When the brain is connected with AI, can we live forever?

20 16 July, the founder of space exploration technology company and Tesla, Elon? Elon Musk founded a new high-tech enterprise, Neuralink, to develop an implantable brain-computer interface, but the news was not known to the public until March 20 17.

In the next two years, Neuralink almost disappeared from the public's sight until the middle of 20 19. Mysterious? The company just held a press conference on brain-computer interface research, and published a peer-reviewed paper [1] in the preprint magazine bioRxiv, which attracted wide attention. At that time, the company had raised about $654.38+58 billion (including $654.38+0 billion from Musk himself) and had 90 employees, including neuroscientists, neurosurgeons, chip designers, biocompatible materials scientists, brain-computer interface experts, micro-manufacturing engineers and experts in other fields. This paper will introduce the origin and development of Neuralink and talk about the author's views.

Musk showed the latest progress of Neuralink at the press conference on August 28 this year. At the scene, Neuralink showed the state of piglets after transplanting brain-computer interface equipment.

Musk is a strange man among entrepreneurs. The companies he founded all have a sci-fi and very grand ultimate goal, trying to solve a major problem of sustainable development of human society; At the same time, there is a short-term goal that can be started, which can solve the immediate problems, carry out commercial operations, and make technical and financial preparations for the ultimate goal. For example, the long-term goal of space exploration technology company is to realize human migration to Mars or other planets, while its short-term goal is to develop recoverable rocket technology, reuse rockets and engage in space commercial activities.

Neuralink also adopts this model. Neuralink was founded because Musk was worried about the rapid development of artificial intelligence. Since 20 14, Musk has repeatedly warned the danger of artificial intelligence on many occasions, thinking that artificial intelligence is developing too fast and may far surpass human beings in the future.

20 14 August:? We need to be more careful about artificial intelligence. It may be more dangerous than a nuclear bomb. ?

Two months later:? If I have to guess what is the biggest threat to our survival, it may be artificial intelligence. ?

20 17 July:? I kept sounding the alarm, but people didn't know how to react until they saw robots killing people in the street, because it sounded so ethereal. ?

2065438+April 2008:? Artificial intelligence is a very important topic. It will affect our lives in ways that we can't even imagine now. ?

Musk also believes that the US government is losing the war to control artificial intelligence. He said:? The existing supervision method is slow and linear, and we are facing an exponential threat. If you have only one linear response to the exponential threat, then the exponential threat is very likely to win. This is the problem. ? [3]

In Musk's view, people become the masters of this planet because of their high IQ. Once artificial intelligence surpasses human beings in this respect, they will treat us like pets today. However, it is impossible to completely limit the development of artificial intelligence, so his prescription is to develop a brain-computer interface that can be implanted in the brain and integrate the two before artificial intelligence completely surpasses human beings.

The ultimate goal of Neuralink is to create a "whole brain interface" so that almost all neurons in the brain can communicate smoothly with the outside world. This device will be fully integrated into the brain, so that the whole brain interface is "a part of the body" in the sense. This interface allows the brain to wirelessly connect to the cloud, so that it can connect computers and even other brains with the same interface. In this way, the communication between the brain and the outside world will become easy, and it will feel like thinking in your own mind.

The purpose of creating a whole brain interface is to make people "live with artificial intelligence" and become superman. Musk believes that by that time, when a person dies, he already has his own computer expansion and online expansion, just like an online ghost? ,? Do you exist more? Cloud? In it, not in your body? This is what some people call digital immortality.

Do you really want to build every brain neuron? Microelectrode-neuron interface? We should not only consider the huge number of neurons in the brain (more than 80 billion), but also consider the technical limit that only a few hundred electrodes can be placed in the brain at present, and each electrode can measure about 500 neurons at the same time at most. In this case, it is impossible to measure 80 billion neurons at the same time. So, the number of neurons that can be recorded at the same time (what does Musk call it? Bandwidth? ) has become the bottleneck of the whole brain interface.

The second difficulty encountered by the Neuralink project is how to implant electrodes into the brain. At present, all non-invasive techniques for recording brain activity have poor spatial resolution, which is far from the level of recording the activity of a single neuron, or poor temporal resolution, which cannot record the rapidly changing brain activity in real time (Figure 2). Needless to say, a single neuron record, even a small number of neuron records, requires traumatic craniotomy, which is acceptable only if the patient is in danger. But Neuralink's long-term goal is man-machine integration, which requires healthy people to undergo craniotomy. Therefore, security has become a major obstacle. In addition, intracranial implantation is expensive and can only be done by highly skilled neurosurgeons, so it is difficult to popularize. According to Musk's vision, this technology should be as convenient as laser correction of myopia.

