The Human Brain Project: You asked - we answer

    Source: Nature

    Some time ago, a note was published on Habré about the possibilities of 3D SEM microscopy as applied to the study of the structure of the human brain in the framework of the European megaproject “The Human Brain Project”. Under the cut, we tried as detailed as possible - and that means there will be a lot of text - to answer the questions asked, but by tradition we will start with some introduction.
    Attention! Ahead of a lot of text


    As the practice of the previous article and comments on it showed on three resources: HabraHabr , and Dirty . To explain what chemists, physicists, biologists, mathematicians, engineers and programmers are actually doing in the framework of The Human Brain Project (HBP), so I would like to first dwell in more detail on the main goal of the project and the description of 12 subprojects (SP), united in HBP

    The main goal and objectives of HBP

    Let's start with the global one. So, HBP sets as its task neither more nor less, but an understanding of the human brain. And if we can make this dream a reality, we will understand how to treat various diseases of the brain, we will be able to realize what makes us human beings, and we will also create revolutionary computer technologies.

    A brief description of all 12 subprojects (at the end of each video is a list of organizations working on their implementation):

    SP 1 - Strategic Mouse Brain Data - The main goal of this project: to get all the missing basic information about the structure of the brain of mice and to simplify the comparison between the mouse brain and human.

    In fact, part of this subproject for the study of the brain using three-dimensional electron microscopy was disassembled in a previous article . As correctly noted in the comments to the post, mice can solve simple puzzles, for example, purposefully press a button to get food or look for the shortest path in the maze with it.

    SP 2 - Strategic Human Brain Data - The main goal: obtaining a multi-level set of data on the human brain, which is similar to the data obtained for the brain of mice, and simplifying the use of data from the SP 1 project to predict processes in the human brain.

    Thus, the team of this project will fill the void between the microscopic data obtained in project SP 1 and the macroscopic data that can be obtained, for example, using MRI. In particular, it is planned to create a 3D “Google” -map of the brain, each user of which will be able to see where and what areas are located, how signals are transmitted, and so on.

    Habr user vaborg let slip that he was involved in an MRI (SP2) project, so please write to him in more detail about studying the brain using MRI and NMR and the relationship of this subproject with SP1 and SP3.

    SP 3 - Cognitive Architectures - The main goal: a deeper understanding of the cognitive architecture of the brain (sorry for the tautology), that is, how we read, how we recognize things, faces, what motivation is and how it works and so on.

    In other words, these guys are engaged in determining which areas of the brain and how they interact (for example, how visual or auditory information is processed, which parts of the brain are involved in this), that is, this is a classic “top-down” approach, unlike the first two - "bottom-up".

    SP 4 - Mathematical and Theoretical Foundations of Brain Research - The main objective of the subproject: the development of theories, mathematical theories that will form the basis of all further projects to create a model of the brain. In fact, this subproject is about how, knowing the structure of the brain at different levels, to build a single mathematical model, which also includes such an important property of the brain as flexibility.

    SP 5 - Neuroinformatics is one of the main subprojects within the framework of HBP. Its main goal is to organize and make as much accessible as possible for neuroscientists the entire array of various data, knowledge and tools for studying the brain that has been developed by an international team of scientists. In particular, to implement this, a web site is already functioning , on which you can, for example, look at the mouse’s brain and see in which regions of the brain this or that protein is produced.

    More details on how data is collected and processed can be viewed on TED with Russian or English subtitles.

    SP6 - Brain Simulation - The main goal of the subproject: to create such tools that will allow us, based on biological data, to conduct simulation, modeling of individual parts and even the whole brain. It will be an open platform. It is also possible that in modeling it will be possible to take into account the effects and interactions at the molecular level.

    SP7 - High Performance Computing - This subproject aims to provide everyone involved in HBP with wide access to supercomputers, Big Data and cloud technologies, as well as support in the visualization, creation and simulation of multi-level brain models and so on.

    SP8 - Medical Informatics - The main objective of the subproject: to develop methods that will allow the detection of diseases (this will include clinical and genetic data, MRI data collected in hospitals and so on), as well as to understand the symptoms of the disease and how they could be treated on an actually molecular level.

