Editor's Note:

Brain-inspired intelligence is becoming an important breakthrough direction for next-generation artificial intelligence. Recently, a Science and Technology Daily reporter conducted an exclusive interview with Zhang Xu, Academician of the Chinese Academy of Sciences and President of Guangdong Institute of Intelligence Science and Technology, holding a dialogue on the conceptual boundaries, core advantages, industrialization bottlenecks, and future trends of brain-inspired intelligence. The full text is reproduced below:

Brain-Inspired Intelligence Expected to Reach Millions of Households Within Five Years

Multiple brain-inspired intelligence products were released at the 2026 World Artificial Intelligence Conference, drawing widespread attention; Yangpu District in Shanghai issued the "Several Measures of Yangpu District on Promoting the Development of the Brain-Inspired Intelligence Industry"... In recent years, the popularity of brain-inspired intelligence has continued to rise. According to the blue book "China Future Industry Science and Technology Innovation Development Report (2026)," the total investment and financing in China's brain-inspired intelligence field has exceeded 3 billion yuan.

What core advantages does the increasingly popular brain-inspired intelligence possess? What challenges does its industrialization path face? How can the rapid development of the brain-inspired intelligence industry be promoted?

Recently, a Science and Technology Daily reporter conducted an exclusive interview with Zhang Xu, Academician of the Chinese Academy of Sciences and President of the Chinese Neuroscience Society, holding an exchange on topics such as the conceptual boundaries, technology development trends, and industrialization paths of brain-inspired intelligence.

AI Breakthroughs Require Brain-Inspired Intelligence

Reporter: At present, the cross-disciplinary integration of artificial intelligence and neuroscience is attracting increasing attention. What is the relationship between the two?

Zhang Xu: Artificial intelligence and neuroscience have formed a bidirectional, mutually reinforcing closed loop. On the one hand, the development of neuroscience provides bionic ideas and theoretical references for artificial intelligence. On the other hand, artificial intelligence has become a core tool for neuroscience research.

Over the past decade, the biomedical field has accumulated massive amounts of digital research data. It is difficult to tap the deep value of these data relying solely on traditional analytical methods. The systematic and comprehensive analytical capabilities of artificial intelligence can help researchers discover patterns that were previously difficult to detect and identify potential application scenarios.

Today, artificial intelligence has become an essential tool for researchers and clinicians. It can even be said that without understanding the application logic of AI for Science, one will fall behind the times.

Reporter: In the interdisciplinary field of brain science and artificial intelligence, important concepts such as "brain intelligence" and "brain-inspired intelligence" have emerged successively. What is the difference between these two concepts?

Zhang Xu: "Brain intelligence" and "brain-inspired intelligence" are not the same concept. "Brain intelligence" is the cross-disciplinary integration of brain science and artificial intelligence, a bridge connecting brain science and artificial intelligence. The development of artificial intelligence today mainly relies on disciplines such as computer science, mathematics, and statistics, and is not closely connected with the frontier progress of modern neuroscience; there is an obvious research gap between the two. In 2014, the original intention behind my leading the establishment of the "Shanghai Brain-Intelligence Engineering" research program was to bridge the gap and fill this void.

With the deepening of the cross-disciplinary integration of artificial intelligence and brain science, as well as the iterative development of related technologies, the concept of "brain-inspired intelligence" emerged as the times required.

Brain-inspired intelligence is an intelligent science and technology inspired by brain science and belongs to the category of technical science. Its core lies in drawing on the principles of the human brain's perceptual and cognitive neural networks and their working mechanisms to engineer intelligent frameworks, algorithm models, and intelligent computing systems inspired by brain information processing mechanisms.

Compared with brain intelligence, the concept of brain-inspired intelligence is clearer and more accurate, and its industrialization path is also more definite. Today, we have made it clear that brain-inspired intelligence is a completely new artificial intelligence theory and technology system oriented toward the future.