Temporal resolution and spatial resolution of main techniques of brain activity recording.

Even if the above two bottlenecks are overcome, Neuralink still faces other serious obstacles: the communication between the brain and the computer should be wireless; Implants need to have good biocompatibility, do not cause rejection, and can be used in the brain environment for a long time; How to place a large number of electrodes in a limited intracranial space; How to process massive data in real time and get useful information from it.

According to Musk's consistent strategy of founding a company, Neuralink's short-term goal is to develop a medical implantable brain-computer interface, which can not only realize the current practical application, but also make technical preparations and raise funds for the long-term goal.

This device can be applied to patients with epilepsy to give an early warning before the seizure to remind patients to take medicine in time. Or help quadriplegic patients control robot hands or computer screen cursors with brain signals. Neuralink plans to bring this device to market around 2023.

Musk's timetable

Before the end of 2020

Brain-computer interface (BCI) is implanted in the human brain, which is first used for quadriplegic patients.

8- 10 year

It is possible to implant brain-computer interface devices into normal people's brains, although this "largely depends on the approval time of the regulatory authorities and the work of our devices for the disabled".

/kloc-within 0/0 years

Hope it can be realized in healthy people? Telepathy? Telepathy, that is, two people with brain-computer interface can communicate directly with brain signals.

25 years later.

It is expected to develop a whole brain interface, that is, all neurons of a person will be connected with the carrier of artificial intelligence, and artificial intelligence will be regarded as the expansion of one's own mental activities. In addition, the brains of healthy people with whole brain interface can form giant brains that can communicate with each other directly, and what new phenomena will be produced by such giant brains is beyond our imagination.

Has been realized

In the paper published in early August of 20 19, Musk and Neuralink introduced their three achievements so far: [1].

1 flexible multi-wire electrode array

This kind of silk is very thin and soft, and its width is only about 4-6? M, the electrodes containing gold are covered with polymer insulation layers, and each electrode extends out of the wire to receive signals. These small pieces are arranged in a string along the wire. Compared with the electrodes commonly used in brain-computer interface at present, this electrode is very soft and can move with the tiny activities of the brain, so it has less damage to the brain. At the same time, compared with the electrode number of the commonly used multi-electrode array, the electrode number of the multi-wire electrode array has increased by an order of magnitude, and each array has 96 strands of electrodes as many as 3072.

(1) The brain is connected with AI, can we? Immortality? ?

(b) electrodes? Silk? . Each wire includes 32 electrodes. (a) The appearance of the electrode wire, in which the electrode pieces protruding from the wire are arranged in a row, and the center distance of the pieces is 50? M .(b) enlarged electrode wire. [ 1]

Surgical robot with implanted electrodes

Electrode wires are thin, soft and numerous, which need to be implanted into the brain accurately in a short time. It is obviously impossible to implant them by hand. Musk's team has developed a neurosurgical robot that combines a microscope with a sewing machine. The robot has automatic implantation mode, and it can implant up to 6 wires per minute (192 electrodes). Each filament can be implanted into the brain independently with micron-scale accuracy, and can avoid surface blood vessels and aim at specific brain regions.

Although the whole implantation process can be carried out automatically, the surgeon still maintains complete control, and if necessary, he can manually fine-tune the position of the electrode wire before each implantation of cortex. Using this system, the team achieved 87. 1 in 19 operations. 12.6% (average? Implantation success rate.

3 implantable brain-computer interface chip

Neuralink's third achievement is to develop a chip that can be implanted into the brain-computer interface. The electrode array is packaged in a small implantable device with a low-power custom chip, which can amplify and digitize the signals recorded on 3072 electrodes. After the whole 3072-channel package, it only takes up less than (23? 18.5? 2) volume 2)mm3. A USB-C line can simultaneously transmit data recorded by all channels.

The Musk team has installed this system in the brain of mice and read the information of 3072 electrodes, which is at least one order of magnitude higher than the brain-computer interface currently buried in the human brain. Later, they applied a similar system to monkeys and realized the monkey brain control computer.

Brain-computer interface device implanted in rats for a long time. [ 1]

Next target

At present, the prototype developed can only use the USB interface placed on the animal's head when transmitting brain signals, which is quite inconvenient. Therefore, the next goal is to realize wireless transmission.

