Chapter I . Artificial intelligence – its essence concept

theories

1.1. Artificial intelligence concept

1.2. History of the development of artificial intelligence systems

1.3. Approaches to building artificial intelligence

1.4. Alan Turing's approach to artificial intelligence

1.5. Self-learning artificial intelligence

1.6. Artificial intelligence is a new information technology

revolution

Chapter II Quantum computers and neurocomputers

2.1. Quantum computer

2.2. Neurocomputer

Chapter III Basics of neural networks

3.1. Some information about the brain

3.2. Neuron as an elementary link

3.3. Neuro-like element

3.4. Neuro-like networks

3.5. Training a neural network

Chapter IV Can a computer think?

4.1. Is computer thinking real?

Conclusion

Bibliography

Introduction

It is impossible to imagine today without a computer. The use of computer technology today affects all areas of human activity, be it construction, industry, education, science, economics, etc.

Every year, computers become more powerful and productive, and technology is developing so quickly that analysts who made forecasts for the future of the computer industry 10 years ago now realize that they miscalculated greatly.

The development of computer technology is not only about increasing power, productivity and reducing the cost of materials and technologies, but also about the development and creation of new types of computers capable of thinking like humans.

Today, a home computer with a processor with a clock frequency of 5000 MHz is not a fantasy, although this was not even thought about before. If today the topic of my work is “Can a computer think”, it has a more philosophical orientation than a mathematical one. Then, after several decades, a thinking computer will not be able to surprise anyone, just like today a computer with a 2-core processor with a frequency of 5000 MHz. And if a student of the future writes a similar paper on a similar topic, then most likely by that time all the mathematical and logical principles of constructing artificial intelligence will be outlined.

The creation of artificially thinking machines can help humanity cope with those complex tasks that humans are not able to cope with today. For example, you can send robots to distant galaxies in search of extraterrestrial civilizations. Also, with the help of intelligent robots, it is possible to replace such professions as a hotel steward or a waiter, or use robots to ensure the safety of people. Robots are police officers or even sapper robots.

On the other hand, the creation of artificial intelligence can harm humanity. Many books in modern fiction are devoted to this, and quite a few films have been shot, the most famous of which is the film Terminator. This film shows in the future that a war has broken out between people and the robots they created.

Well, if we approach the problem more seriously, is it possible today to create a computer capable of thinking like a living person? Capable of communicating with a person in such a way that a person will not notice the difference between communicating with a person and a machine. So would machine communication be similar to real human communication?

Many scientists will say - yes, this is possible, but not today, while humanity has not yet fully studied the principles of the brain, and even more so cannot create an electronic brain or the so-called “cyber brain.” After all, the principles of operation of modern computers are quite far from the principles of operation living “biological” brain.

This is what pessimists say, and optimists are working on the creation and development of principles for the operation of artificial intelligence.

The topic of my work is not only relevant, but also interesting. In my work, I will try to outline the essence of artificial intelligence, to tell the story of the emergence of the theory of artificial intelligence. And I will try to answer the question - “Can a computer think?”

Chapter I

Artificial intelligence, its concept, essence, theories.

1.1 The concept of artificial intelligence.

To get closer to the problem of solving the question - can a computer think, it is impossible not to mention such a concept as artificial intelligence. This is exactly what I would like to devote the first chapter of my work to.

Artificial intelligence. Not so long ago, he was on a par with starships, extraterrestrial civilizations and other images of the future born from the imagination of science fiction writers.

Today, the phrase “artificial intelligence” has almost ceased to have a science-fiction character. Increasingly, this phrase appears in the description of new computer programs and complex technical devices. One can increasingly hear the assertion that at the current rate of growth in computer productivity and software improvement, the creation of artificial intelligence is only a matter of time.

The term intelligence comes from the Latin intellectus - which means mind, reason, mind; human thinking abilities. Accordingly, artificial intelligence (AI) is usually interpreted as the property of automatic systems to take on individual functions of human intelligence, for example, choosing and making optimal decisions based on previously gained experience and rational analysis of external influences.

In this definition, the term “knowledge” refers not only to the information that enters the brain through the senses. This type of knowledge is extremely important, but not sufficient for intellectual activity. The fact is that objects in our environment have the ability not only to influence the senses, but also to be in certain relationships with each other. It is clear that in order to carry out intellectual activity in the environment (or at least simply exist), it is necessary to have a model of this world in the knowledge system. In this information model of the environment, real objects, their properties and relationships between them are not only displayed and remembered, but also, as noted in this definition of intelligence, can be mentally “purposefully transformed.” It is essential that the formation of a model of the external environment occurs “in the process of learning from experience and adapting to various circumstances.”

