Nanjing University Establishes Institute of Biological Intelligence: Human-Machine Collaboration from Data Intelligence to Biological Intelligence

Humanity's exploration of "intelligence" is crossing the boundaries of traditional disciplines. In particular, between computer science and the life sciences, research paths that were once relatively independent are moving toward deeper convergence.

This trend is evident in the scientific breakthroughs recognized by the Nobel Prizes. In 2024, the Nobel Prize in Physics focused on artificial neural networks and machine learning, while the Chemistry Prize honored computational protein design and structure prediction; in 2026, the Physiology or Medicine Prize turned its attention to optogenetics—a technology that enables scientists to precisely regulate the activity of specific neurons through light.

These seemingly different breakthroughs reveal two mutually reinforcing lines of research: on the one hand, AI draws inspiration from living systems and in turn becomes an important tool for exploring the mechanisms of life; on the other hand, advances in neuroscience are continually deepening humanity's understanding of the biological foundations of intelligence.

Along this two-way evolutionary path, an early-stage frontier interdisciplinary direction is emerging—biological intelligence.

Behind its rise are also the practical needs of AI's own development. In recent years, AI represented by large models has developed rapidly, but problems such as the growing difficulty of obtaining high-quality data and the sharp increase in computational energy consumption have become increasingly prominent; although embodied intelligence has driven AI from digital space into the physical world, its ability to learn autonomously, adapt continuously and make flexible decisions in complex and ever-changing real-world scenarios still has room for improvement. By contrast, the capabilities that living systems demonstrate in energy efficiency, learning and environmental adaptation provide important inspiration for exploring new paradigms of intelligence.

As neuromorphic computing draws on neural mechanisms, brain-computer interfaces establish human-machine channels, and embodied intelligence brings machines into physical space. Along these directions, how will biological intelligence collaborate with data intelligence and physical intelligence to support higher-level intelligence?

In plain terms, data intelligence enables AI to "read data," represented by large models that learn patterns and generate content from massive amounts of information; physical intelligence enables AI to "step into reality," with embodied intelligence as an important carrier, perceiving the environment and taking action through sensors and physical devices; biological intelligence enables AI to "learn from life," attempting to draw on the operating mechanisms of biological systems such as the brain to find lower-energy, more efficient ways of computing and interacting.

The capital market provides a real-world footnote to this evolutionary path. Currently, OpenAI is seeking its latest funding round, with a pre-money valuation set at $1.25 trillion; in October this year, the embodied intelligence company Figure and Nvidia are in talks for an additional investment of about $1 billion.

In the field of biological intelligence, early-stage capital has already begun to place its bets. The Australian company Cortical Labs, which focuses on combining lab-grown human brain cells with computer chips, received $10 million in funding in 2023 led by Horizons Ventures; the U.S. company The Biological Computing Co. (TBC), which focuses on applied biological computing, completed a $25 million seed round in 2026; the U.S. company Bios Life, which is advancing biological digital twins, completed a $25 million seed round in 2026, using genomic, clinical and other data to build continuously updated patient digital twins for cancer screening and monitoring.

These financings reflect different stages of the industry. Large models have entered a period of heavy capital and large-scale investment, and embodied intelligence has begun raising funds for mass production and scenario delivery; biological intelligence is still at an early stage of the technology curve, with the industry landscape not yet formed, and capital flows more toward technology validation and engineering.

However, the market growth potential of related sub-sectors has already begun to emerge. The global market research and consulting firm The Business Research Company, in a 2026 report, projects that the global AI neuromorphic computing market will grow from $2.04 billion in 2025 to $9.28 billion in 2030, expanding about 4.5 times over five years.

And neuromorphic computing is only one of the many technology directions within biological intelligence. From biological digital twins, neuromorphic chips and brain-computer interfaces to living-cell computing, organ-on-a-chip and biological agents, biological intelligence is exploring different industrialization paths, and its potential applications extend from low-power intelligent terminals to disease research, drug development and new forms of human-machine interaction.

