPECTIVE

Gary Marcus plays a peculiar dual role in the current AGI debate. He is neither among those who consider AGI to be an empty concept in principle, nor among the enthusiasts who already view large language models as a direct path to it. Rather, he argues that while AGI may be a possible long-term goal, large language models are not the right path, either technically or morally. [51] His criticism is directed both at the overestimation of current systems and at the rhetoric that portrays AGI as being almost immediately imminent.

Marcus’s objection is central to a scientific critique of AGI because it highlights the difference between performance illusion and robust generality. LLMs may appear impressive in many open-ended language tasks, but they often remain fragile in areas such as causal explanation, consistent planning, reliability, reality-saturated modeling, and unusual problem situations. Precisely for this reason, Marcus considers it misguided to hastily conclude from the success of such systems that general intelligence will be achieved soon.[52] His position thus supports a sober thesis: there is a vast conceptual and technical gap between statistically powerful text generation and robust general problem-solving ability. Even more politically significant is Marcus’s meta-critical point. Exaggerated warnings about AGI and promises of salvation through AGI can have the same effect: further fueling the hype surrounding a handful of leading models and companies.

When governments and the public believe that a revolutionary general AI is just around the corner, extraordinary investments, concentrations of power, and special security privileges easily appear rational. Then, supposed caution becomes an accelerator of concentration.

This diagnosis connects Marcus’s critique with my own. Both perspectives highlight that AGI rhetoric is not neutral. It influences financial markets, regulation, corporate valuation, and geopolitical strategies. The difference lies in the fact that Marcus primarily emphasizes the cognitive and technical inadequacy of today’s LLMs, while my critique focuses on the economic appropriation of collective intelligence, architectural flaws, and the question of institutional power. Both perspectives complement each other. Taken together, they form a strong thesis: The dominant AGI discourse overestimates the autonomy of today’s systems while simultaneously underestimating the role of social infrastructures, human contributions, and institutional power.

11. HAUSMANN AND VELASCO: AGI, THE GLOBAL ECONOMY, AND THE LOGIC OF GLOBAL PROFIT EXTRACTION

Ricardo Hausmann and Andrés Velasco offer the most compelling macroeconomic perspective on the current AI debate. Their starting point is not the technical question of whether AGI will be achieved soon, but rather the financial and geopolitical implications of today’s AI valuations. They argue that the market valuations of a core group of major AI companies—including Nvidia, Alphabet, Apple, Microsoft, Meta, Broadcom, Tesla, OpenAI, Anthropic, SpaceX-xAI, and AWS—appear plausible only if these companies generate massive additional annual foreign revenues by 2036.

Under conservative assumptions, they estimate approximately $2.4 trillion in additional annual foreign revenue; this figure would roughly correspond to the total value of current U.S. goods exports and would be significantly higher than the U.S. current account deficit. [53] At the same time, they point out that these returns would flow to companies that, collectively, employ fewer than one million people. Their pointed conclusion is therefore that this is not a story of broadly distributed employment effects, but a story of a small group’s claims on the future income of the rest of humanity.[54]

This diagnosis is of the utmost importance for any normative AI theory. For it suggests that the AI boom should not be understood primarily as a boost to a broadly based innovation economy, but rather as a potential transition to a global economy in which access to indispensable AI infrastructure is priced as a rent. In economics, rent is a regular income earned without any direct consideration. Hausmann and Velasco explicitly write that U.S. power in the 21st century could increasingly be based on the ownership of indispensable AI infrastructure; the challenge for the rest of the world will lie in how this access is to be financed.[55] Thus, AI appears not merely as a technology, but as a strategic bottleneck.

This analysis directly intersects with my critique of AGI. If AGI—or even the anticipation of AGI—serves as a narrative to justify extraordinary valuations, massive infrastructure investments, and geopolitical primacy, then the concept represents not merely a technical project, but a claim to future global cash flows. In this interpretation, AGI becomes the ideological and economic hub of a new form of rentier capitalism. Rentier capitalism refers to an economic system in which landowners lease their land to tenants for cultivation in exchange for a substantial share of the harvest (50 percent or more).