    SP9 - Neuromorphic Computing - Within the framework of this subproject, it is planned to create, introduce into production and develop configurable neuromorphic platforms and chips, inside which will sit the brain models developed in SP6. Each such chip, as conceived by the authors, will work on the principle of signal mixing (in fact, an analog processor) and will contain up to four million neurons with billions of synaptic connections, which should provide a speed of up to 10,000 times faster than an ordinary brain.

    SP10 - Neurorobotics - The goal of the subproject is to create a software and hardware platform that allows scientists to combine an approved model of the brain or its individual parts to the body of the robot to create neuromorphic robot systems.

    SP11 - Applications - this subproject is more interesting, most likely, for science fiction writers and futurologists, because it is in it that the concepts of Future Medicine are developed:

    Neuroscience of the Future:

    Computing Technology of the Future:

    SP12 - Ethics and Society - in fact, this subproject will develop ethical standards, calculate the social consequences and potential risks of the entire HBP.

    Actually answers to questions

    After the publication of the first part of the article, we gave about a week for everyone to be able to express their questions and comments, after which they put everything together and started to cry, were delighted at the incredible amount of questions that came to us. After several sleepless weeks of searching for information, phoning neighbors and acquaintances, we are still ready to provide answers.

    Each question contains the nickname of the person who asked this question, so if you do not want to read the whole thing, you can safely proceed to the desired answer. Also, some of the questions are combined, since the subject and purpose of the question are approximately the same (for example, consciousness vs brain simulation).


    Question from NadezdaSh :
    The goal of the project is to create a single open platform for experiments with the simulation of the functions of the human brain, a single open framework. It will be possible to develop new computer models of emulation and test new methods of treating diseases.

    From what is described as a target, it becomes unclear why a neuron should be emulated in iron, if it can be done programmatically. For the purpose, it is written to create an open framework.

    In games, you can emulate the universe and flights to the stars, in low-level languages, control the movement of bytes. Is it really impossible to emulate a neuron programmatically. Perhaps this is not the goal the project is pursuing ...
    1. Explain the main goal of the project. The goal is to make a new storage device?
    2. How can emulating a neuron cure diseases? The human brain consists of a number of other parts, not just neurons. Besides the fact that there is still a body.

    Similar questions from user Dedushka_shubin:
    1. What is the ultimate goal of the project? What is expected to result from?
    2. (answer 3) What significance will this have for medicine and social sciences?

    1. The goal of the project is exactly what is stated: to create a single database of neural structures and mechanisms that are applicable to them, to be able to simulate various neural connections in particular, and to reach the scale of the human brain as a whole. The simulation of neurons just happens programmatically, after a careful study of what happens to neurons at the real micro and macro levels. All accumulated data will be exactly in bytes, in the structures of the neural network, but at the same time, it is planned to implement them within the framework of SP9 in hardware, in neuromorphic chips.

    2. With the success of the project, this will be a colossal breakthrough in the field of medicine. Now patients with diseases associated with various brain lesions are treated with pills, although they have no idea exactly what they affect. When analyzing neural structures, it will be possible to see the disease “from the inside”, and from this start off, which must be changed at some level in order for a person to become “healthier” (that is, more adequate according to generally accepted norms of society). For example, this video talks about the potential benefits of studying the brain for medicine:

    As for the social sciences, this is already a question of the interaction of different types of neural structures, roughly, different brains. This question is still prematurely asked, and there is no stated goal of the project in this regard.

    If it is interesting, thenin the public domain there is a small simulator of the work of neurons at the membrane level, in it you can build synoptic, ionic connections, process the spike system, and so on.

    3. All signals emanating from our body pass through brain processes. In the brain there is a physical-chemical-electronic medium, which is part of the simulations. Models not only include the neurons themselves and their topology, but also suggest activity outside the neurons in the environment where they are located.

    About the interaction of the brain with muscles, you can watch this video with subtitles or in the original:

    Question from Anc :
    1. In such studies, the main essence is to simulate real neurons. Will someone in the project deal with the interpretation of simulated processes? The first artificial neural networks were based on the assumption of the summation of incoming signals in a neuron. Then spike neural networks appeared. Now neural networks have arisen with various effects on the result of input signals by dendrites or depending on the type of synapse. Your research may well become the basis for new models of artificial neural networks.

    A similar question was asked by waphyld :
    Tell us about the criteria for matching their model for compliance with a functioning brain. If there are any. In other words, how do they understand when to stop?