Here I used the term intellectual task. In order to explain how an intellectual task differs from a simple task, it is necessary to introduce the term “algorithm” - one of the cornerstone terms of cybernetics.

An algorithm is understood as a precise prescription for performing a system of operations in a certain order to solve any problem from a given class (set) of problems. The term “algorithm” comes from the name of the Uzbek mathematician Al-Kho Rezmi, who proposed the simplest arithmetic algorithms back in the 9th century. In mathematics and cybernetics, a class of problems of a certain type is considered solved when an algorithm is established to solve it. Finding algorithms is a natural goal for humans when solving various classes of problems. Finding an algorithm for problems of a given type involves subtle and complex reasoning that requires great ingenuity and high skill. It is generally accepted that this kind of activity requires the participation of human intelligence. Problems related to finding an algorithm for solving a class of problems of a certain type will be called intellectual.

As for problems, algorithms whose solutions have already been established, then, as the famous expert in the field of AI M. Minsky notes, “it is unnecessary to attribute to them such mystical properties as “intelligence.” Indeed, after such an algorithm has already been found , the process of solving the corresponding problems becomes such that it can be performed exactly by a person, a computer (properly programmed) or a robot, without having the slightest idea about the essence of the task itself. It is only required that the person solving the problem be able to perform those elementary ones. operations, from which the process is formed, and, in addition, that it is pedantically and carefully guided by the proposed algorithm. Such a person, acting, as they say in such cases, purely mechanically, can successfully solve any problem of the type under consideration.

Therefore, it seems completely natural to exclude from the class of intellectual problems for which there are standard methods of solution. Examples of such problems include purely computational problems: solving a system of linear algebraic equations, numerical integration of differential equations, etc. To solve problems of this kind, there are standard algorithms that represent a certain sequence of elementary operations that can be easily implemented in the form of a program for computational cars. In contrast, for a wide class of intellectual problems, such as pattern recognition, playing chess, proving theorems, etc., on the contrary, this formal division of the process of finding a solution into separate elementary steps often turns out to be very difficult, even if their solution itself is not difficult.

Thus, we can rephrase the definition of intelligence as a universal super-algorithm that is capable of creating algorithms for solving specific problems.

Another interesting remark here is that the profession of a programmer, based on our definitions, is one of the most intellectual, since the product of a programmer’s activity is programs - algorithms in their pure form. That is why the creation of even AI elements should greatly increase the productivity of its work.

Recently there was an article about the parametric buildings of Zaha Hadid, but from the text it is not too clear what parametric architecture is in principle. Are parameters something to do with the equations that describe the “fashionable, stylish, youthful” lines of modern buildings? No, it's getting more interesting. In fact, parametric design is not so much beautifully curved three-dimensional objects as genetic algorithms, polymorphism, mobility, analysis of complex systems and other nonsense. If you are interested in what is currently happening at the intersection of architecture and information technology, read on.

Let me add some intrigue: the architects themselves call a lot of different things “parametric”. As usual, only the past can be sorted into pieces, but the present is pure boiling and confusion.

0. Parametricism is a visual style

Either a lot or nothing needs to be said about the style of smooth lines and curved surfaces and the corresponding design tools. They simply exist and give that recognizable result that you have all seen more than once.

By the way, what you just read is just a stereotype. In fact, a parametric building can look like anything, even a strict parallelepiped without windows. So there will be no pictures in this section. What lies behind the beautiful renderings is much more interesting than them.

1. Parametricism is when a form is created by analyzing the processes that will occur within it

Information modeling takes functionalism to a new level, when the processes occurring in a building are considered as a separate entity, like Hawking’s “four-dimensional carrot,” and the building, as it were, wraps itself around it, without introducing anything unnecessary.

The list of Zaha Hadid's buildings includes a mobile exhibition pavilion, the load-bearing structures of which are simultaneously walls, roofing, furniture, and interior decoration, and direct the movement of visitors between exhibits along the desired trajectory. It corresponds to the function as much as possible, and even disassembles when not needed - pass! Large buildings differ in materials and design, but the underlying principle is the same.

Among traditional buildings, there are also such examples, for example, an antique amphitheater in a natural relief depression, where the bottom is a stage, and the slopes are seats, steps, a supporting structure, and an acoustic surface that controls the distribution of sound. Neither add nor subtract.