In a sentence: biological intelligence is still a blue-ocean market at present, but it has high growth elasticity. Once living-cell computing, neuromorphic chips and brain-computer interfaces cross the threshold of product validation, biological intelligence is expected to move from early exploration into an accelerated industrial phase, and to achieve differentiated breakthroughs and phased development potential in some sub-scenarios, evolving in parallel and complementing and integrating with the two tracks of data intelligence and embodied intelligence.

Beyond market space, the energy consumption constraints of traditional computing paths are another reason biological intelligence has attracted attention.

"Why do we say that intelligence driven purely by data is not sustainable in the long term? We can do an energy account for all of humanity," Professor Miao Feng, dean of the School of Frontier Sciences at Nanjing University and dean of the Institute of Biological Intelligence, told 36Kr. "The computing volume of current AI training and inference is exploding exponentially. If we add up the global energy spent on data movement, it would even be equivalent to the generating capacity of several Three Gorges Dams. And in the future, the computing power demand of future large models will jump by several orders of magnitude."

By contrast, looking at the total energy available to humanity, even including wind power, photovoltaics and nuclear power, global energy supply still relies heavily on fossil fuels and overall maintains slow linear growth.

The energy consumption dilemma is also rooted in the "separation of memory and computing" of the von Neumann architecture: frequent data movement between memory and processors generates enormous energy consumption and latency.

In contrast, the human brain does not need to encode the real world into "0" and "1." After millions of years of evolution, the human brain exhibits outstanding and efficient perception, processing, learning and reasoning capabilities with only extremely low power consumption.

In fact, research related to biological intelligence has a long history. Early on, it mostly sought to understand the mechanisms by which human or animal intelligence arises from the levels of DNA, cells and behavior. Zheng Hairong, member of the Standing Committee of the Party Committee and Executive Vice President of Nanjing University and academician of the Chinese Academy of Sciences, has combined long-term research and practice to offer a new interpretation of this concept, incorporating human-machine integration into the scope of biological intelligence research. He regards biological intelligence as a core pillar of superintelligence and has proposed original concepts such as "brain-to-brain communication."

If described along a more easily understood path, biological intelligence is moving from "reading life" to "fusing with life": first studying how the brain and living organisms produce cognition, then drawing on neural mechanisms to develop new technologies such as brain-inspired computing and brain-inspired vision, then intervening in nervous system diseases through means such as brain-computer interfaces, and ultimately exploring deep collaboration among the human brain, digital twin brains and cloud-based large models.

"Today when we open our phones and chat with large models such as DeepSeek, although in essence there is also a combination of human brain thinking and machine knowledge, the interaction efficiency through typing or voice is actually very low." Professor Miao Feng made a vivid analogy.

He mentioned that large language models and the human brain each have their strengths. If in the future, based on more efficient brain-computer interface interfaces, brain-inspired intelligence technologies and even an individual's "digital twin brain," efficient direct connection and deep collaboration among the human brain, twin brain and cloud-based large models can be achieved, allowing "Human-in-the-loop" to fully participate, then human-machine integrated intelligence of "1+1>2" can be formed.

It can be seen that although biological intelligence is still at an early stage of the technology curve, the three curves of demand, technology and capital have already begun to converge.

How it will collaborate and integrate with large models and embodied intelligence still needs product and market validation. But what capital is positioning itself for in advance is precisely this possibility of moving from marginal exploration to a mainstream intelligence architecture.

Industry practice: What barriers must biological intelligence cross before it moves toward industry?

It should be made clear that a blue ocean does not mean technology will naturally grow into an industry.

Biological intelligence still needs to cross three barriers: forming synergy through interdisciplinary integration, moving from laboratory prototypes to stable products, and aligning with real-scenario demand.

English text is being prepared.