    2. What structural parts of the brain will be scanned, modeled and distinguished in the design? Only neurons? Myelin? Glia? Blood vessels? The blood-brain barrier in general? Microtubules, neocortex layers, some part of the neocortex, or other parts of the brain?
    3. Will the model be dynamic or static in terms of establishing new connections and dying off old ones? After all, a huge number of synapses are established and torn in the brain during the day of a lived life.

    1. In order to build an adequate model of neural networks, each slice (slice) of the brain undergoes a lengthy and tedious procedure of checks, emulations, and regular checks. That is, to establish a connection between individual neurons, electrodes and sensors are used that are implanted in certain areas of the brain tissue. A certain impulse is applied to these electrodes and the direction of impulse propagation is recorded using sensors.

    Then it is transferred to a specific model and a similar stimulating effect on the model is simulated. If everything works well, then this part is left, if not, then new experiments with brain tissue are carried out. Thus, computer results are compared with the actual response of neurons in the brain. Naturally, for the same areas, the entire procedure is performed several times to obtain statistically valid results.

    2. It is not immediately possible to simulate everything. At first, these will be only neurons, as the main cells and structures involved in the processing of information, and by and large only the neocortex, the latest in terms of evolution of the part of the brain that determines our thought processes.

    As for the simulations, the third or fourth level is the construction of neural networks, where the neuron is considered as a whole, without further scaling, that is, the neuron and certain connections with other neurons - the network. And already on the basis of this, one can easily simulate 75 million neural networks. But this is a completely different topic: such a NEST simulator is called , and it is also part of our Blue Brain Project.

    3. Currently only static links are considered. When the simulations on existing models have reached a certain level of perfection, we can begin to talk about changing relationships. Before looking ahead, you must first establish the legitimate results of the current tests. Now the largest model in the research is calculated within 12.5 milliseconds of real time. I do not think that changes in ties on such a scale are very relevant for consideration.

    Question from whileNotFalse :
    At what level do you model neurons: in detail-physical or abstract-functional? Perhaps there are any analogies and interesting facts about the work of neurons?

    How is the simulation of neurons: first, according to the slices of the mouse (subproject SP1), using the automated complexes of microscopes, topologies of neurons are compiled. In principle, such work can be carried out automatically, but so far it is in its infancy (see the previous article on 3D-reconstruction of the brain). Therefore, now this is done in mixed mode: manually, plus some automatic processing / correction / post-processing. Photographs and the accuracy of such maps are determined by the high resolution of the microscopes. Further, as already noted above, the connection between neurons is checked by applying electrical impulses to the corresponding areas of the brain tissue.

    Secondly, data from other subprojects help to understand what the role of certain brain regions is. As a result, at the output we have a copy of the neural network, which is built as a physical copy of the real brain with the functionality assigned to it.

    Question from esir_pavel :
    I’m worried about this question. The Human Brain project aims to create a model of the whole brain. And then, using this model, what imagination wants to at least simulate neurodegenerative diseases and test drugs for them, at least simulate consciousness.

    Clearly, computational drug testing requires biologically plausible neuron models like Hodgkin-Huxley . But as for consciousness, there is such a hunch that it can be simulated using much simpler models of neurons that do not take into account all sorts of physiological features, such as the phenomenological model of Izhikevich.
    Therefore, my question in a revised version sounds like this: "Are there studies being carried out within the framework of the Human Brain project on the possibility of realizing a strong or at least weakly conscious being based on simple phenomenological non-physiological models of neurons?"

    Also similar questions were asked by kvz :
    1. Is it planned to try to simulate consciousness on the basis of the results of this project, or is this a completely different kind of research? If yes, then from the outside it looks like if we examined magnetic hard disk drives under a microscope and cut the processor with nano-knives in order to understand the principles of Windows XP.

    The question of the human brain is always a question of consciousness. How is consciousness born? Where is the place where the meaning goes into action?

    Is it possible to programmatically simulate the evolution of the emergence of consciousness?
    Is a way of creating artificial intelligence possible through independent software evolution?

