The first 8,000-word long text in China and the United States analyzes the global hotspot brain-computer interface (heavy dry goods)

"Everything we imagine will become reality."

If any technology today is closest to science fiction, it must be the brain-computer interface.
Brain-computer interface research has achieved conscious typing (an average of 39 letters entered in 1 minute), and has also achieved mind control, such as human control of mouse behavior, allowing it to complete complex tasks. It also achieved a part of consciousness uploading, which even makes people doubt whether the will is free.
In the future, it is even expected to achieve some lost perception capabilities, such as vision; it is also possible to convert non-human perception capabilities into human perception capabilities, which is actually very anti-sky, such as the perception capability of ultrasonic waves (like getting it from a bat). This ability is the same), such as sensing magnetic fields, etc., it is like having a super power!
As a new way of control and communication, brain-computer interface can also be applied to the wider field of brain-computer fusion, which is the so-called fusion of silicon-based organisms and carbon-based organisms, creating superhuman beings and allowing people to further extend their brains naturally.
The development of brain-computer interface has put forward new requirements on the mechanism of EEG, brain cognition, brain rehabilitation, signal processing, pattern recognition, chip technology, computing technology and other fields, and people will greatly deepen the understanding of the structure and function of the brain. awareness.
With the continuous improvement of technology and the efforts of multi-disciplinary integration, the brain-computer interface will be gradually applied to reality and benefit mankind.
Silicon Valley Live (service number: guigumitanv), together with scientists from Harvard University Brain Science Center and industry experts and scholars, jointly created the first long-term analysis of the brain-computer interface industry in China and the United States, in-depth deconstruction of the technical route in the field of brain-computer interface, and depicting the commercialization trend of brain-computer interface And subject maps, foresight never before seen.
Continuing this article, there is also the first series of courses on brain-computer interface in China and the United States. The cross-border evangelism of Harvard brain science and brain-computer interface practical experts is bound to bring unique insight and practical experience to brain-computer interface practitioners and enthusiasts. (See the end of the article or pay attention to the Silicon Valley Live service account guigumitanv)
Definition of brain-computer interface
First , what is a brain-computer interface?

　　Image source: www.engineering.com
Brain-Computer Interface (BCI): It is a direct connection pathway established between human or animal brain (or culture of brain cells) and external devices.
In this definition,
"brain" means the brain or nervous system of organic life forms, not just "mind" (abstract mind).
"Machine" means any processing or computing device in the form of simple circuits to silicon chips to peripherals and wheelchairs.
"interface" = "mediator for information exchange".
The definition of "brain-computer interface" = "brain" + machine "+" interface".
That is, a connection pathway for information exchange created between a human or animal brain (or a culture of brain cells) and an external device.
Brain-computer Interface is a multidisciplinary field, and the core disciplines involve cognitive science, neural engineering, neuroscience, etc.
Technical knowledge of brain-computer interface: implementation steps and analysis The basic implementation steps of
brain -computer interface can be divided into four steps: signal acquisition >>Information decoding and processing>>Recoding>>Feedback.

　　Image source: www.engineering.com
1. Information collection The division of
brain -computer interface is generally based on the way of information collection, which is usually divided into invasive and semi-invasive Type, non-invasive type (outside the brain).
Invasive type: This type of brain-computer interface is usually directly implanted into the gray matter of the brain, so the obtained neural signal quality is relatively high. However, its disadvantage is that it is easy to induce immune response and callus ( Scar), resulting in the deterioration or even disappearance of the signal quality.
The signals obtained by the invasive method are direct neural signals.
Partially invasive: The interface is generally implanted in the cranial cavity, but is located outside the gray matter, and its spatial resolution is not as good as that of the invasive brain computer Interface, but better than non-invasive. Another advantage is that it has less chance of eliciting immune response and callus, mainly based on cortical electroencephalography (ECoG) for information analysis.
Non-invasive: It does not enter the brain and can be easily worn on the human body like a hat. However, due to the attenuation of the signal by the skull and the dispersion and blurring effects of the electromagnetic waves emitted by neurons, the resolution of the recorded signal is not high and it is difficult to determine The firing of a signalling brain area or associated single neuron.
A typical system is electroencephalography (EGG), which is one of the main information analysis techniques for potential non-invasive brain-computer interfaces, mainly because of its good temporal resolution, ease of use, and portability. and relatively low prices.