However, in the pursuit of optimizing space, architects know how to forget about the psychological comfort of the users of their creations, which is why not all “parametric” creations are loved by citizens.

2. Parametricism is when an object changes in response to the properties of the environment or new functional requirements

Transformability, mobility, and the ability to interact with the environment are important guidelines for all modern architects, but for avant-garde artists this is of particular importance.

Do you know that the scales of an ordinary cone are pressed tightly together in rainy weather and spread out in dry weather due to the swelling and drying cycle of the wood fibers? This is what advanced architects are now striving for: so that the structure responds to changes with its basic elements, and not with complex and expensive technical devices.

A “smart” building, stuffed with sensors and controlled by a program, has already become mainstream, now researchers are looking for non-electronic methods and non-standard materials. For example, using thermocouples so that when heated in the sun, the element is deformed in the desired way.

“Inflatable” office building Media-ICT.

Modern buildings breathe, move, open and close their “eyes” with the help of diaphragms, generate clouds of nitrogen inside, dynamically change the optical and thermal insulation properties of the glass itself, and so on - in general, they live a rich and interesting life.

I will not list countless examples, turning the post into a cabinet of curiosities. An overview of several famous parametric buildings can be seen here. Let's take a better look at the designers' computers.

For projects where an object exists in dynamics, conventional CADs are not enough, and it is necessary to use BIM (building information modeling). The world is gradually moving towards “virtual building” technology, but our land area is still lagging behind this trend (if you are wondering why, read the analysis with graphs here).

3. Parametricism is when an object is created according to a pre-developed algorithm based on a large amount of incoming data

With the help of BIM programs, you can literally parameterize the design, turning it into a “3D equation.” That is, to create a model that, thanks to the given dependencies, will itself adapt to circumstances. Or create a set of rules that, based on available information, will generate something new. Algorithmic morphogenesis is applicable both in minimal objects, such as bus stops, and in large-scale ones, at the level of urban planning. By the way, the bureau of Zaha Hadid and Patrick Schumacher is “widely known in narrow circles” for parametric city master plans generated using scripts (KDPV demonstrates one of them, this is Singapore).

Here is an interesting video that illustrates parametric polymorphism. The text there is not in Russian, so I’ll explain a little what’s happening in the frame.

The main object is a block of several buildings. The number of its sides, their lengths and angles between them can be changed. The buildings themselves adjust their number of storeys, the number of apartments and rooms depending on these parameters. Interconnecting with each other, the blocks form a network, which, depending on the number of residents and the activity of traffic flow, changes the width of the streets, identifies the main ones and can replace part of the housing there with public institutions, for example, shops on the ground floors. This is a research work and not an actual design tool, but it can be used to understand the trend.

For a building, you can write a morphogenesis algorithm that, when applied to different source materials, will produce different results, but they will be members of the same “population.” The result is architectural polymorphism, a modern replacement for standard construction: it is possible to typify in such a way that there are no identical buildings, but there are identical technological and constructive techniques.

Picture from here.

By specifying dependencies, you can also generate objects that organically fit into an existing urban environment. It is in this context that architects use the concept of “genotype”, meaning a set of basic parameters, properties, connections that characterize a building or place. Big data analysis and numerical analytical methods are increasingly being used to identify these “genes” and the matrix of interactions between them. For example, in the 1970s, the structure of a city could be simultaneously analyzed according to 2-3 characteristics, and this was cool, and the most modern example of analysis I came across elucidated the patterns of development of about 400 regions according to 25 parameters.

Why are architects generally not satisfied with “normal human” design methods?

Not so long ago, any structure was created and perceived as an integral static object: a residential building is one piece. Now there is a paradigm shift, each building begins to be viewed as a dynamic system, the elements of which are not only material objects, but also invisible ones: connections, associations, points and axes of perception, and so on. The emergence of architectural systems provides a vast field for research, and parametric architecture is some of what emerges as a result.

Abroad, both private firms and specialized laboratories of large universities work in the field of introducing system analysis, automation and algorithmization into design. In Russia there are still only the first shoots, for example, the educational initiative “Branching Point” is engaged in the popularization of computational methods in architecture.

So can a computer think for an architect? Not yet. But the future is already somewhere nearby.

The computer is a rather lazy creature. A kind of fat guy who can only be given free rein to lie on the sofa and think for his own pleasure. The computer is not at all trained to work with its hands - it has no hands! Hundreds of years ago, people dreamed of a completely different computer - a hard-working, obedient one that would take the burden of everyday work off their shoulders...