Whoever controls the basic models, chips, clouds, security standards, and integration interfaces holds a privileged position in the extraction of global productivity gains. Scientifically, a clear distinction can be drawn from this. As a general-purpose technology, AI can unlock significant productivity potential; the OECD notes that generative AI exhibits key characteristics of such general-purpose technologies, including scalability, continuous improvement, and innovation-generating spillover effects.[56]

However, the nature of a general-purpose technology does not automatically imply an egalitarian distribution of its benefits. On the contrary: foundational technologies in particular can generate strong concentration tendencies in early phases when complementary goods, infrastructure, and standards are concentrated in the hands of a few. The question of growth and prosperity for all cannot therefore be answered by technological characteristics alone, but only through ownership, competition, labor market institutions, and policy design.

12. PAUL KRUGMAN: THE OPEN QUESTION OF DISTRIBUTION, DEMOCRACY, AND EQUITY

Paul Krugman’s article „The Economics of Technological Change“ published on March 1, 2026, is significant precisely because it does not claim to offer a simple prediction. Already in the freely accessible introduction, Krugman states that his goal is to create an intellectual framework within which to consider possible economic scenarios involving AI; such an overview cannot say “what will happen next,” but it does provide important context for any economic scenario-building regarding AI.[57]

The concluding thought I cited—that AI might prove to be a “balancing force”—should be read in this context: not as a forecast, but as the opening of a space of possibilities. This caution is economically justified. The history of technological change shows neither a linear progression toward universal prosperity nor an automatic tendency toward impoverishment. Technological upheavals can increase productivity, depress wage shares, create monopolies, open up new industries, and concentrate or decentralize political power. Krugman’s approach reminds us that technologies do not “intrinsically” carry fixed distributional effects.

Their impacts depend on institutions, market structures, and political regulation. On this point, his stance can be directly linked to that of Hausmann and Velasco. The same AI surge can manifest either as a tax-driven concentration of power or as a broad boost to productivity—depending on how ownership, access, and competition are organized. Krugman is therefore particularly fruitful for my argument. For he allows me to frame my critique of AGI not as a technophobic rejection, but as a call for institutional reform. The alternative is not “AI or no AI,” but “a concentrated AI economy or a democratically embedded AI economy.” Only this distinction makes it clear why AI, despite real risks of concentration, could still become an instrument of broad prosperity.

13. UNDER WHAT CONDITIONS DOES THE ECONOMY ACTUALLY GENERATE GROWTH AND PROSPERITY FOR ALL?

It follows from the above considerations that the impact of AI on growth and prosperity does not depend on whether AGI is achieved in the strong sense. Rather, what matters is how AI is organized institutionally and architecturally. Four conditions are particularly important in this regard.

First, there needs to be broad access to infrastructure rather than proprietary bottlenecks. If powerful AI is available only through a few proprietary clouds, models, and interfaces, this may create application opportunities for many users, but the structural costs remain with the infrastructure owners. Widespread prosperity therefore requires open or at least competitively accessible foundational layers: interoperable standards, fair licensing and access models, connectivity options for smaller companies, public research, and, if necessary, public or cooperative alternatives. Hausmann and Velasco point to the geopolitical sensitivity of AI infrastructure precisely because control over this layer can generate disproportionate global income claims.[58] This insight can be directly linked to the architectural logic of federation. A distributed intelligence framework, in which knowledge, models, data, and responsibilities are not necessarily consolidated into a single central authority, improves not only the balance of power but also the prospects for a broader economic diffusion of productivity gains.

Second, technological change must complement human work rather than merely replace it. The form of technological change that is most beneficial in terms of productivity and income distribution is generally the one that complements human capabilities, not the one that indiscriminately replaces them. Marcus’ skepticism regarding the primacy of general-purpose chat systems can be interpreted in economic terms: The more AI functions as specialized, verifiable support in medicine, science, technology, education, law, or administration, the more it enhances the performance of many professions. Conversely, the more it is deployed as a generic substitute for human judgment in error-critical environments, the greater the risks of quality loss, devaluation of expertise, and damage to trust. [59] This is precisely where the hybrid HCAI model comes in. It conceives of socially sustainable AI not as a replacement architecture, but as a cooperative architecture. Machines should not relieve humans of responsibility, but rather expand human and collective capacity for action.