    And M_Romul :
    If I’m not mistaken, the HBP said in a declaration of intent that the goal is to simulate a person’s brain (um) WITHOUT a simulation of consciousness. How so? If the human um works entirely, there will be consciousness. Bioethics insisted on this formulation? And yes, will the simulation work in real time? Who is the donor who gave you the brain? His gender \ age \ culture? This will be considered a load of consciousness, when everything works out?

    2. There are two major state-level projects - The Human Brain Project and Blue Brain Project. However, the impression was that all of them are connected primarily with the study of the biological and chemical processes of the brain using all kinds of microscopes and other measuring equipment. Are large grants allocated specifically for modeling consciousness? Scientists in which direction are closer in their specialty to modeling consciousness precisely as an information shell (and not at the level of biological structure) - Computer Science or Neuroscience?

    Long and tedious answer:
    1. Consciousness is a side effect, but by no means an advanced, priority goal of HBP. Consciousness is how to see what is happening on the monitor screen, while not understanding what the filling of the computer is made of - here you are absolutely right. In HBP, roughly speaking, they are engaged in building various systems of the same filling (separate CD, processor, memory), and only then it will be possible to assemble various configurations from them.

    Or another example of how medicine was created. For a long time, people had no idea about DNA, the causes of disease, and so on, but they were treating, cutting, and bloodletting. And now we have access to molecular medicine, when an effective medicine or candidates for them can be calculated on a computer, DNA therapy, growing organs and tissues, as well as cloning. HBP - an attempt to skip the "Middle Ages" in understanding the principles of the brain over 10 years.

    The project is based on the goal of studying the mechanisms of the neuro-physicochemical interaction of neurons between themselves and the environment. This is exactly the foundation that can give answers to fundamental questions about the work of the neurons themselves. The brain is a biological structure, which, in fact, is studied in HBP, and consciousness is a by-product of the evolutionary complication of this biological structure. At some point in time, the brain of our distant ancestors had to start processing all the growing volumes of information, and evolution found an excellent mechanism for this - neocortex columns that actually work like nuclei in our multi-core brain, and then consciousness developed on the basis of this evolutionary tricks.

    A source

    It’s too early to speak about consciousness in this context, because we only want to learn how to simulate brain activity. But, if it, the consciousness, depends only on neural, synaptic and molecular interactions, it will certainly appear as a side effect of the increasingly complex interactions of neurons in large numbers. If so, then we ourselves will witness its occurrence from the moment of simulation of rather complex neural structures that allow us to achieve a certain level of development of brain activity.

    The issue of artificial intelligence (AI) is directly correlated with the issue of consciousness. What people call AI now is an attempt to come up with a formula for pseudo-consciousness, and to fit it into general cases. BBP has a completely opposite task: to create structures, the complication of which will possibly lead to AI simulation to one degree or another, or will give great development to the existing AI subsystems. But this “bottom-up” approach, although much more time consuming, is much more fundamental and correct. We do not use phenomenological constructions for this.

    Of course, the creation of artificial intelligence through the independent evolution of software is exactly what we all expect from HBP, but only after scientists understand the principles of the brain and learn to simulate it even with the simplest examples.

    Moreover, it has not yet been proved that all neural structures will be sufficient for self-consciousness. So ethically this can be easily circumvented using such a statement of the statement. As for how it will actually be, it seems that so many are getting ahead of themselves. No one else knows what they will have to face until models of neurons appear that are close to the real human brain, in which there are 300 billion neurons. Simulation calculations will come closer to real time, and then ahead of them, because no one will refuse faster calculations.

    For simulation, a donor-person, as such, is not needed, since our brain works according to certain algorithms that are 90% identical for all people.
    You can, for example, refer to this lecture :

    What is meant by loading consciousness? Consciousness is aware of itself at some stage: it is either there or not. Be sure that the times when people with banners “do not touch the T2030 model, she realized herself” are not far away, although this is just a simulation, and the line there is very thin.

    And once again, the question of the brain and consciousness is more a philosophical question than a scientific one. Who can answer: do animals have a consciousness if they themselves cannot tell us about it?

    For example, this video claims that if a living creature has a connection between the brain stem and the cerebral cortex, then such an animal has self-consciousness, is able to remember and reproduce patterns:

    As for the meaning, it, the meaning, goes into action all in the same biological structure, the brain, which receives the input signal, "evaluates" it - comprehends it - and gives out some output information, for example, that you need to rearrange the leg from position A to position B.