Neuralink said the proposed wireless sensor? N 1 sensor? . They plan to implant four N 1 sensors, three of which are in the motor area and 1 in the somatosensory area, so that they can not only control external devices with brain signals, but also receive sensory feedback to realize two-way communication between the brain and external devices. The sensor will be wirelessly connected to an external device installed behind the ear and can be controlled by a mobile phone application.

The company believes that this can help people solve a series of medical problems quickly, and hopes to test it on five paralyzed patients before the end of 2020 to see if the technology can help patients move the mouse cursor and type with their brains. At present, the epidemic situation in COVID-19 is serious, and it is not clear whether the company can apply for a license from the US Food and Drug Administration.

At present, when installing this system, the surgeon must drill a hole in the skull to implant the thread, which will bring discomfort to the patient. The company hopes to use laser beams to punch a series of tiny holes in the skull in the future, and it can be as automatic, painless and convenient as the current myopia laser surgery.

Schematic diagram of wireless brain-computer interface to be implanted in human body. [2]

Problems in the scientific community

Although the achievements of Neuralink are impressive, many scientists are deeply skeptical about its ultimate goal. 2065438+In July 2009, an article published in the podcast Verdict [5] questioned Musk's ambition. In this paper, Noel The Last Month Of Summer, emeritus professor of artificial intelligence and robotics, thinks that it is absurd for Musk to interfere with the brain in order to keep up with the pace of artificial intelligence. There is no research or evidence to support Musk's claim? Will artificial intelligence rise and kill us? In my opinion, AI is just a tool and it is people who decide how to use it.

Toby, professor of artificial intelligence and data at the University of New South Wales? Toby Walsh believes that it is doubtful that human beings need to integrate with artificial intelligence to be redeemed. We can't compare speed and memory with computers, but our emotional intelligence and social skills, creativity and adaptability are the places where human beings can always stay ahead of machines. He also said that Musk is famous for failing to keep his promise on time, and the neural connection of healthy people may have to wait for decades.

On May 13, 2020, Pesenti, the head of Facebook's artificial intelligence department and a computer scientist at Carnegie Mellon University, published a tweet criticizing Musk for being unintelligible in artificial intelligence, arguing that there is no universal artificial intelligence at present, and the development of artificial intelligence is still far from human intelligence.

Obviously, there are still great differences in people's views on Musk's idea of man-machine integration. Whether it is right or wrong remains to be tested.

As far as Space Exploration Technology Company and Tesla are concerned, it is enlightening and promising for Musk to combine short-term goals with long-term goals when he founded the company.

In my opinion, Neuralink's short-term goal is also expected to bring good news to the disabled, although the timetable may not be as optimistic as Musk imagined. However, its long-term goal is quite problematic in principle and practical feasibility, and it is very likely to become a castle in the air.

Although the long-term goals of Space Exploration Technology and Tesla are ambitious, they are basically pure technology and backed by a solid theoretical foundation. In contrast, what is the long-term goal of Neuralink? Establish whole brain interface and brain-computer fusion? However, our understanding of the human brain is still very superficial, so far there is no theoretical framework about the functional mechanism of the brain, and the foreseeable future is hard to find. Therefore, Neuralink's long-term goal is not only engineering and technical issues, but also scientific issues. However, the key problems in science are often difficult to achieve as planned by relying on a lot of manpower and material resources, and sometimes rely on opportunities and luck, and rely on the flashes of genius.

What's wrong with Neuralink's long-term goals?

First of all, there are 86 billion neurons in the brain, and it seems impossible to measure the activity of such a huge number of neurons at the same time.

The next best thing, the Neuralink team set the long-term goal to record one million neurons at the same time (I don't know how to estimate this number. As far as its sampling ratio is concerned, it is equivalent to saying that if 80 thousand people are investigated, the situation of people all over the world will be mastered. This presupposition is questionable. ) although? Stevenson's law shows that the number of neurons we can record at the same time seems to double every 7.4 years. [4] However, this is a rule of thumb, and whether it can be proved is always a big problem. Even if this speed is maintained, it will not reach one million until the end of this century, and every neuron in the brain will not be recorded until 2225. However, any exponential growth process will inevitably slow down or even smooth down at some point because of other factors.

In addition, due to the limited intracranial volume, even if electrodes thinner than wire electrodes are used, 86 billion electrodes cannot be implanted in the brain. Of course, supporters of the whole brain interface can also argue that completely different new electrodes may be developed in the future. For example, the neurodust technology developed around 20 1 1 is a silicon sensor with the size of 100 micron, which can be sprinkled into the cerebral cortex and communicated with neurodust through ultrasonic waves with a device with the size of 3 mm above the nearby pia mater. However, the nerve dust itself is already equivalent to the size of large neurons, so it is still impossible to spray the same amount of nerve dust into the brain. It has also been suggested that new methods such as optogenetics or carbon nanotubes can be used, but so far they are only some ideas.