Third, there is a need for fair feedback mechanisms in the production of social knowledge. My critique of the appropriation of collective intelligence points to a structural distribution problem: When users continuously contribute to improving systems through interaction, corrections, and knowledge contributions without being part of the resulting value chains, an asymmetrical system of exploitation emerges. Broad prosperity therefore requires mechanisms for fair feedback: transparency regarding data use, participation models, accountability, collective negotiation of usage rights, and institutional forms that do not fully translate societal knowledge production into private wealth accumulation. In a different context, the OECD emphasizes the importance of human agency, oversight, and responsibility; for a democratic AI economy, these principles would need to be expanded to include issues of ownership and participation.[60]

Fourth, democratic governance of information, power, and quality is indispensable. This is where my four fundamental questions become particularly relevant: How does the right information reach the right people at the right time? How are interactions organized on an equal footing? How can values, power, and information flows be dynamically balanced? And how can a platform improve in quality as its user base grows? These questions precisely describe the institutional requirements for an emancipatory hybrid HCAI architecture. Such an architecture would not simply be “more AI,” but a form of collective decision-making infrastructure: participatory, adaptive, verifiable, deliberative, and capable of learning. Only when the quality of coordination increases with the size of the system—rather than devolving into noise, manipulation, or centralized control—can AI become an instrument of responsible collective rationality.

It is precisely at this point that it becomes clear that my vision goes beyond classical AGI debates. It does not aim at the reproduction of an individualized machine mind, but at the institutional enhancement of society’s problem-solving capacity. Scientifically, this could be described as a transition from a machine-centered to a socio-technical theory of general intelligence.

14. WHY THE AGI DEBATE NEEDS TO BE FRAMED ANEW

The conventional AGI debate is too narrow because it sets the wrong priority. It asks above all: When will machines become generally intelligent? The more important question, however, is: What form of social intelligence emerges from the interaction of people, institutions, and machines—and whom does it serve? If one takes this shift seriously, AGI no longer appears primarily as a technical threshold, but as a political, institutional, and architectural problem of order. A society must then decide whether to organize AI as a monopolistic extraction infrastructure, as a military-strategic lever, as a labor-replacing cost-cutter, or as a democratically embedded resource for productivity and coordination. The term AGI does not lose its significance as a result, but it is brought back from the realm of myth into the realm of institutions.

Within this framework, my position can be clearly articulated: AGI in the desirable sense would not be the rule of a centralized superintelligence, but rather the decentralized symbiosis of many self-determined human intelligences with the computational power and scalability of machine AI systems. The measure of success for such an order would not be the media spectacle of individual models, but rather the question of whether it strengthens freedom, productivity, judgment, democracy, and prosperity on a broad scale.

The reconstruction of the path from the BCM model to hybrid HCAI and on to federated neuro-symbolic hybrid collective intelligence shows that this approach extends far beyond a narrative of organizational history or technological development. It articulates a fundamental architectural principle that is central to the future of advanced AI: It is not the isolated performance of individual systems that determines their societal maturity, but rather the way in which intelligence is embedded in orders of self-organization, responsibility, communication, distribution, and governance.[61]

BCM can be understood as an early organizational prototype of a human-centered, feedback-based, and role-oriented self-organization architecture. Hybrid HCAI extends this logic by systematically linking human, symbolic, and subsymbolic intelligence. The proposed extension to federated neuro-symbolic Hybrid HCAI highlights that two additional functional logics are indispensable for a complete societal architecture: federation and governance.[62]

This yields the integrated architectural formula: Subsymbolism scales. Symbolism regulates. Humans decide. Federation distributes. Governance is accountable. This formula does not denote a loose collection of desirable characteristics, but rather the fundamental constitution of a complete architecture of intelligence and responsibility. Only through the recursive interplay of these five logics does a form of organized collective intelligence emerge that is not only powerful, but also transparent, legitimate, pluralistic, controllable, and correctable.[63]

