On January 30, 2024, Elon Musk, a business tycoon and CEO of SpaceX and Tesla, announced on his social media account that "Neural Connections was implanted in the human body for the first time yesterday, and the surgeon is recovering well", adding that the first product of the company he founded called Neuralink will be named Telepathy. This news itself is enough to attract attention, coupled with Musk's personal influence, this news in the field of brain-computer interface was immediately reprinted by major media.
On March 21, Neural Connections revealed the identity of the implant in the form of a live video. Noland Arbaugh was paralyzed eight years ago from a cervical spine injury caused by a diving accident. In a short video showcase, Abo uses his mind ("want") to control his computer's mouse to play chess, and says that he played the large-scale turn-based strategy game Civilization 6 in this way. The video and the topic of brain-computer interfaces have once again attracted a lot of attention and discussion, and the video has been viewed nearly 100 million times on Neural Connect's own social media alone.
On the 23rd, Abo posted a video on his social media, which was filmed at the headquarters of Neural Connection, which was a sharing meeting with Neural Connections employees after Abo was invited to visit the headquarters of Neural Connection. During the sharing session, the host showed a video of Abo playing the game Mario Kart with his mind. If the chess in the video on the 21st can still be "played slowly", Mario Kart obviously cannot be "played slowly", and this game requires a quick reaction from the player. To play the game well, Abo's brain-computer interface system needs to be very fast and accurate in its response to commands. Judging by the video, Abo was able to play the game very smoothly. For this reason alone, it is not an exaggeration to say that the performance of the "neural connection" implant is "excellent".
Screenshot of Nolan Arbo's video sharing session at the "Neural Connection" headquarters, after the N1 implant is implanted, he can already play games such as Mario Kart, which require high reaction speed and accuracy. (Data map)
Because of this, the media coverage of "neural connection" and the public's reaction to it have been highly evaluated, and some media even used news headlines such as "the world's first human transplant of a brain-computer interface device", "Musk announced: mankind's first brain-computer interface chip implantation has been completed", "Musk's sci-fi empire 'next city', mankind's first brain-computer interface chip implantation" news headlines, giving people a feeling that the new progress of "neural connection" has ushered in a major breakthrough in the field of brain-computer interface.
However, in fact, in the field of brain-computer interface, whether it is technology expansion or clinical application exploration, Musk and his "neural connection" are neither a pioneer nor a leader.
What is a brain-computer interface?
Each and every one of us interacts with the world all the time. The realization of these interactions can be broadly divided into three stages. First, perceive the external environment. This perception can come from a variety of senses, such as sight, hearing, smell, taste, touch, etc., and the sensory information is transmitted to our brain through the relevant nerves. Second, information processing, which generates subjective feelings and instructions. Incoming sensory information is processed, processed, and integrated in the relevant brain areas of the brain, resulting in feelings that carry our subjective consciousness: pleasant music, delicious food, wonderful paintings...... If needed, our brains also produce instructions that direct our bodies to respond, such as, "Reach out, grab that piece of cake, and feed it to your mouth." Third, the instruction is sent out and executed. Neural signals from the brain reach their counterparts in the body and are executed. In this case, when you pick up a cake and feed it to your mouth, nerve signals travel through your spinal cord to your hands, where you can control the contraction of the muscles there to complete the instructions and enjoy the meal.
Regardless of which of the three goes wrong, our interaction with the world around us can be hindered. Needless to say, the brain is the commander-in-chief of the organism. Impaired sensory input can affect our perception of the world: people with color blindness perceive the colorful world differently from others because of abnormalities in the photoreceptor cells on the retina. The same is true of the information coming out, as is the case of Abo, who was powerless even if he wanted to play chess or play Mario Kart before the implant of the "neural connection".
Imagine if a person has a problem with the physiological function of one of the above-mentioned links, is it possible to use an artificial device to perform the corresponding function on his behalf? If it is blindness caused by retinopathy, an artificial "retina" is implanted to convert the received light signal into a correct electrical signal and transmit it to the brain; if there is a problem in a certain area of the brain, a "brain chip" is implanted In the case of total paralysis, some kind of device is used to transmit the brain's motor instructions to the receiver and executor, which can be a robotic arm, a mouse, or even the paralyzed person's own hand.
Broadly speaking, these are brain-computer interfaces.
What does "Neural Connection" want to do?
