This wrinkled layer on the outside of the brain is called the cerebral cortex, and is the key area for all important brain functions. Understanding the brain requires knowing not only the structure and function of the cerebral cortex, but also the function of those complex nuclei in the cerebral cortex. It has taken scientists at least 200 years to understand them.
What advances have actually been made in brain science in the last 200 years?
Now, what we know about the brain, such as how the brain processes information, how nerve cells encode and conduct information, and how information interacts from one neuron to another ...... these conduction mechanisms are better understood; and what different neurons do, and what responses they produce in various functions, is also very well understood.
Over the past century, the Nobel Prizes have involved important discoveries in neuroscience related to the encoding and storage of information in the brain. But we only know a great deal about how nerve cells process information, and not much about the complex network structure of the brain as a whole.
We don't know much about what makes nerve cells react in certain ways; we don't know much about information processing in the brain, and we don't know much about sensation, emotion, or some of the higher cognitive functions - thinking, decision-making, and even consciousness.
Although there has been considerable progress in brain science, the unknown is much more than the known. To draw an analogy, brain science is now in a situation equivalent to that of physics and chemistry in the early 20th century, where many things have been clarified, but major understandings and breakthroughs have not yet occurred.
So brain science is now a relatively mysterious field in the biological sciences, and from this point of view, brain science will become a very important field in the development of life sciences in the future. If young people want to study science in the future, brain science is a frontier science, not only in this century, but even in the next century.
The most pressing problem in brain science
The most critical problem in brain science is that we know very little about how the brain functions and how neural networks work.
We know that different parts of the brain's cortex have different functions. For example, the back of the brain is in charge of vision, the top most front of the brain is in charge of motor, sensory, olfactory, and in front of that there is an area in charge of language. If there is damage to the brain, such as after a stroke (commonly known as a stroke), the corresponding functions of the damaged area will be lost.
At the moment, we only have a general understanding of the relationship between brain regions and function, but more details are not clear.
For example, a very widely used brain imaging technique, called PET, which is available in major hospitals, can tell people that the brain is not functioning as it should, but it can tell them that the brain is functioning as it should. It can tell people which areas of the brain are electrically active, and if there is electrical activity, it indicates that there is function going on in that area. If there is abnormal electrical activity, it means that there is an abnormality in that function.
For example, if we test a person with normal brain function, and we have the subject lie down in a machine, and we show him a few words, you'll see electrical activity in the back of the brain, which is characterized by an increase in glucose use. The glucose in the subject's body is radioactive and is injected into the bloodstream separately. From this, researchers can quickly tell that there is activity in the subject's brain.
The subjects were given a few words to listen to, and there was electrical activity in their auditory areas. We can now observe this in real time. Ask a subject to say a few words and the left side of the brain responds in the language area.
But when the subject closes his eyes and doesn't speak or listen, and thinks back to what the words he just saw mean, there's electrical activity all over his brain. This strange phenomenon suggests that thinking is a very complex thing that involves many areas of the brain.
Why is it that just thinking about the meaning of a few words causes all the brain networks to start moving? Understanding this is currently quite difficult, and requires knowledge of the full range of unknown mysteries of the brain.
When Science celebrated its 125th anniversary, it asked hundreds of scientists around the world to list what they considered to be the most important cutting-edge scientific questions in the world today, and ended up with 125, 18 of which were in brain science.
At the top of the list were the biological basis of consciousness, memory storage and recovery, human cooperative behavior, the biological basis of addiction, the causes of schizophrenia, and what triggers autism, or autism, which are all major issues of interest and unresolved. Although the questionnaire was made 10 years ago, the major brain science questions that we now recognize remain unchanged.
To understand these questions, it is important to know the neural networks of the brain. Neural networks are as complex as wires (cables), and in the human brain, hundreds of billions of cells are connected together, sending out many wires - which we call axons - to make connections with other cells, ultimately forming this network.
The brain network is very complex, with a large number of neurons. There are 100 billion neurons in the brain, and each neuron has a different pattern of firing, a different pattern of coding, and a different way of processing information. So it's going to be a big challenge to understand how this complex system works.
We can understand this network better at three levels.