15. CONCLUSION: AGI AS A QUESTION OF JUST ARCHITECTURE AND COLLECTIVE INTELLIGENCE

The question “What is AGI?” can only be meaningfully answered from a scientific perspective if it is not reduced to the supposed autonomy of individual machines. This analysis has shown that while the classical definition of AGI as intelligence that is generally human-like or surpasses human capabilities has heuristic value, it is insufficient for a political economy and a responsible architecture of AI. It overlooks the fact that modern AI systems are embedded in social infrastructures, build upon human knowledge, and are mediated through ownership, standards, data centers, data access, role structures, platform architectures, and institutional procedures.

Gary Marcus reminds us that, despite their capabilities, large language models do not yet represent robust general intelligence and that the AGI hype can jeopardize both analytical clarity and political sobriety.[64] Hausmann and Velasco demonstrate that behind today’s AI valuations lies a conceivable global economy in which a few companies assert enormous claims to income against the rest of the world. [65] Krugman, on the other hand, leaves open the possibility that AI could also be a balancing force under different institutional conditions.[66] Taken together, these positions suggest that the future of AI is not predetermined by technology. It is open—but not neutral.

In this light, my alternative conception of AGI takes on greater clarity. Its normative core lies in conceiving of intelligence not as the property of individual machines or corporations, but as an organized, democratically framed symbiosis of human and machine capabilities. This idea is scientifically viable, provided it is interpreted not as a mere technological vision, but as an institutional and architectural program: as a call for participatory governance, fair distribution of knowledge gains, open infrastructure, qualitative scaling of collective decision-making processes, federated distribution of authority, and effective anchoring of AI to human responsibility.

I therefore come to the following conclusion: The true purpose of artificial intelligence, extending all the way to AGI, does not lie in “ever-increasing performance.”

What matters far more is whether such systems help humanity live better lives, make wiser decisions, and solve major shared problems without undermining freedom, dignity, and responsibility. The crucial question, therefore, is not primarily what AI or AGI can do, but whom it serves, who controls it, and according to what values it acts.[67]

Artificial intelligence does not create growth and prosperity for all simply because it appears as spectacular, autonomous, or highly valued as possible. It creates them when its productivity gains are widely disseminated, its infrastructure remains accessible, its use complements rather than devalues human capabilities, and its institutional rules are subject to democratic oversight. If these conditions are not met, AI will primarily serve to increase the power and wealth of a small elite. If they are met, it could indeed become the balancing force that Krugman holds open as a real possibility.

In this sense, the AGI debate should be framed less as a race toward artificial omnipotence and more as a search for a more equitable architecture of collective intelligence. It is precisely by this standard that the federated neuro-symbolic hybrid HCAI architecture, in my view, proves to be a superior model. It links technical power to clarity of rules, human authority of judgment, distributed sovereignty, and institutional accountability. In doing so, it describes not only a powerful AI, but a form of organized intelligence that can serve humanity without disempowering it.[68]

METHODOLOGY AND ACKNOWLEDGMENTS

ChatGPT (OpenAI, version GPT-4o/5) was used to assist in drafting individual sections of text. This tool helped with spelling, grammar checking, restructuring sentences, and improving clarity. The generated content was critically reviewed and revised by the author, who is responsible for the final version. The actual ideas, arguments, and interpretations in this document are the author’s own.

In this context, I found that ChatGPT either crashed or provided contradictory and illogical answers to complex questions. Only after an extensive dialogue with ChatGPT did I receive coherent results. In other words: Through our dialogue, ChatGPT learned to understand and reproduce causal and logical relationships.

Since I assume that other ChatGPT users have had similar experiences, I would like to thank all the authors mentioned and those unknown who, consciously or unconsciously, shared their knowledge with ChatGPT!

The translation into English was done automatically using DeepL.

© 2026 Friedrich R. Schieck – BCM Consult

FOOTNOTES 

[1] See the standard encyclopedic description of AGI as a hypothetical AI capable of performing “virtually all cognitive tasks.”

[2] OECD, Explanatory Memorandum on the Updated OECD Definition of an AI System, OECD Artificial Intelligence Papers, No. 8, March 2024.