In contrast, the last of the three types of brain-computer interfaces mentioned above is the least difficult to develop. The wide variety of environmental stimuli from the outside world means that the neural signals coming to the brain can be quite complex. Smell is a prime example of this: humans can perceive and distinguish 1 trillion scents, and it is not easy to design and develop brain-computer interfaces that are up to the task. Brain-computer interfaces that replace a brain region are also challenging, given the limited understanding of the complex neural activity in the brain, and the astonishingly complex nature of neurons and firing patterns in even a small brain region.
The brain-computer interface for the last link is much simpler: these outgoing signals are almost all signals that control muscles, and the characteristics are relatively simple. Because of this, both basic research and industrial exploration have focused considerable resources and energy on this kind of brain-computer interface.
Neural Connections was co-founded in 2016 by Musk with seven scientists and engineers, and Musk himself invested $100 million as start-up capital. The company's positioning is to develop brain-computer interfaces that can be implanted in the brain, with the short-term goal of treating serious neurological diseases, and the ultimate goal is to enhance all aspects of human abilities (such as memory, perception, etc.) through these devices. In other words, the ultimate goal of "neural connection" is to "integrate man and machine", and through "artificial intelligence that coexists with people" (Musk's words), the disabled can be restored to their lost abilities, and even let people have abilities beyond human beings.
Since its inception, Neural Connections itself claims that they have tested hundreds of implant implants on animals such as monkeys, pigs, sheep, etc., with nearly 300 in 2022 alone. In addition, Neural Connections and Musk have more than once shown videos of monkeys implanted in their devices playing computer table tennis games or "typing" with their minds. But in fact, the monkey in the video is not consciously "typing" ("Can you give me something to eat?"): the system simply uses a different position of the highlighted flash to indicate the monkey, and the monkey moves the mouse to the position of the flash (the position on the keyboard corresponding to the position of each letter of a word) through the mind, and then the system completes the input. At this point, Musk is obviously very marketing-minded, after all, the gimmick of monkeys typing with their minds is more eye-catching. This is also the case, with many media reports headlining stories about monkeys playing games and "typing".
Despite attracting a lot of attention in the media and the general public, Neural Connections has remained tight-lipped about the specifics of its experiments and hasn't revealed much. In fact, until today, in the field of brain-computer interfaces, there is only one article published in scientific journals with the signature Musk and "Neural Connection". Musk and Neural Connections called the October 2019 article in the Journal of Medical Internet Research a "white paper," and it provides a conceptual introduction to the brain-computer interface technology of Neural Connections.
The idea is alternative
Neural Connections first introduced its brain-computer interface concept to the public in July 2019. According to the company at the time, their technology tried to find a different way, both in terms of implants and the way of implant surgery. The form of the implant is different from the sheet form of the traditional technique, but is filamentous. These "nerves" are made of a flexible, polymeric material, each measuring only 4-6 microns (1 micron = 0.001 mm), thinner than a human hair, and can reduce damage to brain tissue when implanted in the brain. According to the white paper published in the Journal of Medical Internet Research, which was published online in July 2019, there are 32 electrodes on each nerve line, which can read information about the electrical activity of nerve cells.
In order to implant these nerves into the brain, Neural Connections has developed a neurosurgical autonomous surgical robot called the R1 robot. Both in terms of appearance and the way it works, the R1 robot is a lot like a sewing machine. The robot is equipped with multiple cameras and sensors that can quickly and accurately implant nerves into the brain with a "sewing needle". According to Neural Connect's own introduction, the robot is capable of implanting 6 nerves into the brain every minute. If necessary, the robot can also be controlled by a human, trying to avoid the microvessels on the surface of the cerebral cortex during surgery. The white paper also says that the entire brain-computer interface system of "neural connection" can contain 96 neural wires with a total of 3,072 electrodes. Using the signals of cranial nerve activity read by these electrodes, Neural Connections hopes that the person implanted in this brain-computer interface device will be able to interact with the outside world with their minds.
Neural Connections was originally expected to begin trials of the system in humans in 2020, but plans were postponed again and again until May 2023, when the trial was approved by the U.S. Food and Drug Administration (FDA) (the 2022 application was rejected by the FDA on safety grounds).
Musk's personal influence, the ongoing marketing campaign for Neural Connect, the long wait for human trials, and the potential for this system to come together, it's no surprise that Nolan Arbo has attracted a lot of attention and discussion as the first human subject of the Neural Connectivity brain-computer interface system.
What does "neural connection" do?