Reporter: Why is it said that brain-inspired intelligence is an important breakthrough direction for next-generation artificial intelligence?

Zhang Xu: Artificial intelligence absorbed basic discoveries from brain science as early as the 1950s and 1960s, but later it was mainly driven by internet big data, mathematics and statistics, and computer technology. Today it has entered a technological bottleneck—excessive reliance on mainstream artificial intelligence computing power represented by graphics processing units (GPUs) and massive data, resulting in high energy consumption, high costs, and data security risks.

Brain-inspired intelligence, by contrast, takes a different path, simulating the neurons of biological brains and their neural network computing systems. Compared with traditional artificial intelligence computing power solutions, its hardware scale, energy consumption requirements, usage costs, and computing power construction costs are substantially reduced. It can ensure data security and enable proprietary multimodal model training and inference, making it particularly suitable for developers and users of artificial intelligence application technologies, as well as related enterprises, research institutions, and hospitals, and can serve as the foundation for intelligent agents, intelligent robots, and embodied intelligence.

Here I will give an intuitive comparison: the human brain has 86 billion neurons and consumes only 20 watts of energy, while an artificial intelligence system of the same scale consumes as much as 8 megawatts. Behind this enormous difference lies the key to the next generation of intelligent technology transformation.

Reporter: Could you elaborate on the core advantages of brain-inspired intelligence?

Zhang Xu: From the underlying technology perspective, traditional von Neumann architecture computers, relying solely on engineering means such as 3D stacking and advanced packaging, can only slightly reduce energy consumption and improve computing power, and cannot solve the fundamental problem of separating storage and computing. Brain-inspired intelligence, however, is a fundamental transformation of the underlying architecture. Relying on bionic neural units and brain-inspired neural network architectures, it solves the problem of high energy consumption in large models at the root, and at the same time possesses the characteristic of transparency and interpretability, which can avoid the hallucination problem of traditional large models.

In terms of privacy protection, brain-inspired intelligence also has significant advantages. Current large models pursue publicization and standardization, whereas brain-inspired intelligence supports localized and privatized training—personal health data, household data, enterprise data, and the like can all be kept entirely local, achieving a high degree of personalization while safeguarding personal privacy and enterprise security.

Broad Prospects for the Brain-Inspired Intelligence Industry

Reporter: What are the common bottlenecks currently facing the industrialization of brain-inspired intelligence?

Zhang Xu: Through years of effort by researchers, the brain-inspired intelligence field has gradually formed a connected innovation chain: from basic research on brain perception and cognitive neural networks, to brain-computer interface encoding and decoding and regulation technologies, to the development of algorithm models, chip devices, and computing systems. Each link can give rise to corresponding enterprises, and the innovation chain and industrial chain are highly aligned. Brain-inspired intelligence technology is gradually advancing from the research side to the application side.

At present, the most prominent bottleneck in the industrialization of brain-inspired intelligence lies in the fact that the technology system is new, the comprehensive threshold is high, relevant talent is in short supply, and industry-wide popularization has not yet been completed. Technology implementation requires a large number of engineers to master brain-inspired computing technology and develop dedicated equipment, algorithm models, and application scenarios adapted to different industries, which requires a certain development cycle.

Reporter: How do you view the future development trends of brain-inspired intelligence?

Zhang Xu: What is certain is that the development prospects of brain-inspired intelligence are very broad. Its core value is mainly reflected in two aspects: first, achieving the inclusive popularization of intelligent computing power—in the future, ordinary households, small and medium-sized laboratories and research institutions, hospitals of all sizes, and small and medium-sized enterprises can all deploy lightweight brain-inspired intelligent computing devices and independently develop vertical large models adapted to their own needs, truly putting intelligent computing power into the hands of the public; second, the technology will deeply penetrate all sectors of society, giving rise to a large number of entirely new intelligent application scenarios while safeguarding data privacy and core technology security.