According to legend, back in the Middle Ages, a great alchemist created an artificial man from clay - the Golem. But the hard-working idol soon lost control and was destroyed by its creator. And in the twentieth century, robots appeared - artificial creatures invented by the Czech writer Karel Capek. The word “robot” itself is Czech, and it simply means a worker, a hard worker. By the way, Capek's robots were human-like creatures, made of flesh and blood. Today they would be called "clones".

Were these creatures computers? Of course not - after all, they were created only for monotonous, routine work, and could not learn anything else. Even the most advanced robots, which today have already moved from science fiction to reality, remain “narrow specialists.”

But nevertheless, it was the advent of the personal computer that opened the way for new robots - more skillful, and most importantly - capable of learning.

In the late 90s, Sony released the first “robot dog”, Aibo. This mechanical creature is difficult to confuse with a real dog... However, it already knows how to play with a person and independently carry out simple commands. It won't be long before robots equipped with computer intelligence will become as integral to our everyday lives as a refrigerator or a vacuum cleaner.

By the way, robotic vacuum cleaners with a computer “on board” are already on sale! They know how to independently navigate around obstacles, distinguish garbage from necessary things, and independently clean your room!

There remains only one, the most important question - when will the computer learn to think? Will he then compete with man?

However, many are convinced that the computer is ALREADY capable of thinking! Or is able to learn this with the help of clever programs.

For example, in the West several programs have been created to create a “virtual interlocutor”. Such programs even compete with each other - every year an authoritative jury selects a winner. The task of the program is to convince a person that he is communicating not with a computer, but with a live interlocutor.

The famous test for determining “artificial intelligence” poses approximately the same task to a computer. It was created long before the advent of today's personal computers by the American mathematician Alan Turing. But since then, not even the most powerful computer has been able to pass the “Turing test”...

So are we on the verge of the emergence of computer intelligence?

But take your time. After all, no matter how “talkative” a computer interlocutor is, he can still utter only those phrases that are stored in his memory by the creator of the program. He cannot yet compose a phrase on his own in response to a person’s question. Namely, the ability not only to choose one action from many options, but also to come up with a new one, not included in the program, is what distinguishes the computer mind from the human one...

Start: 17.08.2010 | Ending: 17.09.2010

Yan David Evgenievich

David Yan, Chairman of the Board of Directors and founder of the ABBYY group of companies, candidate of physical and mathematical sciences, laureate of the Russian Government Prize in the field of science and technology.

In 1989, David Yan, at that time still a 4th year student at MIPT (Moscow Institute of Physics and Technology), founded the Bit Software company together with an employee of the IPTM RAS Alexander Moskalev. In 1998, Bit Software was renamed ABBYY. Along with the main activities within ABBYY. Today, ABBYY is one of the world's leading software developers and service providers in the field of document recognition and input, linguistics and translation. David is involved in a number of other projects. Among them it should be noted:

• Creation of the world's first pocket communication computer for teenagers, Cybiko (Russia, USA, Taiwan, 1998-2003);
• Participation in ATAPY Software (2001);
• Founding and participation in the work of the company iiko, read "Aiko", creating a new generation system for managing restaurants and services in the hospitality industry (2005);
• Participation in a number of creative projects, such as the workshop FAQ-Café (2004), restaurant, club, gallery ArteFAQ (2007); creative club "Squat" (2009); club "Sisters Grimm" (2009);
• Participation in charitable and educational projects, such as the Ayb educational foundation (Ayb, 2005), the MIPT Supervisory Board, the Tumo educational center (2006) and a number of other projects.

Questions and answers:

Question:

Dmitriy
David Evgenievich, in your opinion or in the opinion of your experts, on the basis of what system-forming principle does the brain build abstract (generalized) images (models) of reality? If it builds at all.

Question:

Anton
S. V. Savelyev, in his interview on the same site 4 years ago, regarding AI, stated that AI will not be created, because: [i] "... because there are no real grounds for creating artificial intelligence. And what is called a neurocomputer is an absolute fake, an invention of people involved in technical sciences who have absolutely no idea of ​​either the principle of the brain or the principles of thinking. The brain is a morphogenetically active system. This means that connections between neurons change all the time throughout life. since we have 150 billion neurons, then in order to even poorly simulate this system, we need to take 150 billion processors, inside which we put people who will randomly and continuously solder the connections between them. At the same time, we still need to know some of the regularities of this. resoldering. People who are building “pseudo-neurocomputers” build them on algorithms adopted in mathematics, that is, on those concepts that are casuistry for the brain. Therefore, there will be no next system of evolution in the form of artificial intelligence, because intelligence, those. thinking is built on completely different principles." Do you agree with his opinion, and if not, then where is Savelyev’s mistake?