On the specifics of Nolan Arbo's trial, "neural connections" still reveal very little information. Because of this, industry experts are generally positive but cautious about the evaluation of this trial.
According to the "Neural Connections" webpage (limited information on the webpage is not available until April 12, 2024) and the white paper previously published in the Journal of Medical Internet Research, the implant used in the trial (known as the N1 implant) consists of a number of components, including wireless charging coils, batteries, signal processing devices, neural access elements, etc. The elements are assembled together in layers of nesting, like matryoshka dolls. The entire unit is less than 23 mm× 18.5 mm × 2 mm. This figure comes from the white paper's description of a device containing 96 nerves with a total of 3,072 electrodes. However, it is worth noting that the description of the device implanted in Nolan Arbo's brain on the "Neural Connections" webpage is "64 nerves with a total of 1,024 electrodes". On the one hand, the number of nerves is less than what is described in the white paper. On the other hand, the number of electrodes per nerve is only half the number described in the white paper (16 vs 32). It is unclear whether this difference is due to the caution of the first human trial, or for other reasons.
The implant procedure was done at the Barrow Neurological Institute in Phoenix, Arizona. According to Wikipedia, this is the world's largest institution for the research and treatment of neurological disorders. As Musk claimed on social media, the surgery went well, and Abo was discharged the next day.
Prior to the start of the surgery, the investigator used functional magnetic resonance imaging (fMRI) to select the location for the N1 implant. Functional magnetic resonance imaging (fMRI) is an imaging method that detects brain activity, and which area of the brain has active brain activity will "light up" in fMRI. By asking Arbo to "imagine" moving his hands and arms, the researchers selected an area of the precentral gyrus of the brain. The precentral gyrus is the location of the primary motor cortex in the brain, and the instructions that direct our activities are derived from the brain activity in the corresponding areas of the primary motor cortex, the hand has the area that directs the hand, and the foot has the area that directs the foot. One of the interesting things about the nervous system is that even if you don't actually do a certain action, but just imagine it, the corresponding area in the motor cortex will produce the corresponding neural activity (so Arbo's imagination can also "light up" this area). This means that for Arbo, simply imagining the movement of the hand also triggers the corresponding area in the motor cortex to generate nerve signals that direct the hand to move.
The procedure is not much different from conventional brain surgery, in which the surgeon makes a hole in the skull corresponding to the brain area where it is implanted, and then the implantation is done by the R1 robot. Not many details have been released about "Neural Connections," but according to an article published in Bloomberg Businessweek by veteran technology journalist Ashlee Vance in November 2023, preoperative preparation and craniotomy can take several hours, while the R1 robot implants the N1 implant in a very short time, taking less than half an hour. Ashley Vance is a long-time journalist covering Musk and his company, having written more than one book related to Musk, including a biography of Musk. In addition, judging from this article in Bloomberg Businessweek, "Neural Connection" provides Vance with considerable information knowing, so Vance's information is still more credible even though "Neural Connection" does not disclose more details.
N1 brain-computer interface implants. (Data map)
After the N1 implant was implanted, Abo did not immediately gain the ability to interact with the outside world through his mind, and he still needed a training process. According to the moderator (who is also a participant in the trial) at the sharing session at the headquarters of "Neural Connections", Abo had a total of 12 sessions with the brain-computer interface in about a month after the N1 implant was implanted. Each session consists of many trials. In each attempt, Abo imagines moving the mouse on the computer screen to a designated position and clicking the mouse. The N1 implant collects nerve signals from this process and feeds back the accuracy of Abo's ideas. Through these trainings, Abo can become more and more accurate and control his "hand of mind" more and more freely. As the moderator explained, compared to the training process of other previous studies, Abo's training intensity is amazing: 8 hours a day, 5 days a week. The 12 sets of sessions consisted of a total of 89,285 attempts. Abo clicked the left mouse button a total of 111315 times and the right mouse button 35,045 times with his mind.
The rewards of training: Today's Abo can spend an entire night playing Civilization 6, nearly eight hours at a time.
Controversy continues
Despite Abo's exciting post-operative performance, the "neural connection" and its brain-computer interface have been controversial from its inception to today's human trials.