Reporter: I have heard that you are cultivating brain-inspired enterprises in Guangdong.

Zhang Xu: Yes. The period from 2014 to 2020 was the initial stage of the brain science and artificial intelligence integration industry, during which a number of landmark enterprises such as Cambricon and United Imaging were born. After 2020, enterprises with clear brain-inspired intelligence attributes have continued to emerge.

Previously, we promoted the "Shanghai Brain-Intelligence Engineering" research program in Shanghai and incubated a number of benchmark enterprises. At present, we are actively building a brain-inspired intelligence industrial ecosystem in Shanghai and Guangdong, and have also incubated a number of brain-inspired enterprises, covering tracks such as brain-inspired algorithm software and high-end intelligent hardware.

Brain-inspired intelligence is a typical high-end manufacturing industry with its own laws of development. Relevant enterprises are currently in a stage of rapid ascent, with good development momentum and enormous potential for social value.

Reporter: What will be the focus of your team's efforts going forward?

Zhang Xu: Research and development in brain-inspired intelligence has shifted from a single direction in the early days to an all-round layout, but we have always focused on tackling fundamental issues such as underlying basic theories and core technologies, which is the core responsibility of research institutions.

Brain-inspired computing systems are highly systematic and interdisciplinary, with long research and development cycles. Ultimately, they must be realized as physical products, verify the theoretical computing power advantages through actual operation, and then iteratively optimize basic research in reverse. Going forward, we will also join forces with industry players to promote the adaptation of brain-inspired technology to existing industrial intelligent systems and solve practical problems at the application level.

Reporter: You mentioned that ultimately the technology must be realized as physical products and basic research iteratively optimized in reverse. What practical experience do you have in connecting the entire chain from basic research to achievement translation?

Zhang Xu: First, basic research and applied research are not contradictory. Original innovation "from 0 to 1" determines whether the industry can achieve technological leadership, and integrated innovation "from 1 to 100" also has enormous industrial value.

We have two main paths for achievement translation. One is to independently incubate entirely new science and technology innovation enterprises around underlying technologies and build entirely new industrial chains. The other is to empower traditional industries, transform existing technology systems, and enable traditional CPUs to fully unleash their computing power advantages under brain-inspired algorithm frameworks.

To connect the chain between basic research and applied research, we must first aim at translation from the very source. Scientific research cannot stop at the level of publishing papers and applying for patents; it must adhere to a problem-oriented approach from the very beginning of research project initiation and focus on major common problems in industrial and social development. For example, artificial intelligence currently faces industry pain points such as high energy consumption and high computing power costs, and these are precisely our core directions for tackling challenges.

How do we find problems? I believe that distilling frontier scientific questions is inseparable from interdisciplinary exchange and collision. I come from a medical background, and during my time at the Shanghai Branch of the Chinese Academy of Sciences, I came into contact with outstanding scientists in many fields such as computer science, mathematics, and physics. It is precisely this multidisciplinary intersection that expanded my research thinking.

Cultivating Young People's Interdisciplinary Thinking

Reporter: Looking to the future, how can the brain-inspired intelligence industry truly achieve breakthroughs?

Zhang Xu: The biggest challenge in the current development of brain-inspired intelligence does not come purely from the current mainstream artificial intelligence computing power competition, but from the insufficient deep intersection and paradigm coordination of multiple disciplines such as neuroscience, brain-computer interfaces, mathematics, computer science, and chip manufacturing. For brain-inspired intelligence to truly achieve breakthroughs, it is necessary to connect the entire chain covering brain cognitive mechanisms, algorithm models, chips, and systems.

If China seizes the opportunity of this round of technological transformation, it is expected to gain the upper hand in certain areas of next-generation artificial intelligence and achieve a leap from "following" to "leading."