　　Image source for EEG equipment: www.engineering.com
However, one problem with EEG technology is its sensitivity to noise; another real-world obstacle to using EEG as a brain-computer interface is the extensive training the user has to do before working.
2. Information Analysis Once enough information has been
collected , the signal is decoded and re-encoded to deal with the interference. There are many interferences in the process of EEG signal acquisition, such as power frequency interference, eye movement artifacts, and other electromagnetic interferences in the environment.
The analysis model is the key to the information decoding process. According to the different acquisition methods, there are usually models such as electroencephalogram (EGG) and cortical electroencephalogram (ECoG) that can assist in the analysis.
Signal processing, analysis and feature extraction methods include denoising filtering, P300 signal analysis, wavelet analysis + singular value decomposition, etc.
3. Recoding
The parsed information is encoded, and how it is encoded depends on what is desired to be done. For example, to control a robotic arm to pick up a coffee cup and drink coffee for itself, it needs to be encoded into the motion signal of the robotic arm. It is very complicated to accurately control the movement trajectory and force control of objects in a complex three-dimensional environment.
But the coding forms can also be varied, which is why the brain-computer interface can be combined with almost any engineering discipline. The most complex cases involve output to other organisms, such as a mouse, to control how it behaves.
4. feedback
It is also very complex to get feedback from the environment and then act on the brain. Humans perceive the environment through perception and transmit feedback to the brain, including vision, touch, and hearing.
The realization of this step of the brain-computer interface is actually very complicated, including the mixed analysis of multimodal perception is also difficult, because the process of feedback to the brain may be incompatible.
An important milestone in the history of brain-computer interface
In 1924, German psychiatrist Hans Berger discovered EEG.
In 1969, the University of Washington School of Medicine used monkeys for EEG biofeedback research.
In the 1990s, after Nicolelis completed a preliminary study of mouse motor brain waves, he realized an experiment in night monkeys that could extract signals from cortical motor neurons to control a robotic arm.
In 1999, Garrett Stanley of Harvard University attempted to reconstruct visual images by decoding the firing of neurons in the lateral geniculate body of the cat's thalamus.
After 2000, Donoghue's group implemented motor control of a cursor on a computer screen in rhesus monkeys to track visual targets, in which the monkeys did not need to move their limbs.
In 2009, Theodore Berger's group at the University of Southern California developed a neural chip that can simulate the function of the hippocampus. The team's neurochip is implanted in the rat brain, making it the first advanced brain function prosthesis.
2012 Brazil World Cup - Robot armor, an amputee disabled in a robot armor, scored a goal with a brain-computer interface and a mechanical exoskeleton.

　　In 2014, researchers at the University of Washington achieved direct brain-to-brain communication by transmitting EEG signals over the network.
In December 2016, Bin He of the University of Minnesota and his team made a major breakthrough, allowing ordinary people to control objects in a complex three-dimensional space with only "ideas" without implanting brain electrodes. Including manipulating the robotic arm to grab, place objects and control the flight of the aircraft. The research results are expected to help millions of people with disabilities and neurological diseases.
The experimental results of Bin He and his team
In February 2017, Stanford University electrical engineering professor Krishna Shenoy and neurosurgery professor Jaimie Henderson published a paper announcing that they had successfully allowed three paralyzed subjects to precisely control the cursor on the computer screen through simple imagination. Three paralyzed patients successfully entered what they wanted to say on a computer screen by imagining, and one patient could type an average of 39 letters in a minute.
Challenges
of brain -computer interface Moore's law of brain-computer interface:

According to the above chart, at an average rate of 7.4 years to double the number of neurons that can be recorded simultaneously, it will take until 2100 to record one million neurons simultaneously, while To record all neurons in the human brain (5-10 billion), it will take until 2225.
Therefore, how the brain-computer interface solves the bandwidth problem has become the key point of academic research breakthroughs. Neuralink, founded by Elon Musk, is working to accelerate this problem.
Brain-computer interface is also a complex interdisciplinary subject, which generally has two challenges, one is engineering challenge and the other is theoretical challenge.
Theoretical studies are all grappling with one or both of these two questions:
1) How to get the right information from the brain?
2) How to send the correct information to the brain?
The first is "brain to machine," capturing the brain's output -- recording what neurons say.
The second is "machine-to-brain," where information is fed into the brain or otherwise alters the natural flow of the brain -- which is to stimulate neurons.
At present, there have been some research results of "from brain to machine", but "from machine to brain" has almost no clue. Basically, it can be said that it is pitch black with only a few lights.
What does "from machine to brain" mean? That is, reverse encoding perception into signals that can be read by the brain. For example, whether you can record the touch when you touch a kitten or a piece of your imagination and reproduce it back to you through a machine, it is also a good idea to help the blind to reconstruct vision.
Machine-to-brain research is much slower than brain-to-machine research, because the specific methods of neural coding in neuroscience are still unknown. The demand for neural coding knowledge from machine to brain is far greater than that from brain to machine. Neuroscience's study of single neurons is gradually becoming clear, but the magic of the brain cannot be explained at all.
Moreover, the difficulty in engineering lies in the fact that the brain-computer interface industry involves a large number of disciplines such as mechanical dynamics, machine learning, neuroscience, cognitive science, and information engineering, which require a large number of talents from various industries and cannot have shortcomings.
In addition, the greater difficulty in engineering also includes cost control, whether commercialization can be achieved by reducing costs through reasonable processes and processes.
Commercialization of brain-computer interfaces
Medical and health: The
medical direction is mainly divided into two directions, namely "intensification" and "recovery", both of which have extremely ambitious "money prospects", especially the direction of strengthening. At this stage, the recovery class is mainly used because it is easier to implement.
The "enhanced" direction mainly refers to the implantation of chips into the brain to enhance memory and promote direct connections between the human brain and computing devices. This is the so-called "human enhancement" (Human Intelligence, HI). The shallow level of research is one-way between the brain and the machine, and the deeper level will be the two-way between the machine and the brain. At present, the "strengthening" direction includes Neuralink founded by Musk and Kernel, which has received a $100 million investment.
The "recovery" direction mainly refers to the corresponding recovery training for ADHD, stroke, epilepsy and other diseases. The main method is neurofeedback training. This direction has been widely used in some hospitals, clinics, and rehabilitation centers around the world, and many start-up companies are making wearable devices in this area.
Reasons for the lack of “enhanced” directions: first, because of the high difficulty of implementation; second, because the market has not been fully educated, the thinking paradigm is difficult to change in the short term, and the willingness to pay has not reached the critical value due to insufficient technical capabilities, but in the military field In fact, there are already many applications, and the military has invested a lot of money.
Finally, it is worth adding that the "health care direction", which is meditation to reduce stress, some startups have launched a brain wave detection headband to help users improve the effect of meditation through real-time audio feedback. In fact, in North America, the meditation market is very large, and this is a market segment that can definitely be tapped.
VR direction:
At this stage, the interactive experience of VR/AR needs to be improved. The current solution is to use voice recognition and gesture recognition, but if you use a brain-computer interface, you can use your thoughts to control the menu navigation and option control of the VR interface. Greatly improve the user experience. At present, the company that is relatively advanced in this area is MindMaze, whose total financing has exceeded 100 million US dollars.
Education technology:
This direction is actually somewhat similar to the "recovery" direction in the medical direction. Education technology is a market worth 100 billion yuan. Currently, BrainCo, a Boston-based startup, is working on this direction, mainly to detect students' attention in real time, so as to help teachers understand the classroom situation in time and change their teaching methods. The market development in this field is currently mainly on the B side.
Smart home:
Smart home is an imaginary space for the cross-domain combination of brain-computer interface and IoT (Internet of Things). In this field, brain-computer interfaces play a role similar to a "remote control", helping people control lights, doors, curtains, etc. with their thoughts, and further control home service robots.
Global Brain-Computer Interface Market Size