Answer:

Yan David Evgenievich

Neurocomputer technologies really have nothing to do with the concept of “artificial intelligence”. Neurocomputers are a beautiful name for a purely statistical, mathematical idea, while artificial intelligence includes technologies related to putting forward a hypothesis and having a structural model about the world around us.

As for the term “artificial intelligence,” scientists actually mean two things by it. This is, firstly, a modeling of thinking. Secondly, technologies that replace humans in certain areas of their activity (this is machine translation, object classification, recognition of three-dimensional scenes, images, etc.). These technologies have nothing to do with modeling thinking. They are usually built on questions of structural information about the world. There are a number of approaches in this direction, the task of which is to replace a person in a number of his activities.

Do I agree with Savelyev that in the near future a person will not be able to simulate human thinking using computer processors? I think yes. Recent research shows that the human brain and its impulses are not entirely digital in nature. This is not a binary system (1 and 0), in some cases it resembles an analog system, which is why digital technology may fundamentally not be able to simulate the brain.

What to do in this direction? First of all, you don’t have to imagine yourself as the Lord God. And you need to understand: the fact that we could not simulate a bird or a dragonfly does not mean that we have not learned to move through the air. After all, man has created the most complex technical systems (hang gliders, helicopters, airplanes, rockets), thanks to which we can fly. To paraphrase Savelyev, it turns out that since we cannot model the brain in the form in which it exists in humans, we will never learn to perform complex intellectual work to a certain extent. This is, of course, not true. Already today, intelligent computer systems are replacing people, performing their intellectual work in certain areas more accurately and quickly.

Question:

Sergey
Is it really necessary to teach a computer to think? Is this so that people don’t think, but just press buttons like complete idiots? Is this what we should strive for?

Question:

Vladimir
Hello! Please tell us what prospects you see for the development of AI and robotics in the world and in Russia. Thank you!

Question:

Ushcheko Vyacheslav
I'm interested in fuzzy logic when building artificial intelligence circuits. Do you think fuzzy logic can be used to create new theories?

Question:

Gleb
Is it possible for you to work remotely? for example from Riga? or do you have branches in Latvia? :)

Question:

Ruslan
I have several questions: 1. How closely do artificial intelligence specialists interact with biologists who study the brain? 2. Is it currently possible to take a biological brain and scan the entire architecture of neurons and interneuron connections layer by layer? 3. Is it possible to then reproduce it in the form of a microcircuit or simulate it in software, reproducing the entire number of neurons and connections between them? 4. In your opinion, what path should we expect to take to solve the problem of artificial intelligence - on the path of studying and copying the human brain, or on the path of improving computers and computer programs on our own? 5. Is there conscious opposition to the creation of artificial intelligence on the part of states or other structures for ethical reasons or for fear of losing control over this product?

Answer:

Yan David Evgenievich

I am not a specialist in the field of brain research and do not study the problem of modeling thinking, so my answer to this question will be of an amateurish nature. But, nevertheless, I am absolutely sure that modeling the activity of the human brain at the physical level will not give any practical results for a long time. The reason is the colossal complexity of the technical system called the “human brain”. At the same time, attempts to simulate thinking using computer processors will also not be able to produce serious results due to the significant simplification of this system.

In my opinion, experiments that combine these approaches are interesting and promising. The first positive results are already known, when self-learning chips are created on the basis of biological systems of a population grown on silicon, reacting to external stimuli and making simple decisions. It seems to me that revolutionary discoveries await humanity in this direction in the very near future.

Question:

iq
Why make a machine think? The machine must work. If you teach her to think, she will force people to work for her.

Answer:

Yan David Evgenievich

Iq, brilliant pun! Let's take it into service! But, seriously, we all understand that the word “work” no longer has the same meaning as it did 100 years ago. If the main resource in the era of industrialization was material raw materials, then today, in the era of knowledge, it becomes information. Converting information into useful knowledge is the work that hundreds of millions of people around the world do. This is where the main help is needed today, and modern computer systems help a person solve this problem.

And the fact that a machine will force a person to work for it is rather a far-fetched fear by science fiction writers. Artificial intelligence is a very specific term that combines a number of technologies related to the automatic classification of objects, decision making, and so on.