At present, China already possesses the basic conditions for "overtaking on a new track" in the brain-inspired intelligence field. The key lies in seizing the opportunities of the times, adhering to both independent innovation and open cooperation, and promoting the translation of technological achievements into industrial practices that benefit the people. This is not only a need for scientific and technological competition, but also a strategic choice to serve the country's high-quality development and satisfy the people's aspirations for a better life.

Reporter: Will brain-inspired intelligence soon reach millions of households?

Zhang Xu: Yes, I expect it will not be long. Conservatively estimated, I believe that within five years, brain-inspired intelligence personal computers (BI PCs) will reach millions of households. Of course, underlying theoretical research and technological development will still require sustained investment over a considerable period, but the pace of technological iteration will be very fast.

To promote the development of the brain-inspired intelligence industry, we cannot rely solely on research funding investment; the core is to grasp two directions.

The first is to build an application ecosystem. Intelligent technology must have scenarios and users to iterate continuously, which requires multi-party coordination among the government, research institutes, enterprises, and industrial funds to build diversified application scenarios at low cost and improve the industrial ecosystem. The second is to strengthen the full-chain connection between the innovation chain and the industrial chain. Research funding cannot focus only on a single link; it is necessary to connect the entire chain from basic research and device development to industrial implementation, avoiding scientific research achievements becoming detached from the innovation chain and unable to be translated.

I believe that brain-inspired intelligence needs to expand in a differentiated manner on the basis of the existing artificial intelligence technology system, and through crowd innovation and resource integration, jointly build an industrial ecosystem and development ecosystem with developers, investors, and consumers.

Reporter: As President of the Chinese Neuroscience Society, how do you think the society can play a role in promoting interdisciplinary integration?

Zhang Xu: The society is a very important think tank platform and plays an important role in promoting interdisciplinary integration. Societies are usually not bound by the interests of institutions or individuals, conduct affairs in a fair and transparent manner, and are truly the home of scientific and technological personnel. At the same time, the academic atmosphere of the society is equal and open, where everyone can freely discuss technical roadmaps and solutions, making it an important carrier of cross-disciplinary innovation.

In recent years, with the support of the China Association for Science and Technology, the Chinese Neuroscience Society has produced a large number of research recommendations and industrial roadmaps on brain-inspired intelligence and related fields. For example, the Chinese Neuroscience Society has organized the compilation of a series of reports such as the "Roadmap for the Development of the Brain-Inspired Intelligence Industry and Technology" and the "Roadmap for the Brain-Computer Interface Industry and Technology," providing important support for national decision-making.

However, benchmarked against outstanding international societies, I believe domestic societies still need to improve in two aspects. First, they need to strengthen internationalization, attract more scholars from different countries to participate in academic conferences, broaden research horizons through cross-national collision of ideas, and enhance the international influence of academic conferences. Second, they need to attract more young people to participate. At present, many young researchers focus excessively on their own paper output and lack awareness of participating in industry exchanges. I hope societies will strengthen support for young people, cultivate young people's interdisciplinary innovative thinking, and broaden their academic horizons, which is crucial to the long-term development of young talent.

Reporter: What suggestions do you have for young researchers in this field?

Zhang Xu: Brain-inspired intelligence is an entirely new track with a large number of directions awaiting exploration, which is a rare opportunity for young researchers. When doing research, do not fixate only on the few papers and sets of data in your own hands; you should proactively step out, participate more in interdisciplinary academic exchanges, come into contact with more frontier achievements, and open your eyes to the world.

Innovation in interdisciplinary fields often comes from the collision of ideas from different disciplines. Young people should maintain an open mindset, respect the research thinking of different fields, learn more from others, and achieve their own growth in the process of solving major common problems.

The future of intelligent technology depends on the younger generation to explore. I hope the younger generation can take root in the fundamentals, orient themselves toward applications, and make contributions belonging to themselves and their teams on this entirely new track.

The full text is reproduced from Science and Technology Daily (September 24, 2026, Page 5, Viewpoint)