Functional imaging such as PET Imaging or MRI Imaging, as just described, provides a macroscopic view with a resolution in the centimeter or millimeter class. At this scale, one can roughly see where the nerve bundles go between brain regions.
Each nerve bundle is composed of thousands of nerve cell fibers. To know further details, the neural circuits must be studied at the mesoscopic (the state between micro and macro) level, to understand how each nerve cell connects with and conveys information to other different kinds of nerve cells, and what activity there is during various functions.
It is also possible to look at cells under an electron microscope, from the micrometer to the nanometer level, so that the microscopic scale gives a much finer view.
One of the most critical aspects of neuroscience today is to move from the known macroscopic level to the mesoscopic level, and thus to understand the formation and function of brain network structures.
For example, we fluorescently labeled and sliced 52 cortical nerve cells from mice and reconstructed their three-dimensional structure, where each color represented a nerve cell.
It turns out that the brain is unimaginably complex. And that's just 52 cells; the human brain has hundreds of billions of cells, so how difficult it would be to really analyze! Even for these 52 cells, there are different kinds, and the rules for their distribution in the brain are different.
This is a major challenge facing neuroscience today. So, the first key point for future brain science is to figure out the network structure of the brain at the mesoscopic level, i.e., the structure of the map.
Information conduction in the brain relies on electricity, and electrical activity is conducted in nerve cells like waves. It's different from the conduction of electrons in a wire, because this transverse wave is caused by the flow of ions across the cell membrane - cations flow into the cell from the outside, causing fluctuations, which are constantly pushing forward, and are pushed at a much slower rate than the flow of electrons, which is only a few hundred meters per second.
When an electrical wave reaches the axon terminal of a nerve, it passes information to the next cell, which we call a synapse. A nerve cell is able to pass electrical information to the next cell with the help of releasing a chemical called a neuromediator.
When the neuromediators reach the next nerve cell, they continue to trigger electrical activity in the next cell, which is the pattern of electrical signaling.
Issues such as how to observe electrical signals and the processing patterns of electrical signals in a network are key questions for us to understand today.
The one-piece, two-wing structure of China's brain program
Regarding the future of brain science, the first one is to understand the brain, which is one of the ultimate goals of our understanding of nature. We often mention the mysterious outer space, and for human beings, there are many unsolved mysteries in the universe, such as dark matter and dark energy. In fact, there is also a universe in our brain, and what is the structure of this inner universe of the human body and how does it work is what we need to understand in the future.
What are the benefits of understanding this?
On the one hand, it gives us a deeper understanding of nature, and on the other hand, it can have very important applications - simulating the brain and creating machines that are as intelligent as people, which is the ultimate goal of AI and one of the directions in which brain science is developing.
In addition, in terms of population health, the brain is so important that we need to protect it, promote the development of intelligence, and prevent the decline of the brain as well as the emergence of brain diseases, which is also another important direction for the future development of brain science.
It took Chinese scientists 4 years of discussion to formalize the Chinese brain program in 2018. All countries in the world have brain programs, and the brain programs of the United States, Japan and the European Union are not small in scale.
The program is the future of Chinese brain technology. So what is it going to do? Just like the three directions mentioned below, China's brain program has a structure of one body and two wings.
The main structure is the neurological basis of the cognitive function of the brain, which was introduced earlier, that is, the foundation of the network, and we must know the structure of its map, figure out the connection map, and the structural map. On this basis, various platforms are built to help parse the functions of the above maps.
To this end, we hope to launch an international scientific program led by Chinese scientists to do neural connectivity mapping at the mesoscopic level of the whole brain. For mesoscopic mapping, not only Chinese scientists are interested, but also scientists from all over the world. Through this program, people can study the brain maps of animals, especially model animals (including mice, rhesus monkeys, and other primates that are most similar to humans).
One wing is going to be the diagnosis and treatment of brain diseases, resulting in a variety of new medical industries. The other wing is brain-like artificial intelligence, brain-like computing, brain-computer interfaces and other new technologies related to artificial intelligence, which will have a significant impact on the future of the artificial intelligence industry.
This is the current direction of China's brain program, which is also recognized as the best direction. Compared to the brain programs of other countries in the world, although our program started slowly, our design is the most successful, and we hope that its implementation will also be the most successful.