In October 2022, according to a Reuters report, "Neural Connection" was investigated by a U.S. federal agency for violating ethical regulations regarding the welfare of laboratory animals. Internal employees of Neural Connections complained that animal testing of Neural Connections was too hasty, causing unnecessary suffering and death to experimental animals. The Reuters investigation found that at least four experiments involved avoidable human error, involving a total of 86 pigs and two monkeys. According to information revealed by insiders at Neural Connection, some of the errors appear to be quite low-level and amateurish, for example, in a 2021 study, 25 experimental pigs were implanted with the wrong size implants. In the other two surgeries, the implant was misplaced and implanted into the wrong vertebrae of two pigs. Although a preliminary July 2023 investigation into another allegation by the USDA (the regulator of these animal experiments) concluded that the "neural connections" did not violate the rules, while the investigators said that they would continue to investigate further, the USDA itself has been widely criticized for its lax regulation of animal testing in the past few years. In fact, in an independent investigation by the FDA in June 2023, the FDA found that the animal experiment of "neural connection" violated basic experimental requirements such as incomplete experimental records and lack of quality control.
In September 2023, Wired magazine published an investigative report. Citing information from various sources, the report pointed out that after the implant of the "neural connection", more than a dozen monkeys were in poor health, and the conditions involved included cerebral hematoma, bloody stool, and partial paralysis. There were even implants that broke in the monkeys' brains, resulting in infections that could not be completely eliminated, and eventually the monkeys had to be euthanized. These situations are very different from Musk's claim that no experimental monkeys died from implanted experiments. In November 2023, some members of the U.S. legal profession even asked the U.S. Securities and Exchange Commission to investigate whether Neural Connections had concealed details of the death of experimental animals in order to deceive investors.
Even Nolan Arbo's implant surgery, which seems to have been successful so far, has been criticized and controversial. In the field of medicine, it is a common practice for clinical trials for the trial party to upload information about the trial to the National Institutes of Health's clinical trial information website ClinicalTrials.gov. Neural Connections did not upload this information. However, according to the interpretation of experts in the relevant field, the FDA authorized "neural connection" to carry out early feasibility trials, and the purpose of such trials is proof of concept, not to prove its safety and efficacy, so it is not mandatory to upload detailed information to the ClinicalTrials.gov. Despite this, there are still many experts in the field who believe that the opaque method of "neural connection" is not advisable. In addition, the academic evaluation of a trial is usually based on a paper published in a professional journal, and many experts in the industry are generally cautious about the evaluation of this trial when the paper has not yet been published, and often express a "wait and see" attitude when interviewed by the media.
First mover or leader?
In terms of the innovation of "neural connection" brain-computer interfaces, the evaluation of experts in the field is not as high as that of the public and the media. In fact, in the field of brain-computer interfaces, Musk and "neural connections" are neither pioneers nor leaders.
The exploration of brain-computer interfaces began in the 70s of the 20th century. The concept was developed in 1973 by Jacques Vidal, a professor at the University of California, Los Angeles. Later, in 1977, Vidal experimentally demonstrated the possibility of a brain-computer interface for the first time. In this study, changes in the experimenter's brain waves, generated by visual stimuli, were able to control the movement of a mouse-like object on a computer screen.
Research is also being carried out on the relationship between the activity of neurons in the motor cortex and movement. Through a series of studies on macaques in the 80s of the 20th century, researchers at Johns Hopkins University have clarified the relationship between the two: many neurons in the motor cortex have a preference for the direction of movement, and when the arm or imaginary arm moves in this preferred direction, the corresponding neurons have the strongest neural activity. These insights then became the principles of algorithms used for motion decoding in brain-computer interface systems.
Later, with the advancement of neural electrodes and computer technology, the field of brain-computer interface ushered in rapid development. One of the most well-known is the work of Brazilian neuroscientist Miguel Nicolellis. In addition to a series of studies using monkeys to control robotic arms, Nicolellis drew the world's attention to brain-computer interfaces at the opening ceremony of the 2014 World Cup in Brazil. With the help of a brain-computer interface and a mechanical exoskeleton developed by the Nicolellis team, a paralyzed young man kicked off the tournament and kicked off a football feast.
Chinese researchers have also done excellent work in the past few years. In 2020, researchers at Zhejiang University completed the first clinical study of an implantable brain-computer interface in China. Patients who receive implantation can control the robotic arm to eat and drink, thus achieving a "zero breakthrough" in the clinical practice of implantable brain-computer interface in China.