    And I would like to end this long answer with a TED lecture by one of the project’s ideologists - Henry Markram. It is worth noting that during the whole lecture he never mentioned consciousness - consciousness, only perception - feeling, perception:

    I also advise you to refer to the TED-lecture series “How does my brain works?” I highly recommend watching 1.2 and 6.7 lectures.

    2. Blue Brain Project - a purely Swiss project, which was at one time the impetus for the adoption of The Human Brain Project. The main goal of both projects is to understand how the brain is structured and how it works, the question of consciousness is secondary, since consciousness is the highest degree of brain development. Not understanding how the brain works at the level of biochemistry, we simply are not able to build an adequate model of any consciousness, with the exception of phenomenological ones.

    As far as more is concerned, it is rather Neuroscience with the support and active participation of Computer Science.

    Questions from CompleteBrains sent by mail:
    1. Are there any ways for one person to remotely contribute to the project? Jobs at HBP involve moving to Switzerland, as I understand it?
    Also, in the section it is said about the possibility of cooperation with organizations. Perhaps I can at least join such an organization and, thus, contribute?

    A similar question from hidoba:
    How to volunteer to participate in the project? Even if living for half a year of participation at their own expense.

    2. The most important question: is it possible to start from a developer’s perspective and gain a deeper understanding of the topic and begin to contribute directly to research, and not just to the creation of program models? Is HBP suitable for such a development, or is science and development very separate (because the project is huge, for example)?

    3. There are a lot of subprojects in HBP and therefore it is not easy to single out the general essence. I wonder if HBP aims to learn and recreate the principles of the mind in any form, or only human? That is, purely hypothetically, will it be possible, within the framework of this project, to create the rudiments of a rational intelligence, but different from human, if this is appropriate? Or is the project limited only to what relates to the person, his characteristics and problems? To paraphrase in one sentence: what is more important in the project: to correct the shortcomings of the human mind or create something new based on it? A bit strange question, but still.

    4. What is the situation with the candidates, very fierce competition at the moment? It's just that in some respects I am weak (Linux, English at the intermediate level, not fluent) and in others in relative gain (Game Engines, 3D Graphics, GPGPU). In general, do I have a chance, is it worth it to drop everything and urgently tighten English now? Because something is not entirely clear to me what exactly is meant by “Experienced” or “Expert” in the job description.

    5. If it does not work out now, will it be possible to try again, in a year, two? Are there any subprojects that are scheduled to launch in the next few years? Or at least join affiliate organizations.

    6. And finally, purely practical questions, just in case there will be extra time: Is it likely to use C # (Mono, Interop) in addition to C ++? For example, for a client or prototype model. What is the approximate percentage of tasks that support, refactoring, and improving existing code take? Is the work schedule regular or more flexible? I just don’t know how it is accepted in the scientific community.

    1. The Human Brain Project is a project throughout Europe, including also several institutes in the USA and Japan. To participate, it is not necessary to come to Switzerland, where the Blue Brain Project headquarters is located. It is advisable to resolve these administrative issues with the appropriate administration department or human resources department.

    2. Of course, anything is possible. But development and research are closely related; one cannot without the other. People who came to development positions used to go entirely into research. The same situation as everywhere.

    3. HBP is aimed at the human brain, but, of course, all models of lower brain evolutions, starting with mice, will be used on the way to this goal. As for the “other” reason, different from the human mind, there is no such goal, but quite the opposite - the goal is to correct the problems of existing brain activities.

    4. Why guess? Send a resume for the vacancy, and it will be seen. There are not so many candidates for positions, because it is very difficult to find a good specialist with the right skills. We get resumes from different parts of the world, but for some reason nobody is specifically from Russian-speaking countries. Perhaps some kind of effort needs to be made by our PR team so that the project is more recognizable in countries where they speak Russian. After all, we have a lot of smart and qualified developers. On the other hand, without good English getting into the project will be quite difficult, but it's worth a try.

    5. The project is developing, vacancies are being updated, subprojects are growing. So again - what to guess, send a resume.

    6. Of course, in terms of programming, the project uses various kinds of languages, even those that you don’t particularly hear about among most developers. Support, refactoring - all these are urgent problems of all coders, as in any other company. We are also fighting to improve the code and its optimization. Of course, this is one of the key parts of the project. Otherwise, how can simulate neurons in real time with less memory (and used on lshim number of neurons, respectively). To do this, there is a special team of High Performance Computing - Software Engineers.

