Editor's Note
When AI has fallen into the bottleneck of "stacking computing power and consuming huge amounts of energy," brain-inspired intelligence is opening a new path. Recently, Academician Zhang Xu, President of the Guangdong Institute of Intelligent Science and Technology, wrote in China Science Daily that brain-inspired intelligence, by simulating biological neural networks, can achieve localized, private, personalized training with lower power consumption, making intelligent agents "more and more like you." The article systematically explains the breakthrough path from following to leading, and describes the team's important breakthroughs in fields such as Intuitive Neural Networks and brain-inspired chips. The full text is reprinted below:
Brain-Inspired Intelligence Is an Important Breakthrough Direction for Next-Generation AI
Brain-inspired intelligence (BI), as an important breakthrough direction for next-generation AI, is undergoing comprehensive innovation from algorithms and chips to computing systems, and is expected to break free from long-term dependence on traditional AI computing power while solving data privacy challenges.
A New Opportunity from "Following" to "Leading"
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 and working mechanisms of human brain perceptual and cognitive neural networks, and engineering intelligent frameworks, algorithmic models, and intelligent computing systems inspired by brain information-processing mechanisms.
Since the 1950s and 1960s, AI has absorbed basic discoveries from brain science, but later it was mainly driven by internet big data, mathematics and statistics, and computer technology. Today, it has reached a technical bottleneck. Its excessive dependence on mainstream AI computing power represented by GUP and massive data has led to high energy consumption, high costs, and difficulty in popularization.
Brain-inspired intelligence, however, takes another path. By simulating biological brain neurons and their neural-network computing system, it can, compared with traditional AI computing-power solutions, greatly reduce hardware scale, energy demand, usage cost, and computing-power construction cost. It is especially suitable for the broad range of developers and users of AI application technologies, as well as related enterprises, research institutions, hospitals, and others, and can serve as the computing-power foundation for intelligent agents, intelligent robots, and embodied intelligence.
However, the greatest challenge in the current development of brain-inspired intelligence does not come simply from today's mainstream AI computing power, but from the deep interdisciplinary crossing and paradigm coordination of neuroscience, brain-computer interfaces, mathematics, computer science, chip manufacturing, and other disciplines. For brain-inspired intelligence to truly break through, it must connect the full chain from brain cognitive mechanisms to algorithmic models, and then to chips and systems.
At the same time, brain-inspired intelligence needs to expand in a differentiated way on the basis of the existing AI technology system. Through crowd-creation linkage and resource integration, it should work with developers, investors, and consumers to jointly build an industrial ecosystem and development ecosystem.
By seizing this round of technological transformation opportunities, China is expected to gain an advantage in several areas of next-generation artificial intelligence and achieve a leap from "following" to "leading."
"More and More Like You"
Unlike current large models, which pursue public and standardized capabilities, brain-inspired intelligence supports localized and private training. Personal health and family data can be completely retained locally, both protecting privacy and security and achieving a high degree of personalization.
Our team has already achieved a number of original advances. For example, the Intuitive Neural Network (INN) developed by the team differs from traditional artificial neural networks (ANN) and spiking neural networks (SNN), and can better simulate the perception-cognition integration mechanism of layer 5 neurons in the brain's sensorimotor cortex. A brain-inspired computing system based on this model, with only 6 kilowatts of power consumption, can support computation across 1,153 X86 CPU cores, reaching a training speed of 360,000 tokens per second, an inference speed of 1.35 million tokens per second, and latency below 100 milliseconds.
At the chip level, although the team's first-generation brain-inspired wafer chip uses a 40-nanometer process, when processing unstructured data it has an 18-fold energy-consumption advantage compared with an A100 GPU using a 7-nanometer process, and it already supports computation with 4.6 million spiking neurons. The world's first 10-billion-neuron brain-inspired heterogeneous integrated supercomputing system has also been built, requiring only 5 cabinets and 30 servers.
The performance advantages of brain-inspired intelligence may completely change business models: enterprises, laboratories, and even households may all be able to independently complete multimodal large-model training without worrying about data leakage. This is an important step in the liberation of human thought and personalized capabilities.
Therefore, regarding questions that many people care about, such as "whether brain-inspired intelligence will replace human emotions," my view is that a brain-inspired computer is "raised" in your home. It will become more and more like you, not like someone else.
Financing Has Exceeded RMB 3 Billion, and Scenario Applications Are the Key
The greatest characteristic of intelligent technology is that without scenarios there is no application, and without application there is no market.
However, the recently released blue book China Future Industry Science and Technology Innovation Development Report (2026) shows that the total financing in China's brain-inspired intelligence field has exceeded RMB 3 billion, of which the brain-computer interface track accounts for as much as 43%. Shanghai has already formed industrial clusters for brain-inspired computing and brain-computer interfaces respectively. More segmented industrial layouts are expected to appear in 2026.
Brain-inspired intelligence has already shown application potential in fields such as healthcare, industrial simulation, and edge computing, but government and society need to jointly promote scenario development.
In terms of technological integration, cross-innovation among brain science, AI, synthetic biology, and materials science will become the core driving force.
In terms of chip manufacturing and technology integration, it is recommended to build open and shared national-level infrastructure, give priority to developing edge application scenarios, create typical demonstration cases, and at the same time accelerate talent cultivation and standards-system construction.
At present, China already has the basic conditions for "changing lanes to overtake" in the field of brain-inspired intelligence. The key lies in grasping the opportunities of the era, adhering to both independent innovation and open cooperation, and promoting the transformation of technological achievements into industrial practices that benefit the people. This is not only required by scientific and technological competition, but also a strategic choice for serving national high-quality development and satisfying the people's aspirations for a better life.
(Organized by Zhang Nan, reporter for China Science Daily)
Full text reprinted from China Science Daily (June 4, 2026, Page 1, Important News)