From this history, it is not difficult to see that "neural connection" is not a pioneer in the field of brain-computer interfaces. In fact, even the "sewing machine" implantation technology of the R1 surgical robot, which "Neural Connection" is proud of, was not developed after the establishment of "Neural Connection". The concept and technology originated from the team of Philip Sabes, professor emeritus at UCLA, but Sabes was invited to co-founder to found "Neural Connection", and this technology was brought to "Neural Connection" by Childis (Sabeth has now left "Neural Connection", in fact, six of the other seven co-founders in addition to Musk have left "Neural Connection").
In terms of human implant trials that try to break through the limitations of the laboratory, even if the concept of the 2014 World Cup opening ceremony is not counted, the "neural connection" trial is not the first in the world. In May 2023, researchers at the Swiss Federal Institute of Technology in Lausanne published a paper in the journal Nature describing their technique of using a brain-computer interface to allow a paralyzed person with lower limbs to use his mind to control his or her feet to walk (with the assistance of assistive devices). The paralyzed person who received the implant was already able to walk outdoors without much help from others.
In terms of exploration that is relatively confined to the laboratory, there are even more amazing results even earlier. In May 2016, published in the journal Nature, a study by American scientists allowed Ian Burkhart, a quadriplegic man, to use his mind to control his hands for everyday activities such as picking up objects, pouring water and even playing the guitar.
These improvements may not be impressive compared to the all-night games that Abo can play today, but it is necessary to recognize that it is much more difficult to control limbs than to control a mouse, that moving a mouse is unlikely to solve all aspects of a paralyzed person's daily life, and that it remains to be seen whether the technology of "neural connection" can be translated into the technology of controlling limb movements.
Two other companies
In addition, "neural connection" is not the only company that has entered the brain-computer interface industry. Among these companies, Blackrock Neurotech and Synchron are two companies in particular that deserve attention.
Blackrock Neurotech was founded in 2008 and its development of Utah electrodes has been widely adopted by the neuroengineering community. In fact, the electrodes that collect Ian Burkhar's neural signals so that he can control his hand are the Utah electrodes developed by Blackrock Neurotech. In November 2022, at the annual meeting of the American Society for Neuroscience, Blackrock Neurotech showcased its next-generation neural electrode, Neuralace, an ultra-thin, flexible, lace-shaped neural interface with more than 10,000 electrodes, nearly 10 times the number of electrodes for "neural connection" N1 implants.
The technology developed by Synchron is even more exciting. Back in 2021, the application for a human trial of Synchron was approved by the FDA. In January 2023, the company's researchers and their collaborators published their trial in four severely paralyzed patients in JAMA Neurology, one of the top academic journals in the field. Four patients had Stentrode, a brain-computer interface electrode developed by Synchron, implanted in their brains. After receiving the necessary training, these patients are able to mentally perform daily activities on the computer, such as typing, sending and receiving emails, shopping, etc. The researchers also followed the patients implanted with Stentrode for a year, and the paper mentions that the follow-up ended in January 2022, which means that the implant surgery was completed in January 2021, much earlier than the "neural connection". Follow-up results showed that one year after Stentrode implantation, Stentrode's signal acquisition ability remained stable, and patients did not show significant adverse symptoms.
What's even more amazing is that the implantation of Stentrode does not require a craniotomy. In appearance, Stentrode is more like a vascular stent for the treatment of cardiovascular and cerebrovascular diseases. Implantation is done through a neurointerventional procedure: the researcher directs the Stentrode through the patient's jugular vein to the appropriate location in the brain, where the Stentrode collects the nerve signals generated by the patient's mind. Judging from the paper in the Journal of Neurology of the American Medical Association, as well as the video of patients using the brain-computer interface posted on the Synchron website, the performance is not inferior to that of the "neural connection" N1 implant.
Given this history of BCI technology and the performance of competitor technologies – especially Synchron's Stentrode – "neural connectivity" is neither a pioneer in the field of BCIs, nor is it a leader at present. Moreover, at present, there is too little information about "Neural Connection", not only is the video showing its effect on Nolan Arbo very short, but also the paper has not yet been published, so it is difficult for industry insiders to make an accurate and objective evaluation of it. In addition, there are many questions to be answered and explored by "neural connectivity", such as whether the efficacy of N1 implants deteriorates over time (a common problem faced by implanted electrodes), and whether the long-term retention of implants in the human brain will have adverse effects. Synchron and Stentrode already have certain results on these issues.
Neural Connection's current achievements may be commended, but it may not be time to give too much praise.
Chen Bin
Editor-in-charge: Zhu Liyuan