Producing a neural connectivity map of the whole brain is the only way to resolve the ultimate function of neural circuits
So what are the principles of brain cognition?
The first is basic brain cognition. Our senses, our reception of external information, including perception, learning and memory, mood and emotion, attention and choice, these are all basic brain cognitive functions. Fruit flies, mice, monkeys, and even zebrafish, nematodes, and many other animals have this basic function.
As for advanced brain cognitive functions, only higher animals above primates have them. These include ****sentiment and empathy - if you are sad, I feel sad too; social cognition, cognition inside a social group; cooperative behavior, which is very special and complex; various consciousnesses, such as human self-consciousness; and language, which is a very complex language not found in other animals.
Understanding the mechanism of the above cognitive functions is important for designing the next-generation artificial intelligence that is similar to the human brain.
To design AI devices that can not only understand speech and recognize speech, but also understand semantics, you also need to know how the human brain processes language.
For this to work, model animals are necessary. We can't experiment directly on humans because of the ethical issues involved.
Since the macaque's brain structure is very close to that of a human, it is a good model animal. So we have to do all kinds of manipulations on animals like macaques to find out how it works, and then later extrapolate that to see if the human brain is the same.
Within the neural basis of cognitive function, the most critical thing is still to produce a whole-brain neural connectivity map. We need to know the kinds of neurons in the brain, how the types of neurons are set out. This is a very important task, and we are doing research all over the world, and we need to do it too.
Once we know the types of neurons, we need to figure out the output fibers and input fibers of each type of neuron in each brain region, and where they go, and that's the structural mapping.
With the structural map, we can then map their electrical activity, see when the waves are coming and how they are transmitting information, which is the activity map.
It is only after the full mapping is out that the ultimate function of the neural loops can be resolved.
Brain disease treatment faces the challenge of finding specific drug targets
In China, a major application of brain science is to serve a healthy China. How to maintain healthy brain development and intellectual development is a very important social issue. Maintaining normal brain function and slowing down brain degeneration are essential for a healthy life.
Neurodegenerative diseases are a major problem in an aging society. Currently, there are more than 100 million people over the age of 65 in China, making it the world's oldest country, even surpassing India, and it has basically entered an aging society.
Therefore, it is very important to prevent and treat various diseases related to aging. Taking Alzheimer's disease (dementia), which is the most commonly heard of, as an example, if there is no good treatment, by 2050, more than 100 million people around the world will suffer from Alzheimer's disease; and on average, 1/3 of the elderly people over 85 years of age will have the possibility of developing Alzheimer's disease. If the Chinese Brain Program can delay the onset of Alzheimer's from 85 to 95 in 15 years, it will be a huge contribution.
In fact, not only Alzheimer's disease, according to the World Health Organization, brain-related diseases, including various neurological and psychiatric disorders, have the largest social burden of all diseases, accounting for 28% of all diseases, more than cardiovascular diseases and more than cancer. Therefore, the diagnosis and intervention of major brain diseases is a very important research in the field of brain technology in the future.
What are major brain diseases? For example, autism or autism and mental retardation in early childhood, depression and addiction in middle age, Alzheimer's disease and Parkinson's disease and other degenerative brain diseases in old age are all major brain diseases.
Only by fully understanding their mechanisms can we find the most effective solutions. However, our understanding in this regard is limited, especially for depression, bipolar (commonly known as bipolar disorder), schizophrenia and other mental illnesses, and is not clear what exactly causes them. It could take decades to figure these out.
However, we can't wait until we have the causative mechanism completely figured out before we can treat the disease, so it is important to develop early diagnostic indicators for a variety of brain diseases before the causative mechanism is completely clear. Once the diagnostic indicators are available, early intervention can be carried out. For example, if the memory begins to deteriorate, what means are available to slow down or delay the deterioration. These interventions can be medications, physical, psychological or physiological interventions.
Playing a game is also an intervention, it's a mental and physical intervention, you have to move, you have to think, you have to react quickly.
The various interventions that have been developed in the diagnosis and treatment of brain diseases have to be tested on animals before they can be applied to humans, and this involves ethical issues. If they are not clearly developed, they cannot be tested in clinical trials. Therefore, it becomes very important to establish a very good disease model for non-human primates such as rhesus monkeys.
