Gao Xiaorong, a professor at Tsinghua University: In the race of brain-computer interface, China and the United States are "running side by side" on different paths.
In the Silicon Valley narrative, the field of brain-computer interfaces is filled with numerous imaginable visions. The brain is no longer just an anatomical diagram in medical textbooks but the next frontier to be transformed by technology. People are being told about a future where the brain and machines are deeply integrated, and memory and consciousness will break through biological limitations.
At the intersection of reality and imagination, brain-computer interface companies such as Neuralink and Synchron are continuously advancing clinical trials of brain-computer interfaces. Neuralink has completed craniotomy implantation in a small number of patients, focusing on restoring motor and language functions lost due to paralysis or neurological diseases, and plans to conduct speech cortex experiments in the fourth quarter of 2025 to decode "intended language." Synchron, on the other hand, enables paralyzed patients to control external devices with brain signals through minimally invasive intravascular implantation. The safety and partial restoration of daily functions have been verified in clinical trials. These achievements are gradually demonstrating more potential of brain-computer interface technology.
Facing the rapid development of technology, the public's attention is drawn between two distinct visions. On the one hand, brain-computer interfaces bring hope for recovery to those who have lost partial functions for various reasons. On the other hand, tech moguls like Elon Musk have sparked the public's infinite imagination of a better future through frequent marketing and promotion, such as uploading consciousness and "digitizing the self" in science fiction futures, but have also raised public concerns. In this atmosphere, as these technologies continue to progress, the boundary between science and science fiction is becoming increasingly blurred.
Currently, how is the actual implementation ability of brain-computer interface technology progressing? Will the scenario of a human-machine war in the movie "The Matrix" be reproduced in the future world? When artificial intelligence becomes the new "intelligent agent" in society, what role will brain-computer interfaces play? Recently, we had an in-depth exchange with Gao Xiaorong, a tenured professor at the School of Medicine of Tsinghua University, focusing on the actual ability boundaries of this technology.
Gao Xiaorong provided a more sober perspective: Brain-computer interfaces are not a shortcut to "superhumanization" but a technological exploration centered on repair and assistance under ethical constraints. "Elon Musk's initial vision tended towards'superhumanization,' which is ethically unacceptable," he pointed out. Concepts such as uploading consciousness and directly reading brain information through brain chips are still unrealistic fantasies at present.
Gao Xiaorong, Tenured Professor and Doctoral Supervisor at the School of Medicine of Tsinghua University
The following is a transcript of the conversation between Tencent Technology and Professor Gao Xiaorong (with some deletions without changing the original meaning):
Tencent Technology: In recent years, the public has had high expectations and even some fantasies about brain-computer interface technology. Especially with Elon Musk's frequent external promotion of the potential of brain-computer interfaces and his company Neuralink, concepts such as uploading consciousness and mind control in the future have further stimulated the social imagination. In your opinion, where are the real boundaries of brain-computer interfaces? Can things like uploading consciousness really be achieved?
Gao Xiaorong: In terms of principles and goals, brain-computer interfaces can be roughly divided into two categories: one is repair or replacement, that is, when a person's ability is below "100%", technology helps restore it to the normal level; the other is enhancement, which attempts to break through "100%" and enable humans to have abilities beyond nature. Using a joking metaphor, it's like "turning people into supermen."
The so-called "superhumanization" is ethically unacceptable. If there were really technologies that could give humans extraordinary abilities, then "who can use them and who is qualified to use them" would bring about huge unfairness. Elon Musk's initial vision tended towards "superhumanization," but under ethical constraints, he has now turned to clinical applications.
On the other hand, there is also a goal of human-machine collaboration. With the development of AI, in the future society, there will be both human intelligent agents and artificial intelligent agents. How the two communicate is a fundamental problem. If the communication gap cannot be resolved, society may become divided, and brain-computer interfaces are a potential means to fill this gap. It can achieve "making things happen as one wishes," such as directly transmitting a person's intentions to machines.
In this sense, brain-computer interfaces also contain elements of "transcending human natural limits," but this is necessary transcendence - it is not to create supermen but to ensure the coexistence of humans and machines. As for "uploading consciousness" or "digital immortality," there is absolutely no possibility of achieving them at present.
In the context of such human-machine coexistence, we must rethink how to maintain the "human-centered" nature of society. Some people propose to make machines completely human-like, but this is fundamentally impossible. Because a human life is unique, while machines can be turned on and off at any time - they can never truly understand the uniqueness of human existence, especially the ultimate experience of "death." And many human problems are precisely related to death.
Tencent Technology: Putting aside ethical considerations, from the perspective of neuroscience principles and technology, is it possible to achieve functions such as uploading consciousness and mind control through brain-computer interfaces?
Gao Xiaorong: I don't think it's possible now. Neuroscience principles also do not support achieving these.
Tencent Technology: Elon Musk also mentioned that brain-computer interfaces can treat schizophrenia or autism. Can this be achieved?
Gao Xiaorong: This is possible. The brain has a "reward center," and electrical stimulation can produce hallucinations. If the degree of hallucinations can be precisely controlled, the purpose of treatment can be achieved. In the past, we mainly relied on drugs for control, and now electrical stimulation provides new possibilities. Cochlear implants and artificial retinas are mature examples. From a higher level, this is scientifically feasible, but it must strictly follow ethical and regulatory requirements.
Tencent Technology: You also mentioned just now that brain-computer interfaces can serve as a medium between humans and machines and even enhance certain human abilities, but this enhancement has limits. So, to what extent can it be achieved in the future?