    Question from kvz :
    What now looks more promising - to simulate artificial intelligence based on the biological structures of living beings or on a super-computer consisting of classical electronic components?

    What does it mean to simulate artificial intelligence based on biological structures? Grow something containing DNA? This is quite difficult to crank, primarily because of the irreproducibility of one to one biological objects. To do this is much more difficult than, for example, building biomorphic systems in the gland. For example, in the framework of SP9, a project of a neuromorphic chip is created, which will work faster than the biological analog by orders of magnitude. Therefore, it is more promising, spying on Nature, to transfer everything to the existing component base.


    Questions from SAKrisT :
    1. Is it possible now to take a similar 3D image of a part of the brain from a living subject without consequences?
    And if the answer is yes to the first, then the next, maybe a naive question, but still.
    2. Is it possible to track changes in the brain after learning something new? It is very interesting to know how information is stored. I would like to know in more detail and see the changes. I think the scientists themselves are very concerned about this issue.

    1. For this, MRI and NMR tomography are just needed, but this cannot be done using the described method (3D FIB / SEM), since part of the brain tissue is removed by the ion beam, and the tissue itself requires a certain sample preparation and fixation.

    2. Again, this is a question for MRI and functional MRI. Lie down in a tomograph and learn words, for example. Speaking of languages, here people have already done something similar. The sample is true a few people but still. The idea is this: how people learn words and languages. Three groups: bilinguals (i.e., two language environments from birth), children who began to learn the language at an early age and at a later date. Result: a normal brain in the first and second groups of children and an enlarged part on the left, like, somewhere near the cerebellum of the third. We need to find a link to this study ...

    In essence, the question asked: to be honest, it’s hard for me to imagine that we can test it in a living organism at the level of neurons or their rearrangement, as we learn, remember - in vivo , what is called. We can get closer to this or that trick, but the possibilities of science, unfortunately, are limited.

    Question from Nashev (for engine9 comment ):
    “I consider it necessary to mention a project that attracts volunteer forces to build a model of neural connections through the game

    Reference: EyeWire - a project to study the human retina with the help of volunteers. The project was created as a result of cooperation between the Massachusetts Institute of Technology (MIT) and the Institute of Medical Research. Max Planck

    And here is the question in the Blue Brain Project - do they know about this project, and what do they think of it?
    Are you ready to overtake / overtake / support / adopt methods? Do they work more precisely / coarser, etc.?

    A project called “Adopt a Neuron” will soon be launched in BBP, where everyone can participate in the construction of biologically realistic neurons. This is similar to the SETI project, where users who install the application help calculate signals from space. In the same way, everyone will be able to “adopt” a neuron for its construction and analysis in the “screen saver” mode.


    Question from BbIBuX:
    And what are the prospects for building a biological processor?

    Answer: It
    depends on what is meant by a biological processor. If this is a biomorphism, that is, an imitation of living Nature, then the subproject SP9 (neuromorphic chips) is aimed at this. If this is a processor that runs on DNA, RNA or proteins, then this topic is unlikely to be promising.

    In principle, one of the ideologists of the project can answer :