Even if the electrical activities of all neurons are recorded at the same time, how to deal with such a huge amount of data is a big problem, and the electrical signals of neurons may not necessarily reflect the whole picture of the brain. Because the brain is essentially an electrochemical machine, besides electrical activities, chemicals such as neurotransmitters and neuromodulators are also extremely important in brain activities, and these factors are completely ignored in the whole brain interface. Another neglected factor is that the role of glial cells in the brain is still unclear. About these questions and the so-called? Spiritual upload? And then what? Digital immortality? This is just a myth. I'm talking to Carl? This point has been analyzed in detail in Dr. Schlagenhoff's discussion [6], and will not be repeated here.

Supporters of the whole brain interface argue that there will always be some ubiquitous technologies in the future that people in the past could not believe. People always underestimate the collective strength of human beings. ? Although there is some truth in this statement, we can't turn possibility into inevitability. Although people's group wisdom is unparalleled, and the development of technology often exceeds the expectations of ordinary people, it does not mean that no matter what kind of ideas can be realized, especially in a limited time.

The author observed an interesting phenomenon, that is, neuroscientists are often very cautious about uncovering the mystery of the brain, while technical experts are often full of enthusiasm and enthusiasm for bullfighting. This may be because technologists don't know enough about the brain. Ignorant and fearless? . In fact, Musk himself said:? Without a full understanding of technology, I think it is difficult to make the right decision. ? [4] He also admits that he knows the least about neuroscience in the team, so he will make the decision on the whole project, right?

At the beginning of the founding of Neuralink, Tim Urban, a blogger of the well-known technology blog Wait But Why, was invited by Musk to visit the company for a long time and had an in-depth conversation with most of his founding team. Based on this experience, Urban wrote a long article. [4]

He said in his blog post:? AI will run by itself because the conversation with people is too slow. The faster the communication speed, the higher the degree of integration between you and AI. The slower the communication speed, the lower the integration level. The worse our combination with AI is, the more independent AI is, and the more likely it is to betray us. If AI is completely independent and has much higher intelligence than us, how can you ensure that their optimization function will not run counter to the interests of mankind? If we realize the intimate relationship with AI, AI will not be independent, it will become you, and the relationship with your cerebral cortex is similar to the relationship between cerebral cortex and limbic system. In order to minimize the survival threat from AI, his (Musk's) strategy is essentially to make AI power become? By the people, by the people and for the people? . ?

The above may sum up the core idea of Musk's founding Neuralink, but in my opinion, these words themselves are not valid. First of all, artificial intelligence is not a discipline, it has no subjectivity or self-concept, neither consciousness nor will. And so far, these qualities are unique to the brain. However, scientists still don't know why the brain has these characteristics, let alone people have these characteristics. So in the foreseeable future, artificial intelligence can only be a tool, and whether it is good or evil depends only on the people who use it. The author agrees with some artificial intelligence experts that self-conscious artificial intelligence should not develop at all, let alone the possibility of development.

To take a step back, even if artificial intelligence has self and will, the prescription prescribed by Musk can't solve the problem. He wishful thinking that as long as people and artificial intelligence are integrated, artificial intelligence is a part of people and controlled by people. In fact, the combination of the two, one is strong and the other is weak, and it is very likely that the strong is dominant. Since Musk is worried that artificial intelligence will surpass human beings, the result of the integration of the two may not be that artificial intelligence becomes the soul of the consortium, and people become the tools and tools of artificial intelligence. Other people who are not integrated with artificial intelligence will become slaves or pets of this consortium. This fate is not much different from that of adult slaves. As for Musk, as long as everyone is integrated into artificial intelligence, this danger can be avoided, which is purely idiotic. It is impossible to achieve artificial intelligence integration for all people at the same time. Considering that the implantation of the whole brain interface must be expensive at the beginning, it must be affordable for a few rich people like Musk, so that the dominant groups can rule all beings. This prospect is as terrible as Musk's fear of artificial intelligence despotism.

The author believes that in the foreseeable future, there is no danger described by Musk, and his prescription will not solve any problems for ordinary people. As for this question, whether Musk is a prophet to save mankind or another alarmist alchemist, please judge for yourself!