Taking Nanjing University as an example, the university's layout in biological intelligence has lasted more than ten years. Six years ago, the team was among the first internationally to break through retinal-like visual sensors and brain-inspired visual architectures; in 2021 it established the Brain-Inspired Intelligence Technology Research Center.

In recent years, Nanjing University has further strengthened systematic layout in related frontier fields. In 2026, under the overall promotion of Academician Zheng Hairong, the Nanjing University Institute of Biological Intelligence was officially established. In this context, brain-inspired intelligence is an important component of biological intelligence and can be regarded as a subset of it. The earlier brain-inspired research laid the foundation for the Institute of Biological Intelligence.

Biological intelligence spans fields such as physics, information, biology and medical engineering. Truly forming scientific research synergy requires not only scientists to break through existing disciplinary boundaries, but also universities to provide support in disciplinary planning, talent introduction and cultivation, and research platform construction.

In this process, concepts advocated by Academician Zheng Hairong, such as strengthening original innovation, encouraging frontier exploration and promoting interdisciplinary integration, have brought together interdisciplinary innovative forces and effectively advanced scientific research.

Specifically regarding the institute's organizational structure, it adopts an interdisciplinary research design, relying on the new engineering approach of "transforming from science to engineering, with implementation in engineering," allowing different disciplines to collaborate around common technical goals and driving biological intelligence research from single-point technological breakthroughs to multidisciplinary collaborative tackling of key problems.

The Nanjing University Institute of Biological Intelligence focuses on four major research directions:

1. Foundations of biological intelligence and large models: analyzing the cross-scale mechanisms of life intelligence and building interpretable, predictable and intervenable life models;

2. Bio-inspired perception and computing: developing bionic perception, sensing-computing integrated devices and neuromorphic computing;

3. Biological computing and organoid intelligence: exploring organoid living computing and wetware-hardware collaboration;

4. Biological agents and embodied systems: building a "perception-decision-action" closed loop and advancing the integration of biological intelligence and silicon-based intelligence.

The four main lines advance layer by layer along "basic mechanisms-key technologies-system integration," forming a research chain from scientific discovery to intelligent entities.

These main research lines have already produced concrete results. The team has applied a fully connected model of the Drosophila brain containing about 100,000 neuronal connections to a drone system, achieving danger avoidance and obstacle avoidance functions in the laboratory without the need for traditional massive data training. In the future, it is expected to give rise to intelligent exoskeleton "mecha" that help paralyzed patients walk, and may also, like "Avatar," allow humans to precisely control physical entities across space with their minds.

The second difficulty is engineering: performance advantages in papers must become products that can be manufactured, reused and enter real systems.

"Guangyu Chip" is a sample of the team's attempt to cross this barrier. To understand its value, one must first understand the difference between traditional machine vision and human vision.

“For example, if I want to determine what brand the phone in your hand is, traditional computer vision processing methods require first digitizing and modeling 100% of all pixels in the entire image, including redundant background information, and then going through extremely energy-intensive data movement and back-end computation to compare features. A large amount of bandwidth and computing power is wasted on processing useless redundant information,” Professor Miao Feng said by way of example.

The human eye, by contrast, relies on retinal “bipolar cells” to pre-process visual signals, and the cerebral cortex uses an attention mechanism to mobilize the “macular region” to focus on core features, completing rapid recognition at extremely low consumption. This prompted researchers to break the traditional path of “image first, process later” and deeply integrate sensing, memory and computing at the front end.

More than a decade ago, Professor Miao Feng’s team began cross-disciplinary exploration aimed at biological intelligence. Two months ago, the “Guangyu Chip 1.0” result was published in Nature Sensors, a top international journal in the sensing field, and Nature Electronics, a top international journal in the electronics field, also published a dedicated article recommending it. After the result was unveiled, industry and capital circles successively approached the team.