These technologies have long been used in the household appliances around us - from washing machines to air conditioners. So, there is no need to confuse artificial intelligence technologies with artificial thinking technologies.

Question:

Alexander Prusakov
Dear David Evgenievich, I think many are grateful for your support (along with the Russian Foundation for Basic Research) of the Twelfth National Conference on Artificial Intelligence, which will be held in Tver. Many are probably grateful for the formula for happiness, the body painting festival or a delicious lunch for 169 rubles. But it so happened that we live in a moment of little meaningful reform of fading Russian science. Here your opinion is interesting, and your word is significant! This year, funding from the Russian Foundation for Basic Research is being significantly reduced; resources are being redirected to less transparent and efficient structures. Are there any thoughts about the effective organization of government support for the sciences of Artificial Intelligence?

Answer:

Yan David Evgenievich

Our company invests significant amounts in research and development in the field of artificial intelligence. Also, at the Faculty of Innovation and High Technologies (FIHT) of MIPT in 2006, our department of “Image Recognition and Text Processing” was opened, where students study not only Computer Science, but also some specialized subjects in the field of artificial intelligence.

We support the Dialogue conference, which annually brings together the world's leading experts in the field of computational linguistics. There are plans to organize and support other events, conferences and competitions in the field of artificial intelligence, linguistics and computer science. We are now at the stage of selecting target projects.

We, as a commercial organization, cannot take on the functions of the state. In my opinion, the only correct way to support research in the field of artificial intelligence is the cooperation of companies similar to ours and the state in the field of education and science.

A new computer program can understand what a person is thinking in an instant.

The study authors can predict what a person is currently seeing based on the activity of neurons, read using electrodes implanted in the brain. Scientists have discovered that decoding the thoughts of a person who first sees an image occurs within a fraction of a second.

The new findings could one day help mute patients or people who have trouble communicating to express themselves. This was stated by a neurologist from the University of Washington in Seattle, Rajesh Rao. "Clinically, this means that a mechanism can be created to communicate with paralyzed patients, stroke survivors, and others who are 'trapped inside,'" Rao said.

Mind Reading

In recent years, scientists have made significant strides in deciphering human thoughts. So, in 2011, researchers were able to translate the electroencephalogram into a video that people were watching at the time. In 2014, two scientists exchanged thoughts using brains connected together. Other studies have shown that computers can “see” our dreams by using brain activity to analyze them.

Rao and his colleagues also wanted to be in trend. They asked people with severe cases of epilepsy who had electrodes implanted in their temporal lobes to participate in a thought decoding experiment. (The patients had electrodes implanted in their temporal lobes just a week ago so that doctors could determine the location of the source of the seizures).

“They would have had electrodes implanted anyway, so we just gave them a few extra tasks during their stay in the hospital. There's nothing for them to do here anyway,” said study author Dr. Jeff Augemann, a neurosurgeon at the University of Washington Medical Center. The temporal lobes are also responsible for sensory perception, such as visualization and recognition of images that a person sees.

Rao, Ojeman and their colleagues had study participants watch rapidly changing pictures on a screen. These were images of faces and houses. In addition, there were several blank images. Patients were instructed to watch for an image of an upside-down house to appear.

The electrodes were connected to a program that could analyze thousands of brain signals per second, determining which signal meant that a person was looking at a house and which signal meant that a person was looking at a face. For the first two-thirds of the images, the computer determined that “this is what the brain signal looks like when someone looks at a picture of a house.” For the remaining one-third, the computer was able to determine what exactly the person was looking at with a 96% probability. Scientists reported this on January 21 in an article in the journal PLOS Computational Biology. Interestingly, the computer was able to complete the task in that 20 millisecond period of time while the person was looking at the image.

Difficult process

It turned out that depending on what a person is looking at - a face or an image of a house - different groups of neurons are activated. In addition, it turned out that the computer needs two types of brain signals to decipher: event-related potential and broadband spectrum changes. The first is a characteristic wave on the electroencephalogram, indicating the brain’s reaction to any stimulus. The second means a change in electrical activity in a certain area of ​​the brain.

“Traditionally, scientists study individual neurons,” says Rao. - Our research gave a more general picture at the level of large neural networks. We were able to see how a waking person perceives a complex visual object.”

By allowing researchers to determine in real time which part of the brain responds to a particular stimulus, the new technique has opened the door to a comprehensive mapping of human brain activity.