Gao Xiaorong: Some limited "plug-in" abilities are feasible. For example, memory enhancement: Alzheimer's patients often cannot immediately recognize acquaintances when they see them. If a brain-computer interface can detect brain activity and prompt "who this is" and "when and where they met," communication will be much smoother.
A similar scenario also applies to healthy people. For example, when someone sees an acquaintance but can't remember their name for a moment, the system can provide an immediate reminder. This kind of enhancement can be understood as "plug-ins": memory plug-ins, computing plug-ins, cognitive plug-ins, and behavior control plug-ins. They won't turn humans into "supermen," but they can improve efficiency and experience in specific aspects.
Tencent Technology: Based on your research experience, which directions are more reliable in current clinical experiments?
Gao Xiaorong: It's probably several directions related to the application of brain-computer interfaces. First, motor compensation and rehabilitation, such as controlling prosthetics and hand movements. This area has been done quite well. Elon Musk's implantation, the research of Professor Hong Bo in China, the "North Brain No. 1," and the research of the Shanghai team are all in this area. Second is the aspect of speech function, such as helping people who can't speak to regain language ability. This is actually also motor enhancement, except that it enhances the tongue. Later, it may be applied to scenarios such as cognition, emotion, and cultural entertainment.
From a technical perspective, all aspects such as signal reading, information decoding, and stimulation feedback are progressing. Implanted devices can be stably used for one to two years, but there are still challenges in breaking through three years or more. If the electrodes are placed outside the cerebral dura mater, their lifespan is longer; if they are placed on the scalp, they can be used at any time, but the signal quality is weaker.
Tencent Technology: Is the biggest challenge now at the hardware level or the software signal processing level?
Gao Xiaorong: There are challenges in both. The progress in the past 30 years has mainly relied on materials science - electrodes are getting smaller and softer, just like CPUs that are constantly shrinking. But now, there is a new bottleneck: the number of recording channels has increased from hundreds to thousands and may reach tens of thousands in the future. How to analyze such a large amount of data? This requires artificial intelligence.
The progress of AI is much faster than that of materials science. The capabilities of models may double within a hundred days, while the improvement of materials usually only has a one - percent increment per year. Therefore, in the future, the driving force for breakthroughs in brain-computer interfaces will depend more on AI rather than just materials.
Tencent Technology: In the technical path of brain-computer interfaces, there are invasive, semi-invasive, and non-invasive types. Which method do you think is more promising?
Gao Xiaorong: All three paths are being explored, and it's hard to say which is better or worse at present. I once proposed a metaphor called the "Brain-Computer Starlink." Twenty years ago, it was almost impossible for satellites to communicate with mobile phones. Now, with the development of satellite launch technology, the number has increased from one to tens of thousands, and the problem of communication coverage has been solved. Similarly, in the future, if there are enough "nodes" in brain-computer interfaces, it may be like the Starlink, achieving continuous connection.
Tencent Technology: Judging from people's feedback, there are more concerns about the invasive method. How do you view the risks here?
Gao Xiaorong: The concerns are reasonable. Non-invasive devices can be removed at any time, and then brain signals are no longer read; while for invasive devices, once implanted, users cannot confirm whether they are running continuously and are more likely to face the risk of being attacked.
Tencent Technology: Many people associate brain-computer interfaces with robots. Compared with traditional control methods, what are the improvements?
Gao Xiaorong: The difference is very large. Traditional control can only make robots perform simple actions, such as "moving forward, backward, and turning." But it doesn't know your real intention. Brain-computer interfaces can directly transmit the purpose. For example, saying "I want to eat an apple," the robot will independently plan the path and complete the task, rather than waiting for step-by-step instructions. This difference is like navigation. In the past, we needed to manually tell it each step, but now we only need to input the destination.
Tencent Technology: Will there be differences in terms of latency or battery life?
Gao Xiaorong: This is a technical issue that can be solved with iterations.
Tencent Technology: Currently, the applications of brain-computer chips are mainly concentrated in the medical field. What other application scenarios might they expand to in the future?
Gao Xiaorong: I'm particularly optimistic about the elderly care field. For example, when the elderly encounter something they don't know how to operate, they can directly put forward their needs, and the robot can assist them in completing the task. In addition, it also includes sports rehabilitation, cognitive rehabilitation, emotional support, and cultural entertainment. It should be emphasized that when entering the fields of emotion and entertainment, the invasive method is hardly used, and more reliance is placed on non-invasive or wearable solutions.
Tencent Technology: As the global leaders in brain-computer interface research, what are the main differences in the research progress of China and the United States in this field?
Gao Xiaorong: Generally speaking, they are "running side by side." China is more advanced in non-invasive and semi-invasive research, while the United States is stronger in invasive research.
Tencent Technology: How many years will it take for the technology to mature approximately?
Gao Xiaorong: I initially thought it would take 60 years, and then changed it to 30 years. Now it seems that the time may be further shortened, and more mature results are expected to appear within about 15 to 20 years. However, it must be admitted that the current technology still has obvious limitations. Take implanted devices as an example. Their lifespan is still measured in "years," and the longest case has only lasted for a few years. Elon Musk once claimed that the device could work stably for a year, but in fact, the electrodes partially fall off after three months. This also shows that we still have a long way to go to achieve "lifetime implantation."
This article is from the WeChat official account "Tencent Technology," written by Helen, and published by 36Kr with authorization.