The soon-to-be-released “Guangyu Chip 2.0” goes further, shifting toward engineering-oriented architectural innovation. The chip directly realizes “optical signal input, token output” at the sensor end, allowing optical signals to be directly connected to large language models without cumbersome conversion. At present, the chip has completed the engineering route and is carrying out adaptation and industrial incubation for scenarios such as small and micro special robots and commercial aerospace. Professor Miao Feng hopes that Guangyu Chip can become the entry point for large language models to perceive the physical world in the future.

After crossing the engineering threshold, the third challenge is how to coordinate talent and industry so that technology continues to emerge and aligns with real needs.

In response, the institute has adopted the approach of “working backward from future industry needs,” newly establishing the interdisciplinary direction of “Biological Intelligence Science and Technology” and launching global recruitment to gather top leaders and engineering technical talent.

In terms of industrial coordination, Nanjing University has previously signed a comprehensive strategic cooperation agreement with CETC Hikvision Group to promote the early embedding of enterprise scenario understanding and ecosystem capabilities into scientific research. Enterprises raise questions, engineering teams participate in verification, and researchers continue to tackle key problems, thereby giving interdisciplinary research a clearer industrial coordinate.

From disciplinary organization to engineering verification, and then to talent-enterprise coordination, what the institute aims to build is a mechanism that continuously produces technology and continuously calibrates direction. This also sets a starting point for the next stage of industrial transformation.

Industrial formation: how to transform frontier scientific research into future industries?

When biological intelligence changes from a technological proposition to an industrial proposition, competition also extends from papers and prototypes to capital, platforms and scenarios.

From a global perspective, the industrial organization methods of the United States, Europe and China each have different emphases. The United States emphasizes relay between scientific research and capital, Europe values cross-national coordination and public scientific research infrastructure, and China promotes linkage among scientific research, policy, capital and regional scenarios.

First, the United States. The National Institutes of Health’s “BRAIN Initiative” has a fiscal year 2025 budget of US$321 million, continuously supporting brain mapping, neurotechnology tools and interdisciplinary research; on the enterprise side, venture capital is used to push frontier directions such as living neuron computing toward product verification. It is good at using scientific research supply to open up possibilities and then letting capital screen commercialization routes, but frontier results still need to pass through engineering and scale-up screening.

Next, Europe. The EU’s “Human Brain Project” formed the EBRAINS research infrastructure over ten years; in 2024, the European Commission provided another 38 million euros to support the development of EBRAINS 2.0 through 2026. The project connects 59 partner institutions in 16 countries and accumulates brain data, brain maps, simulation tools and neuromorphic computing platforms. A shared foundation both reduces the cost of duplicate research and incorporates data interoperability, ethics and governance into industrial evolution.

Finally, China. At present, China is placing scientific research breakthroughs, industrial policy, capital investment and regional scenarios into the same blueprint.

In 2025, the Ministry of Industry and Information Technology and six other departments issued a document stating: “Realizing the collaborative interaction between biological intelligence and machine intelligence is a frontier technology in the integrated development of life sciences and information sciences.”

The document also deploys key technologies, products and standards systems, proposes building industrial cluster zones and improving pilot testing capabilities, and promotes increased investment from the National Manufacturing Transformation and Upgrading Fund, the National Small and Medium Enterprise Development Fund and others, so that policy goals form a closer combination with engineering platforms, industrial funds and application scenarios.

Compared with the three paths, China’s differentiated advantage lies in relying on a complete manufacturing system and rich application scenarios to rapidly connect source innovation in universities with local industrial organization.

Whether this advantage can be realized depends on whether coordination truly occurs. The joint construction of the Biological Intelligence Institute by Nanjing University and Xuanwu District is precisely a concrete experiment of this path in Nanjing.

It is understood that among the five-year goals proposed by the Xuanwu District Party Congress, building an “all-round intelligent pioneer zone” was placed first. Previously, Xuanwu District had gathered about one-third of the province’s registered large models and had laid out a data exchange and the embodied intelligence industry.

