Paradigm Revolution in AI: A New Theory of General Intelligence Emerges
In the rapidly evolving landscape of artificial intelligence, where breakthroughs are often measured in months rather than decades, a recent publication has stirred profound discussion among researchers and philosophers of science alike. A paper titled Paradigm Revolution in Artificial Intelligence and the Birth of General Theory of Intelligence by Zhong Yixin, a professor at the AI School of Beijing University of Posts and Telecommunications, presents not merely a technical advancement but a foundational rethinking of the entire discipline. Published in the journal CAAI Transactions on Intelligent Systems, this work challenges the very epistemological and methodological bedrock upon which decades of AI research have been constructed. At its core lies a bold assertion: the field of artificial intelligence has been operating under a fundamental misalignment—a “misplaced paradigm”—that has hindered its progress toward true general intelligence. The proposed solution is nothing short of a scientific revolution.
For over half a century, artificial intelligence has followed a trajectory shaped by the dominant paradigms of the physical sciences—mechanistic, reductionist, and deterministic. These approaches, which proved immensely successful in physics, chemistry, and engineering, emphasize decomposition, formalization, and predictability. Researchers have applied these principles to AI by breaking down intelligence into isolated components: simulating brain structure through neural networks, modeling logical reasoning via expert systems, or reproducing behavior through robotics. While each of these paths has yielded impressive results—deep learning, game-playing algorithms, autonomous robots—they remain, according to Zhong, fundamentally fragmented and incapable of synthesizing a coherent, universal theory of intelligence.
Zhong’s critique is not of the individual achievements but of the overarching framework that guides them. He argues that intelligence, whether human or artificial, is not a static object to be dissected like a machine, but a dynamic, open, and complex information process. It emerges from the continuous interaction between a subject (the intelligent agent) and its environment (the object), under conditions of uncertainty, purpose, and value. To treat such a process using the tools of classical material science—where objects are passive, predictable, and devoid of meaning—is, in his view, a categorical error. This mismatch, which he terms “wearing the wrong hat” or “misplaced paradigm,” is the root cause of AI’s persistent shortcomings: low comprehension, lack of generalization, and the absence of a unified theoretical foundation.
The historical inevitability of this misalignment, Zhong explains, stems from the lag between technological emergence and philosophical maturation. The Information Age began in the mid-20th century with the advent of computers and communication networks, yet the conceptual tools to understand information as a fundamental entity—distinct from matter and energy—have not kept pace. As a result, information science, including AI, inherited the methodologies of the preceding material era. This is not a failure of intent but a consequence of intellectual inertia. Scientists naturally reach for familiar tools when confronting new phenomena. However, as the demands of AI grow more sophisticated, the limitations of these inherited methods become increasingly apparent.
Zhong’s remedy is a paradigm shift—from the mechanistic reductionism of the material sciences to what he calls the “information ecological methodology” of a mature information science. This new paradigm rests on a fundamentally different worldview. Where traditional science sees passive matter governed by fixed laws, the information paradigm sees active agents engaged in purposeful interaction with their environment. The goal is no longer merely to describe structure and function, but to achieve mutual benefit—what Zhong calls “subject-object win-win.” This shift redefines the very nature of scientific inquiry in AI: from analyzing isolated systems to understanding evolving ecosystems of information exchange.
Central to this new framework is the concept of “comprehensive information,” or quan xin xi in Chinese, which Zhong first developed in the 1980s. Unlike classical information theory, which focuses solely on the statistical form of signals (syntax), Zhong’s theory integrates three inseparable dimensions: syntax (form), semantics (content), and pragmatics (value). Information, in this view, is not just a pattern of bits, but a meaningful entity that carries significance for a particular agent in a particular context. A message is not understood merely by decoding its form, but by grasping its meaning and evaluating its relevance to one’s goals. This triadic model—form, content, value—is the cornerstone of what Zhong calls “general intelligence.”
This holistic approach directly addresses the limitations of current AI systems. Modern deep learning models, for instance, excel at pattern recognition but fail at true understanding. They can classify images or generate text without comprehending what they represent or why it matters. This is because they operate on syntactic information alone, stripped of semantic and pragmatic context. Zhong argues that without all three dimensions, there can be no genuine intelligence—only sophisticated mimicry. His theory posits that real understanding arises only when an agent can connect the form of a message to its meaning and its value in pursuit of a goal.
Building on this foundation, Zhong introduces a universal mechanism for intelligence generation: the “law of information conversion and intelligence creation.” This principle describes intelligence not as a fixed attribute, but as a dynamic process of transformation. It begins with external stimuli—raw data from the environment—which are converted into perceptual information, then into knowledge, then into strategic plans, and finally into intelligent actions. Each step involves a complex integration of form, content, and value, guided by the agent’s prior knowledge and objectives. This process is not linear but recursive, with feedback loops allowing for continuous learning and adaptation.