In addition to the lack of clarity of the mechanism, brain disease treatment also faces the challenge of finding specific drug targets.
Drugs have side effects, but the side effects of drugs for other diseases are not as significant as those for brain diseases. This is because brain disorders occur when certain networks in the brain become abnormal.
Some of these networks are abnormal to produce this disease, and others are abnormal to produce other diseases. But drugs target molecules and cells, and since brain networks are made up of similar nerve cell-synapse connections, it's hard to find specific drugs.
This is why big pharma has been developing drugs for brain diseases for 20 years, and most of them have failed so badly that most of the big companies have given up. Because the development cycle of each drug is unusually long, ten or twenty years, billions of dollars of investment, the development failure rate of more than 90%, the big companies feel that it is not worth it, so they gave up. Now we can only rely on researchers to make very good products in the lab, and large companies to follow suit and put them into testing.
In the preclinical period, in order to determine whether the drug is usable or not, it is also necessary to carry out animal experiments. The first indicator of testing is the safety of the drug, i.e. whether it is safe for animals to use, whether their health will be adversely affected, and the metabolism of the drug.
Previously, drug testing was often done in primates such as rhesus monkeys, but there is a lack of primate models for efficacy testing. This is because the prerequisite for efficacy testing is that primates such as rhesus monkeys show symptoms of the disease in question in order to conduct efficacy experiments. However, there are currently no relevant models of primates in the hands of researchers, and the previous models are of mice, which are not usable, so scientists are also working on modeling.
The cloned monkey project done recently is to develop disease models of cloned monkeys for application in brain disease treatment.
Brain-like AI with generalized AI is likely to emerge in the next 20 to 30 years
Another important application of brain science research is in the area of brain-computer intelligence technology, brain-like research.
In this field, a very important future development direction is the brain-computer interface and brain-computer fusion of new methods, as well as a variety of brain activity stimulation methods, modulation methods, as well as a new generation of artificial network models and computational models.
Although the current deep network computational models are very good, they are still far from the human brain. If we can go further and develop new computational models and new neuron-like processing hardware for human brain-like computers, and apply them to the new generation of computers, it is possible to make even better and more efficient computers, which will also be much closer to human beings in terms of their computational capabilities, and will consume less energy and be more efficient.
In addition, brain-like computing robots and big data processing are also future directions for brain-like research. I will focus on the Turing test.
You may have heard of the Turing test, how to determine that a machine has human intelligence? Turing put forward such an idea 70 years ago: in the case of each other can not see each other, respectively, with a machine and a person to talk, and in the course of the conversation, to distinguish each other is a machine or a person. If it is not possible to distinguish the identity of the other party, it can be recognized that this machine has the intelligence of a human being. Among them semantic understanding is the most critical.
For years, people have wanted to make machines that pass the Turing test. The criterion for passing the test is that if 1/3 of people can't tell whether they are talking to a machine or a person within 5 minutes, the machine wins.
Xiao-Ice is an AI chatbot launched in China by Microsoft's (Asia) Internet Engineering Institute that can continuously improve itself through conversations, increasing its knowledge base and enhancing its responsiveness.
Although Xiao-Ice, which has been around for many years, has a high level of conversational ability, it's still easy to tell that it's not really a person, but just a machine.
Today, if you really want to make good brain-like intelligence, you have to rely on the new Turing test. What is the new Turing test? In addition to language ability, the test indicators should also include the ability to perceive and process a variety of information.
Specifically, a robot and a person can each operate a robot hand to play with a toy, and at the same time be asked to talk to each other about the action in order to discriminate. It's easy to see how a similar test could be more complex than talking to a computer.
The teamwork aspect is also tested. Ask a robot to work with a human on some activity, such as a race, and see if you can tell which of the team members is the robot and which is the human. These are all things that the new Turing test covers.
We can expect that brain-like AI with generalized AI that can pass the new Turing test may emerge in the next 20 to 30 years.
The author of this article, Pu Muming, is an academician of the Chinese Academy of Sciences (CAS), the director of the Institute of Neuroscience, CAS, and the director of the Center of Excellence in Brain Science and Intelligent Technology of CAS
Speech from the SELF Gezhi Lunar Forum
Organized by the public number of the Academy of Sciences