[3] This shift from a question of performance to one of order encapsulates the central normative question of this article.

[4] See Friedrich Reinhard Schieck, “From the BCM Model to Hybrid HCAI – Part I: The Story of an Idea Whose Time Has Come,” in: Journal of Strategic Innovation and Sustainability 20 (2025), No. 4; additionally, Ben Shneiderman, Human-Centered AI, Oxford 2022.

[5] See Schieck, “From the BCM Model to Hybrid HCAI”; also Shneiderman, Human-Centered AI.

[6] Ricardo Hausmann and Andrés Velasco, “The Real Question About the AI Future,” Project Syndicate, April 8, 2026.

[7] This five-part structure is an architectural-theoretical extension of the three-layer hybrid HCAI model.

[8] For the common popular and encyclopedic usage of the term AGI, see the corresponding overview entry.

[9] Gary Marcus, “Is AGI the right goal for AI?”, October 16, 2025.

[10] OECD, Explanatory Memorandum on the Updated OECD Definition of an AI System, op. cit.

[11] Ibid., on fine-tuning, continuous training, and the overlap between the development and deployment phases.

[12] Cf. Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[13] The older BCM publications from 1996, 1998, and 2003 are cited in the JSIS article;

[14] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[15] See ibid.

[16] See ibid.

[17] Erik Brynjolfsson/Daniel Rock/Chad Syverson, “Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics,” NBER Working Paper 24001, Cambridge, MA 2017.

[18] See Schieck, 2025.

[19] Daron Acemoglu/Simon Johnson, Power and Progress: Our Thousand-Year Struggle over Technology and Prosperity, New York 2023.

[20] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[21] See ibid.

[22] See ibid.; additionally, Shneiderman, Human-Centered AI (as in note 4).

[23] Vgl. Schieck, „From the BCM Model to Hybrid HCAI“ (wie Anm. 4).

[24] Vgl. ebd.

[25] Vgl. Ben Shneiderman, Human-Centered AI, Oxford 2022.

[26] This intensification arises from the combination of the model with the normative guiding principle presented here.

[27] This extension follows from the application of the organizational architecture to larger socio-technical systems.

[28] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4); also Regulation (EU) 2024/1689 and ISO/IEC 42001:2023.

[29] The formulation takes up the dictum “Subsymbolics scales, symbolism regulates—humans decide” and systematically expands it to include federation and governance.

[30] Author’s own theoretical synthesis.

[31] Ibid.

[32] On the institutional dimension of governance, see also the EU AI Act and ISO/IEC 42001.

[33] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[34] On the difference between statistical performance and normative justification, see also Shneiderman, Human-Centered AI (as in note 4).

[35] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[36] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4); Shneiderman, Human-Centered AI (as in note 4).

[37] See Shneiderman, Human-Centered AI (as in note 4).

[38] This distinction corresponds to the federated interpretation of hybrid HCAI developed here.

[39] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[40] See ibid.; also Regulation (EU) 2024/1689 and ISO/IEC 42001:2023.

[41] Author’s own systematic refinement.

[42] This is a systems-theoretical reconstruction of the architecture, not a verbatim reproduction from a single source.

[43] Author’s own synthesis based on Schieck and cybernetic feedback logic.

[44] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4), with reference to second-order stability.

[45] See ibid.

[46] See ibid.

[47] This is consistent with, for example, Dirk Helbing, *Thinking Ahead: Essays on Big Data, Digital Revolution, and Participatory Market Society* (Cham, 2015).

[48] My own elaboration of the argument presented so far.

[49] Ibid.

[50] This passage encapsulates the main normative thesis of this article.

[51] Marcus explicitly writes that AGI could change the world, but that LLMs are neither morally nor technically the right path to achieving this.

[52] The detailed technical weaknesses of today’s LLMs are drawn from the consolidated Marcus reconstruction; what is publicly verifiable above all is his fundamental skepticism toward LLMs as a path to AGI.

[53] The specific figures regarding additional annual foreign revenues, margin, and valuation assumptions are drawn from the argumentation presented by Hausmann/Velasco; the general thrust of the paper is publicly accessible.