Brain -computer interface is a new input and output method, and its application will span several industries. The brain-computer interface operating system is also very likely to become another major human-computer interaction system after Windows (representative of computer operating system), iOS (representative of mobile phone operating system), and Alexa (representative of voice operating system).
Narrow market size: From the perspective of brain-computer interface devices (EEG/EMG) alone, the market size will reach $2.5 billion in five years.
Market size in a broad sense: From the perspective of several technology fields that will be deeply affected by brain-computer interfaces, the market size will reach hundreds of billions of dollars in five years, including: ADHD brain-computer interface feedback treatment of $46 billion, brain detection system $12 billion $250 billion in education technology and $120 billion in gaming. (The data comes from a third-party projection within five years based on the market size of the past two years.)
Imagine that in the future, using brain-computer interface technology to play "Honor of Kings" will be very similar to the mind control plot in "Avatar"?
Factors influencing the commercialization of brain-computer interfaces

　　Source: Allied Market Research
According to the report of Allied Market Research, in 2014, the biggest factors affecting the development of brain-computer interfaces were the lack of professional knowledge and ethical issues. They also speculate that by 2020, people will be more receptive to the technology and ethical issues will be reduced, but with it will come the dangers of conscious information leakage and brain hacking caused by cybersecurity threats. From the perspective of the medical and health industry, as the cures for terminal illnesses are realized one by one, neurological diseases will become the biggest problem in the medical industry in the future, and the incidence of brain disorders is on the rise. In addition, government funding, the miniaturization of components, and the expansion of the game industry will all develop in a positive direction.
Investment analysis of brain-computer interface Memorabilia of
private capital entering the brain-computer interface:
In 2001, John Donoghue and the research organization of Brown University jointly established Cyberkinetics to develop BrainGate, a brain-computer interface implantable system.
In 2009, the research center invested by Japan's Honda demonstrated the results of the brain-computer interface project, which opened the door for the brain-computer interface from scientific research projects to marketization.
In 2016, Braintree founder Bryan Johnson personally invested US$100 million to establish Kernel, a brain-computer interface company, and is currently researching brain-computer interface products that improve human memory.
At the end of 2017, Chen Tianqiao, chairman of Shanda Group, and his wife donated US$150 million to Caltech to set up the Brain Research Center.
In March 2017, Elon Musk announced his investment in the establishment of Neuralink, a brain-computer interface company.
In April 2017, Facebook announced the "Mind Typing" project. Zuckerberg invested a lot of capital and talent to build a brain-computer interface technology team.
Government input, "brain plan" of various governments:
The United States took the lead in proposing a national brain science plan in 1989, and named the last 10 years of this century as "10 years of the brain". In April 2013, the White House proposed the "Brain Project", which is considered to be comparable to the Human Genome Project. therapy. The US government announced that the "US BRAIN Initiative" has a start-up capital of more than 100 million US dollars. After adjustment, it plans to invest a total of 4.5 billion US dollars in the next 12 years.
European Union: In 1991, Europe launched the "European Brain 10 Years" plan. In January 2013, the European Commission announced that human brain engineering was selected as the "Future Emerging Flagship Technology Project", and set up a special research and development plan "Human Brain Project (HBP)", which can obtain 1 billion euros in the next 10 years (2013-2023) expenses. The project brings together more than 400 researchers from different fields.
Japan: In 1996, Japan formulated a 20-year "brain science era" plan, planning to invest 100 billion yen each year, with a total investment of 2 trillion yen. In September 2014, the Japanese Ministry of Science also announced the chief scientist and organizational model of its own "Brain Project". Japan's "Brain Project" focuses on the medical field, mainly using the marmoset brain as a model to accelerate research on human brain diseases such as Alzheimer's and schizophrenia. The Japanese government's budget for the "Brain Project" in 2015 was about 6.4 billion yen (about 63.75 million US dollars).
China: "Brain science and brain-inspired research" has been included in the national major scientific and technological innovation and engineering projects in the "13th Five-Year Plan" outline. The Chinese Academy of Sciences established the Center for Excellence in Brain Science and Intelligent Technology with 80 elite laboratories in 20 institutes earlier this year. For the "China Brain Project", scientists in various fields put forward a layout proposal of "one body and two wings": that is, the research on the neural principles of brain cognition is the "main body", and the development of new methods for the diagnosis and treatment of major brain diseases and new technologies for brain-computer intelligence are the "two wings". ". The goal is to achieve world-leading results in the three frontier fields of brain science, early diagnosis and intervention of brain diseases, and brain-like intelligent devices in the next 15 years. According to a rough estimate, my country's main investment in this field has increased from about 348 million yuan per year in 2010 to nearly 500 million yuan per year in 2013.
Brain-computer interface industry
distribution