What makes this mechanism truly universal is its applicability to both natural and artificial systems. Zhong contends that the same fundamental process governs human cognition and machine intelligence. The difference lies not in the mechanism itself, but in the substrate: biological neurons versus silicon circuits. This insight bridges a long-standing divide in cognitive science and AI, suggesting that intelligence is not exclusive to humans but a general property of information-processing systems that operate under the right principles.
The implications of this theory are far-reaching. First, it offers a unified framework that integrates the previously fragmented schools of AI. Instead of competing paradigms—structural, functional, behavioral—Zhong proposes a single “mechanismist” approach centered on the dynamics of information conversion. This could end decades of theoretical fragmentation and pave the way for a cohesive research agenda. Second, it redefines the academic identity of AI. Rather than being a subfield of computer science, Zhong sees it as an interdisciplinary convergence of neuroscience, cognitive science, information science, systems theory, and philosophy. This broader perspective acknowledges the complexity of intelligence and the need for diverse intellectual contributions.
Moreover, the theory demands a rethinking of AI’s mathematical and logical foundations. Traditional logic and probability theory, while powerful, are insufficient for capturing the fluid, context-sensitive nature of real-world intelligence. In response, Zhong and his team have developed new formal tools, including “flexible logic” and “factor space theory,” which allow for graded reasoning and the integration of heterogeneous knowledge. These innovations are not mere technical refinements but necessary components of a new scientific language capable of expressing the richness of intelligent behavior.
Perhaps most provocatively, Zhong elevates his principle to the status of a fundamental scientific law, placing it alongside the conservation of mass and energy in physics. Just as matter and energy can be transformed but not created or destroyed, information can be converted into intelligence through the right processes. This “law of intelligence creation” suggests that intelligence is not a mystical property but a natural phenomenon that can be engineered and optimized. It implies that machines can genuinely create value through intelligent action, not just simulate it.
The practical realization of this theory remains a formidable challenge. While the conceptual framework is comprehensive, translating it into working systems requires advances in hardware, software, and system integration. Zhong acknowledges this, noting that the implementation details will be addressed in future work. Yet, the theoretical clarity he provides offers a roadmap for researchers seeking to move beyond narrow, task-specific AI toward systems with broader understanding and adaptability.
Critics may argue that Zhong’s vision is overly ambitious or too abstract to yield tangible results. After all, grand theories of mind have a long history of failing to deliver practical technologies. However, Zhong’s approach differs in its grounding in information theory and systems science, disciplines with strong empirical foundations. His work is not speculative philosophy but a rigorous attempt to formalize the principles of intelligence based on decades of research in communication, computation, and cognition.
Furthermore, the timing of this theory is significant. As AI systems become more powerful and pervasive, questions about their reliability, transparency, and alignment with human values grow more urgent. Current models, despite their capabilities, are often brittle, opaque, and prone to errors when faced with novel situations. Zhong’s emphasis on understanding, context, and purpose speaks directly to these concerns. A system that comprehends the meaning and value of its actions is more likely to behave responsibly and adaptively than one that merely processes patterns.
The broader cultural and philosophical implications are also noteworthy. Zhong draws connections between his “information conversion” model and classical Chinese thought, particularly the Confucian and Daoist emphasis on the interplay between knowledge and action (zhi xing he yi). This synthesis of Eastern philosophy and Western science challenges the dominance of purely mechanistic models in cognitive science and opens new avenues for cross-cultural dialogue in the study of mind.
In an era where AI is often portrayed as an inevitable force driven by technological momentum, Zhong’s work serves as a reminder that science is also a human endeavor shaped by choices—choices about what to study, how to study it, and what questions to ask. By calling for a paradigm revolution, he invites the AI community to reflect on its deepest assumptions and to consider whether the path it is on will lead to truly intelligent machines or merely more efficient automata.
The journey from current AI to general intelligence is not one of incremental scaling, Zhong suggests, but of fundamental reorientation. It requires not just more data or faster processors, but a new way of thinking about what intelligence is and how it arises. His theory offers a compelling vision: a future where machines are not just smart, but understanding; not just efficient, but meaningful; not just tools, but partners in the pursuit of knowledge and progress.
As the field stands at a crossroads, Zhong Yixin’s contribution provides both a critique and a compass. Whether his paradigm will be widely adopted remains to be seen, but its articulation marks a significant moment in the intellectual history of artificial intelligence. It challenges researchers to look beyond the immediate horizon of technical performance and to engage with the deeper questions of meaning, purpose, and the nature of mind itself.
Paradigm Revolution in Artificial Intelligence and the Birth of General Theory of Intelligence by Zhong Yixin, AI School, Beijing University of Posts and Telecommunications, published in CAAI Transactions on Intelligent Systems, DOI: 10.11992/tis.202103042