[54] The formulation that the issue is not one of broadly distributed employment but rather of the claims of the few on the future incomes of the many corresponds to the line of argumentation in the paper.

[55] Hausmann and Velasco explicitly emphasize that U.S. power in the 21st century could increasingly depend on the possession of indispensable AI infrastructure.

[56] OECD, Is Generative AI a General Purpose Technology?, OECD, 2025.

[57] Paul Krugman, “The Economics of Technological Change,” March 1, 2026.

[58] Hausmann/Velasco, op. cit.

[59] Marcus, op. cit.

[60] OECD, AI Principles overview and Explanatory Memorandum on the Updated OECD Definition of an AI System.

[61] See Schieck, “From the BCM Model to Hybrid HCAI” (as in note 4).

[62] Author’s own theoretical extension.

[63] Ibid.

[64] Marcus, op. cit.

[65] Hausmann/Velasco, op. cit.

[66] Krugman, op. cit.

[67] This passage formulates the concluding normative thesis of the article.

[68] Summary conclusion of this article.

REFERENCES

  • Acemoglu, Daron/Johnson, Simon (2023): Power and Progress: Our Thousand-Year Struggle over Technology and Prosperity, New York 2023.
  • Brynjolfsson, Erik/Rock, Daniel/Syverson, Chad (2017): “Artificial Intelligence and the Modern Productivity Paradox: A Clash of Expectations and Statistics,” NBER Working Paper 24001, Cambridge, MA 2017.
  • European Union: Regulation (EU) 2024/1689 of the European Parliament and of the Council of June 13, 2024, establishing harmonized rules for artificial intelligence (Artificial Intelligence Act).
  • Hausmann, Ricardo/Velasco, Andrés (2026): “The Real Question About the AI Future,” Project Syndicate, April 8, 2026.
  • Helbing, Dirk (2015): Thinking Ahead. Essays on Big Data, Digital Revolution, and Participatory Market Society, Cham 2015.
  • International Organization for Standardization: ISO/IEC 42001:2023 Information technology — Artificial intelligence — Management system, Geneva 2023.
  • Krugman, Paul (2026): “The Economics of Technological Change,” Substack, March 1, 2026.
  • Marcus, Gary (2025): “Is AGI the right goal for AI?”, Substack, October 16, 2025.
  • OECD (2024): Explanatory Memorandum on the Updated OECD Definition of an AI System, OECD Artificial Intelligence Papers, No. 8, March 2024.
  • OECD: “AI Principles overview,” OECD.AI.
  • OECD (2025): Is Generative AI a General Purpose Technology?, OECD, 2025.
  • Schieck, F. (1996). Systemic and Holistic Management Based on a Behavior-Driven Organizational Method;
  • Schieck, F., & Tauber, H. (1998). Business Communication Management – The Path to Enterprise-Wide Knowledge Management. OrgNews GfO;
  • Schieck, F. (2003). The Paradox of Consulting – Or of Rapid Change and Outdated Concepts. Magazine for Digitalization, Networking & Collaboration.
  • Schieck, F. (2023). Where Is the Flaw in the Digital Transformation System …,
  • Schieck, F. (2024): When will AI become the killer app for productivity growth and bureaucracy reduction in companies …,
  • Schieck, F. (2025): Hybrid-HCAI – A thought experiment on the ultimate symbiosis of human and artificial intelligence …,
  • Schieck, F. (2025): „From the BCM Model to Hybrid HCAI – Part I: The Story of an Idea Whose Time Has Come“, in: Journal of Strategic Innovation and Sustainability 20 (2025), H. 4, DOI: 10.33423/jsis.v20i4.8022.
  • Schieck, F. (2025): From the BCM Model to Hybrid HCAI – Part 1,
  • Schieck, F. (2026): The Individualized and Adaptive Digital Workplace
  • Shneiderman, Ben (2022): Human-Centered AI, Oxford
  • Wikipedia: “Artificial general intelligence,” English version, last crawled in April 2026.

2 Responses

Leave a Reply

Your email address will not be published. Required fields are marked *