AETDEW Fellow Professor Zhong Yixin Publishes His Groundbreaking Book “Unified Theory of Intelligence “
AETDEW Fellow Professor Zhong Yixin is an internationally acknowledged pioneer in Artificial Intelligence and the nationally acknowledged father of AI in China. He published his groundbreaking book “Unified Theory of Intelligence “ in Chinese December 2023. He gave a TV speech in China to introduce his book. AETDEW persuaded him to translate the transcript into English and is arranging to have his book ttanslated into English for the benefit of the rest of the world.
An Brief Introduction to Unified Theory of Intelligence
The book, titled “Unified Theory of Intelligence”, was published at the end of 2023 by Science Press, Beijing, China.The major features of the book, particularly interested for science, include:
- A Surprising Finding in AI Research
AI is a kind of complex information system and should accordingly follow the paradigm for information science as guidelines in its research. However, it is found that the paradigm followed in AI since its beginning has been the one for physical science , leading to a serious mismatching between AI research (the feet) and the paradigm followed (the head).
- The Paradigm Revolution in AI
To overcome the problem of the serious mismatching in AI mentioned above, the paradigm for information science was summarized. And then the paradigm for physical science in AI has been replaced by the one for information science. This is called the paradigm revolution in AI.
- The Achievements
Attributing to the Paradigm Revolution in AI
Under the guidance of the paradigm for information science, the pervasive mechanism for intelligence creation in AI has been uncovered, which is the law of “information conversion and intelligence creation”, and the universal theory for AI has been established based on the law.
- The More Significant Result
It is further exposed that the mechanism for intelligence creation in AI is the same as that in humans. As result, the general theory for intelligence has been established.
The paradigm revolution in AI, the pervasive mechanism for intelligence creation in AI, the law of information conversion and intelligence creation, the universal theory of AI, and the unified theory of intelligence are all the results published for the first time in academic world.
In addition, the following explanations may be necessary for reading and understanding. The “paradigm” of a discipline specifically refers to the “scientific view and methodology” that must be followed in that discipline. It is the ultimate guide that directs and regulates the research and development of the discipline. The paradigm belongs to the category of social consciousness. Artificial intelligence (AI) is an open and complex information system, a higher chapter of information science, and should follow the information science paradigm. However, as is well known that social consciousness inevitably far lags behind social existence. Therefore, since the rise of AI in the mid-20th century till the present, the information science paradigm has not yet been formed, and thus the already existing and flourishing physical science paradigm has been borrowed by AI research. The physical science paradigm has been very successful in the realm of physical science research (material science and energy science included), but its borrowing to the field of AI inevitably leads to severe mismatches: the borrowed paradigm does not match the nature of AI research, especially in next two aspects.
(1) The scientific view of the physical science paradigm is mechanistic materialism: it only studies physical objects and rejects subjective factors; its research objective is to explain the structure of matter. Rejecting subjective factors blocks the source of intelligence.
(2) The methodology of the physical science paradigm is mechanistic reductionism: divide and conquer, purely formalized. The former causes AI research to be fragmented into (neural networks), functionalism (expert systems), and behaviorism (senor motor systems) – three paths that do not recognize each other and go their separate ways, unable to form a unified theory. The latter formalizes AI, discarding content and value factors, and thus fails to produce real intelligence.The absence of a unified theory and the lack of real intelligence are the inevitable chronic diseases of all AI research to date (including large language models like GPT series).The manuscript “Unified Theory of Intelligence” clearly demonstrates that in the early stages of AI development (up to date), the misplacement of paradigm is an unavoidable rule; only through a paradigm revolution in AI can the chronic diseases caused by the misplacement of paradigm be completely eradicated, and AI research can be successful.
《统一智能理论》的视频文稿
这是科学出版社新出版的著作《统一智能理论》。它的特点是：（1）突破了自然科学与哲学间的壁垒，发现了人工智能研究范式张冠李戴的问题：用物质学科范式指导信息学科的人工智能研究；（2）通过人工智能的范式革命，提炼了“信息学科范式”，以此取代了人工智能研究一直遵循着的“物质学科范式”；（3）在信息学科范式引领下，揭示了“普适性的智能生成机制”，即“信息转换与智能创生定律”，在此基础上建立了《通用人工智能理论》；（4）进一步发现，“信息转换与智能创生定律”，既是通用人工智能的生成机制，也是人类智能的生成机制，因而建立了《统一智能理论》。人工智能的范式革命、普适性智能生成机制 — 信息转换与智能创生定律、通用人工智能理论、统一智能理论，都是国内外首次面世的学术新成果。
说明：某个学科的“范式”，特指该学科必须遵循的“科学观和方法论”。它是指导和规范该学科研究发展的最高指南；属于社会意识范畴。人工智能是一类开放复杂的信息系统，是信息学科的高级篇章，应当遵循信息学科的范式。可是，社会意识必然远远滞后于社会存在。这是社会发展的铁律。所以，人工智能自20世纪中叶兴起至今，信息学科范式尚未形成，于是就借用了业已存在而且风华正茂的物质学科范式。物质学科范式在物质学科研究领域非常成功，但借用到人工智能领域就必然产生严重的失配问题，借来的范式与自己的研究性质互不匹配：（1）物质学科范式的科学观是机械唯物科学观：只研究物质客体，戒绝主观因素；研究目的是阐明物质的结构。戒绝主观因素，就堵死了智能的源泉。（2）物质学科范式的方法论是机械还原论：分而治之，纯粹形式化。前者使人工智能研究被分解为结构主义（神经网络）、功能主义（专家系统）、行为主义（感知动作系统）三个互不认可分道扬镳的路径，无法形成统一理论；后者使人工智能形式化，丢弃了内容因素和价值因素，无法产生真实的智能。既无统一理论，又无真实智能，这是迄今一切人工智能研究（包括GPT大模型）不可避免的痼疾顽症。《统一智能理论》证明了：在人工智能发展的初级阶段，范式的张冠李戴是无法回避的铁定规律；只有通过人工智能的范式革命，范式张冠李戴所产生的痼疾顽症才能被彻底清除，人工智能的研究才能取得成功。