Therefore, relying on local application scenarios such as biomedicine, cross-border trade and green low-carbon development, Xuanwu is accelerating its leap toward biological intelligence, providing an ideal “first-launch test site” for the implementation of new technologies.

In university-local cooperation, both sides emphasize the autonomy of universities’ frontier exploration, the enthusiasm of scientific research teams for transforming results, and the synergy of university-local cooperation.

Regarding this cross-entity coordination, Academician Zheng Hairong advocates promoting the deep integration of innovation elements such as healthcare, scientific research, industry, capital and talent, and solving problems such as scattered resources and poor transformation of results in the process of frontier technologies moving from the laboratory to industry.

This concept also provides a perspective for understanding the university-local cooperation model of the Nanjing University Institute of Biological Intelligence: universities complete source innovation, parks provide space, policies and implementation scenarios, and then professional platforms familiar with medical device engineering transformation and market demand are introduced, so that technological innovation and industrial demand can be more closely connected.

In this collaborative system, the National High-Performance Medical Device Innovation Center (hereinafter referred to as the “National Innovation Center”), approved and established by the Ministry of Industry and Information Technology, participates in co-construction as the only national-level innovation center in the medical device field. Since then, the institute, Xuzhuang High-tech Zone and the National Innovation Center have respectively undertaken technology sourcing, achievement hosting and industrial implementation support.

On the technology sourcing side, the Nanjing University Institute of Biological Intelligence focuses on frontier science and advances original breakthroughs from 0 to 1.

On the achievement hosting side, Xuzhuang High-tech Zone in Xuanwu District plans to provide industrial space at Jiangsu International Data Port, along with supporting special policies and upstream and downstream application scenarios.

On the industrial implementation support side, the National Innovation Center relies on a national-level platform to provide full-chain industrial services and bring in industrial ecosystem resources, covering links from inspection and testing, registration application and clinical validation to investment and incubation.

Capital and incubation mechanisms have also been moved forward to the starting point of achievement transformation. The institute is simultaneously advancing the construction of the Biological Intelligence Industry Fund, the Innovation Workshop (Nanbrain Intelligence Incubator) and the Biological Intelligence Science and Technology Innovation Alliance, introducing technology managers and professional engineering teams, and working with leading enterprises, innovative companies and investment institutions to jointly build think tanks and roadmaps, helping scientific research talent cross the gap from papers to products.

This transformation mechanism will also be implemented in a concentrated manner at the 2026 International Biological Intelligence Innovation Conference: the three entities of the institute, the Biological Intelligence Industry Fund and the Innovation Workshop will be unveiled on the same stage, and two major projects, “Guangyu Chip 2.0” and organ-on-a-chip, will be released, enabling industry, scientific research and capital to connect on the same platform.

Beyond the coordination mechanism, Professor Miao Feng also pointed out that many frontier biological intelligence technologies, after proof of principle, have demonstrated clear advantages, but to move toward large-scale commercialization, they still must pass the “scenario validation” threshold oriented toward real industrial needs.

This threshold cannot be breached by a single entity: researchers are good at tackling key problems but may not be familiar with the market, enterprises have scenarios but find it difficult to bear the risk of long-term frontier R&D, and investment institutions need to judge whether the technology curve matches the business cycle. Government-industry-academia-research-medical-capital coordination aims to build a full-chain incubation system before the industry takes shape.

“Through continuous R&D investment and ‘laying eggs along the way’ staged transformation of results, we are expected to drive the biological intelligence industry to usher in an important explosion within 3 to 5 years,” Miao Feng said.

To achieve the above goals, the institute plans to advance engineering and scenario validation around four major research directions, incubate related enterprises and attract upstream and downstream enterprises to gather, while using Nanjing as a base to connect with leading enterprises, startups and industrial capital across the country and around the world, exploring a path for the deep integration of technological innovation and industrial innovation.

This route from digital space and the physical world to living systems therefore also has more specific industrial coordinates.