Top 10 most concerned brain-computer interface companies in the world According to the five dimensions of company technology, team/partners, development plans, products, and financing, the world's top ten most-watched brain-computer interface companies were selected.
Among them, Neuralink and Kernel focus on brain science applications, aiming at the direction of human intelligence (HI). These two companies, together with BrainGate, which focuses on medical and health, use invasive technology in EEG signal acquisition, and the remaining seven use non-invasive technology.
Among the seven non-invasive ones, g·tec and BrainMaster focus on developing high-precision EEG measurement equipment, and the products are aimed at clinical and scientific research levels.
The remaining 5 non-invasive ones are more inclined to consumer-grade brain-computer interface products. Among them, NeuroSky, InteraXon (Muse) and Emotiv are mainly making mobile wearable EEG devices for meditation, games and other needs. These companies often also have supporting APPs and SDKs for users and developers. MindMaze in Switzerland is committed to combining VR/AR and brain-computer interface, cutting into the two major fields of medical health and games. BrainCo, located in Boston, first started in the field of education, but also involved in the fields of medical care and games.
Among the top 10 most concerned brain-computer interface companies, 7 are from the United States, and the other 3 are from Switzerland, Canada, and Austria. The financing situation and introduction are shown in the figure below:

　　The distribution of brain-computer interface scientific research strength
According to the brain-computer interface of major scientific research institutes around the world For the output and influence of research results in the interface field, we have selected these 20 research institutes for your reference and study:

Of course, in addition to these 20 research institutes, there are also the US Defense Advanced Research Projects Agency DARPA and Facebook's research institutes. Mysterious research groups such as Building 8 are engaged in research on brain-computer interfaces.
Brain-computer interface subject map
Finally , for those who do not want to miss another big trend after artificial intelligence, how can they get started quickly? Here, we draw a disciplinary map of brain-computer interfaces. In this typical interdisciplinary field, science students, engineering students, medical students, and liberal arts students will all find their own entry point.

　　Future Prospects of Brain-Computer Interface
At present , mainstream consumer-level brain-computer interface research mainly uses non-invasive EEG technology. Although it is easier to implement than invasive technology, the cost is still high. However, with the influx of talents and capital, non-invasive EEG technology is bound to develop towards miniaturization, portability, wearability and ease of use.
As for invasive technology, in the future, if it can solve the two major problems of human body rejection and the transmission of information from the skull to the outside, the computer will accurately identify people's thinking consciousness in real time. On the one hand, it will help the computer to better understand the characteristics of human brain activity, so as to guide the computer to better imitate the human brain; on the other hand, it will allow the computer to work better with people.
Of course, from the standpoint of Elon Musk, what he is worried about is that humans will be threatened by AI. We need to let the brain-computer interface act as a medium for connecting the brain and the machine to ensure that humans can fight against AI in the future. On this entire development path, topics such as "superhuman beings", "cyborgs", and "uploading consciousness to realize human immortality" that cannot be ignored will become issues that all human beings need to face.