    Full answer in English
    Question 8: What is liquid computing?
    HM: Let me first explain that a Turing machine is a machine that can solve any problem if the problem is given to it in discretely timed batches. So a Turing machine is a universal computer for what is called “batch processing”. But what a Turing machine can't do is to solve problems universally while information is continuously coming in and disturbing it from finishing the operation it just started on. In other words, it can't (without work arounds and cheating) strictly speaking solve problems presented to it on an analog time scale and produce answers on an analog time scale. A liquid computer is however a computer that can solve any problem in realtime and at any time (not discrete time). You can even call it “anytime processing". So it is a universal theory for analog computing. You see, a big problem that the brain has, is to solve how to keep thinking about something that it just saw while the world around it never stops sending it new information. If you sent your computer continuously new information it will not be able to do anything because it can't finish one thing before it has to start on another problem. The way liquid computing works is very much like an actual physical liquid. It makes sense of the perturbations rather than seeing them as a nuisance. We also call it high entropy computing or computing on transient states. This is a very important (but not complete) theory of how the brain works because it shows us how to tap into the vast amount of information that lies in a “surprise”. Another big challenge to understand the brain is that it is always physically changing. Your brain right now is already different from what it was just 1 hour ago, and extremely different from what it was when you were 10 years old. So, because your brain is constantly different and because every moment in your life is potentially (hopefully) also novel, there is a very good chance that most of the time, the responses produced in your brain are new to you (to your neurons) - never “seen” before. So if the brain produces a response that it never “experienced” before, how does it know what it means? The state that your brain is in right now, never happened before so how can your brain make sense of states it never saw before and connect them to all your moments before? Liquid computing provides a partial explanation for this problem by showing that the same state never actually needs to reoccur in the brain for you to make sense of the states - that is why we also call it computing on transient states. Liquid computing can in principle solve any problem instantaneously and keep solving them in real-time and with infinite parallelization. But, it is very difficult to build a good liquid computer. One of the benefits of Blue Brain is that it will be able to design and build extremely powerful liquid computers.

    Simply put, our brain receives information, even when it processes the previous one, the idea of ​​“liquid computing” (computing in transition states) is to build, in a sense, an analog computer that processes information in any given period of time (that is, in real time), not tacts like now.

    Question from father_gorry:
    1. Нельзя говорить о полноценном моделировании мозга, не учтя того, что новые нейроны и нейросвязи образуются в процессе развития, причем в ответ на поступающую с сенсоров информацию. В эксперименте это учтено? Если да, то как именно?

    2. Будет ли этот симулированный малый участок способен создавать новые нейроны и связи в ответ на полученную информацию? Поскольку мозг как фило–, так и онтогенетически — продукт адаптации и развивается адаптационно, эта функция, я полагаю, критически важна для его понимания.

    Reply from user neuroscience:
    You have already climbed far. The adaptive function of the brain, of course, is important, but it is not as critical as understanding the basic principles of the brain. That is, it may be within the framework of HBP that there will be a study of the formation of new neural connections, but so far this is pointless. Think of the brain simply as another organ, like the heart, for example, the most complicated mechanism, but you can create a simulation without taking into account all internal and external factors.

    A crude experiment - “there is a 3D model of all neurons and synapses of the anterior girdle gyrus with prescribed biochemistry. With the introduction of substance A into the synaptic cleft - 6% of the neurons died, the activity of the rest increased. ” And then the laboratory assistant sits and draws conclusions - what was it all about.

    The brain is rebuilt and adapted in response to biochemistry, not information.

    That is, your question is “how to rebuild my brain if they told me“ goat! ”In the tram, the answer is“ no way ”. He will process the information and select something there for himself, and already this selection will jam, develop and completely rebuild the bricks in his head.

    Question from Lukashenko:
    What area of ​​the brain is responsible for shame?

    Ventromedial prefrontal cortex. You do not need a BBP to find out .

    Question from ksotar:
    Is there any initial impulse that triggers brain activity? If so, where does it come from?

    There is no such impulse as there is no, for example, impulse so that a person begins to breathe. As soon as the simplest neural structure appears, it is already functioning. And what decisions are made at the simplest level there - this is a separate issue.

    Question from B0gger:
    1. Will it be possible to consider a verse once learned from the very brain of which you are making a map? Or some emotions, hatred of blacks, for example?

    2. If the brain remembers information by creating a neural network , forming a connection, does this mean that cryonics practice will work? Is the whole personality and memory contained in the connection, or is it encoded somehow, in some tricky microtubules or quantum states of atoms that make up the brain?

    1. So far, unfortunately, this brain (or rather its individual pieces) cannot read. But, in principle, in the future, it is quite possible, because the whole brain will be implemented on a virtual level, connect and watch, read.

    2. Theoretically, this is possible. It’s not necessary to go far for examples . Potentially, proper freezing should not break anything, break the connection between neurons, and so on. And all your memories are just brain biochemistry, which is fixed during cryonic freezing and can be restored.

    Question from sometimes:
    What are the main areas of research? For example, biochemistry, neural network structure, or something else?
    Will the effect of surfactants on consciousness and its mechanisms be considered?

    The main directions are all biological, chemical, physical processes occurring between neurons. Definition of activities and building models with already known mechanisms. Ideally, everything that is known about neurons and their behavior in a certain environment and with certain criteria is taken into account. We don’t know anything

    about surfactants ( psychoactive substances ), but one of the projects is aimed at studying the effect of medications for brain therapy, and surfactants may be studied in this study, since surfactants themselves are nothing more than neuro-mediators .

    Question from KUTKI:
    На каком уровне структуры материи проявляются первые признаки сознания, причем не как психически–социального феномена, а как первичной элементарной субъективной реакции на окружающий мир, разделения внешнего и внутреннего. Есть ли качественный скачек формирования такого сознания у животных, например, оно присуще исключительно высокоразвитым животным, либо животным самой простой нервной системой? Можно ли выделить самую–самую суть, убрав все лишнее?

    Являются ли квантово–механические свойства материи основой для существования сознания, или, теоретически, можно создать абсолютно идентичную эмуляцию системы на будущих супер–компьютерах?

    There have already been many questions about consciousness, as well as answers that remain about the same: so far no one has been able to simulate complex enough neural structures in order to witness the initial appearance of consciousness. And this is provided that the neural structures and processes in them are all that is needed for consciousness. Therefore, talking about it is still very early. For the time being, we can judge consciousness based on available surface data.

    The quantum-mechanical question is unlikely to be considered, because there is nothing “divine” (spiritual, universal) in the consciousness itself, as some have suggested. This is just an idea from the category of misunderstanding that something could be more complicated than human consciousness. And it is necessary, we ourselves take part in it, evolving. Of course, no one denies that consciousness, as it is, will eventually be achievable for simulation. The only question is time and ethics.

    Question from red_ostrich:
    Какова доля учёных из разных областей в Blue Brain Project? Есть ли доля людей, занимающихся теорией, или в основном упор на анализ и описание экспериментальных данных? Кто занимается дизайном экспериментов и как он проходит? На каком уровне нужно изучить биологию, какому-нибудь математику (или физику, или специалисту CS), чтобы участвовать в проекте?

    Most scientists are involved in the theory and construction of models. Although the project itself is primarily neuroscientific, and only then computational. In terms of programming, there is a high-performance computing team, a team of visualizers, and a platform building team (the largest of the computing ones). Some knowledge of neural processes needs to be acquired even by CS guys at the beginning of work, but all this is quickly understood at an abstract level. The main thing is that there should always be a mediator between calculators and theorists, because these parties are not always able to find an equal language for work.

    By the way, a fairly large group of scientists is engaged in neuro-shaped robotics SP10.

    Question from vsh:
    Can you briefly describe the model that you use for the simulation? Tell me, are there things that for some reason she does not take into account, but would you prefer to take into account?

    Unfortunately, we cannot, all models in BBP are under the NDA. Be that as it may, you can read about mechanisms that are part of models without problems ( 1 , 2 , 3 , 4 ).

    Question from X5robot:
    Is it possible to evaluate now how much energy the analogue will be?

    If we mean the energy spent on the simulation itself, then even this parameter is quite difficult to determine (there are estimates in gigawatts of electricity and billions of euros of investments). It’s not entirely fair to take current models and current cars, because energy is now spent on fewer neurons, plus a new generation of supercomputers will consume less electricity, and ideally, the entire human brain would have to be simulated on a 20 megawatt “data center”.


    Question from the Shepherd Evfgrafovich:
    And you can transfer the model of a neuron to a model of any human organization, replacing axons of the type by telegraphic communication. How many people will be required for such a model? Or at least for one synaptic vesicle? Unless, of course, is there any program simulating its behavior? Is there enough for this ordinary sports stadium? And why not, if the ant’s brain consists of just about 250 thousand neurons ...

    A neuron model can be transferred to anything, but why? At the moment, the goal is to study those very neurons and their connections, and then, after building the corresponding mathematical models, they can be transferred to almost anything.

    If we missed something, then you can always refer to the FAQ of the project itself.

    And at the end of this slightly tightened narration, another fun, funny video with TED with subtitles and in the original:

    PS: We understand the text is impressive, it’s difficult to check everything with jewelry accuracy, but if you find a mistake, please write to the PM.

    Sometimes it is possible to read briefly, and sometimes not so much about the news of science and technology on my Telegram channel - we are welcome;)

    Also popular now: