Evgeny Burnaev: Will we create a terminator some day? Maybe.

What is machine learning? How do you handle Big Data? Can a piece of software help to quickly detect a fault in a complex engineering system? We caught up with Professor Evgeny Burnaev, head of research group for Advanced Data Analytics in Science and Engineering at Skoltech’s Center for Computational and Data-Intensive Science and Engineering (CDISE), to get answers to these questions.

In 2017 Evgeny won the prestigious award for young scientists established by the Moscow Government. The awards ceremony took place in the Kremlin on February 5, 2018.

— You work a lot with Big Data analysis and processing. Would you call what you do a profession of the future?

— I’m very wary about statements like this. Over the course of history, the best minds took a fancy to different fields of knowledge. Nuclear physics and space exploration that once fascinated mankind have become an industry just like any other. Globally, volumes of information are increasing, enabling a wealth of useful solutions, such as automation and optimization of processes and production facilities. I can’t be certain whether the demand for these solutions will increase exponentially, but I’m positive that it will grow.

— Is the demand for your expertise growing too?

— Yes it is, and it does so very fast. We are working hard on creating training courses for Skoltech. There are a lot of industry customers waiting for experts to come and explain what Big Data, Machine Learning and Neural Networks are about. Everyone seems to be talking about Big Data, but most have a vague idea about what it means.

— What skills will your graduates have?

— We expect to train managers and engineers capable of running a project with a machine learning component in the right way and distinguishing between the tasks that are or aren’t worth solving using machine learning methods. When you discuss a project with a customer, you don’t only walk them through the statement of work but also educate them by explaining the fundamentals of data analysis. It would be great instead to deal with managers and engineers who are already familiar with data analysis methods and know how to use them.

We can teach frontier disciplines, such as deep learning and neural networks, blockchain, wireless sensor networks for monitoring applications and the Internet of Things, Earth remote sensing data processing, etc. that a modern engineer should be familiar with and that are used in multiple applications.

Besides, our Center has a second area of research mathematical modeling which is a “complement” to the data analysis and machine learning technologies.

— Looking back, how did you discover your true purpose? What did you dream of as a kid?

— Like any other kid, I dreamt of many different things. I had an aptitude for sciences, including both mathematics and physics. It was my father, a MEPhI graduate, a physicist and an expert in high-speed chemical processes who steered me in the right direction. My grandmother who had graduated from MSU Chemistry Department also helped me a lot with choosing an academic path. I attended a regular secondary school which fortunately had an outstanding math teacher, T. Lepekhina.

— Are you a Muscovite?

It so happened that I finished secondary school in Volgograd, although I was born in Moscow and am living here now. Again, it was my math teacher who suggested I apply to MIPT. At the time, her son and some of her former students studied there. I met one of them once, and he talked about MIPT and I thought I should give it a try. What looked tempting is that the department I selected offered a full-fledged math program including advanced courses in mathematics (at MSU Mechanics and Mathematics Department level) and courses in general physics.

I took part in MIPT Olympiad where I scored the number of points required for admission. Out of pure curiosity, I also applied to MSU Physics Department and got admitted too. I even thought of applying to MSU Mechanics and Mathematics Department where I was asked to provide some additional documents before I could take exams. I thought it was too much of a bother and gave up. After all, I was already enrolled at MIPT, so I decided to stay there.

— What other universities can offer education of the same standard as MIPT?

— When I studied at MIPT, there were few universities in Moscow which offered programs in applied mathematics with a data analysis component aside from pure mathematics or engineering. Actually, I had to educate myself. After completing general courses at MIPT, I started studying wavelet-transform signal analysis.  It was a new scientific topic, so at that time I was not able to find a research supervisor for my bachelor studies. I am very grateful to Senior Researcher at the RAS Computing Center Nikolay Olenev and Dean of the Applied Mathematics and Control Department at MIPT Alexander Shananin who were extremely helpful with organizational matters. I continued working on my own and was eventually awarded the bachelor’s degree.

As time went by, it became obvious that my knowledge was not sufficient, so I joined specialized courses in probability and random processes at MSU Mechanics and Mathematics Department. These two topics form the core of the signal analysis methods. In the long run, I realized that mathematical statistics, data analysis and related applications were the right fit for me. It’s at MSU that I met my future PhD supervisor, Full Member of RAS Albert Shiryaev.   

When I became a PhD student, professor Shiryaev suggested I focus on the so-called quickest disorder detection problem which amounts to observing a random process in real time and detecting the moment its random properties change. Currently, the problem is of high relevance, especially for designers of complex engineering systems where timely detection of a change in behavior is crucial for preventing a future failure or malfunction.

You might have heard about the “industrial Internet of Things” which is a popular topic, because hardware and software systems are growing increasingly complex. For example, airplanes have a lot of systems and assemblies and a delay in replacing any of them results in noticeable financial losses, let alone the security issue. Delayed delivery of a part or a malfunction in a heat power plant, an aircraft or an IT service leads to financial and reputational losses.  Businesses are looking for solutions to prevent such failures.

Modern hardware systems are fitted with sensors to measure various physical parameters, such as vibration, temperature, pressure, etc. If something goes wrong in the system, the data gives an indication of an imminent breakdown or failure. Analyzing the signal behavior helps to capture these indications.

In my PhD thesis I looked into theoretical aspects of the quickest disorder detection problem. Soon I got my PhD degree and started working with Professor Alexander Bernstein, at the time Head of Laboratory at the RAS Institute for System Analysis, and his team. We were faced with a new research task in a joint project with Airbus.

The project was more complex than mere analysis of time series and signals that I focused on in my PhD studies. It involved predictive modeling and construction of models that could predict dependencies of some parameters, for example, aircraft wing lift, from other parameters, such as wing geometry and flight mode.

Initially, predictive modelling relied on mathematical models based on the ‘first principles of physics’ and describing complex physical processes and phenomena by partial differential equations with boundary conditions. To solve PDE, one should apply complex and time-consuming methods, both in terms of computations and preparation of input data and computational meshes, so these models were really difficult to use in designing complex objects. This poses an obvious difficulty at the early design phase when a large variety of solutions has to be considered and mistakes are too costly.

However, if simulations and experiments (for example, wind tunnel tests) yield enough data to compile a database, it can be used to construct the so called meta-model (or surrogate model) using data analysis and machine learning methods.

Meta-models are known to have much higher computational efficiency as compared to initial data sources; this allows engineers to sort through and compare a much larger number of design solutions before selecting the best fits.

Another example of using meta-models in practice is predictive maintenance of production systems and economic processes typically described by a huge number of parameters (production of sugar from sugar beets, automated adjustment of the loan allocation process parameters, etc.)

For example, the efficiency of the granulated sugar extraction (diffusion) process depends on the shape of beet slices, beet quality, sugar content, pH, beet slices and water temperatures, temperature distribution in the diffuser and so on. Understandably, production parameters should be set to optimal values to minimize costs and reduce losses, but even a skilled and experienced process engineer is unable to keep track of multiple control actions, environmental conditions, their interrelations and impact on the production efficiency.

Meta-models based on extensive operational data and integrated in the recommender platforms make it much easier for process engineers to select the production process control parameters (by providing a ‘second opinion’) and eventually help to reduce costs and enhance product quality.

It transpired later that predictive modeling was closely linked to the subject matter of my thesis. The point is that creating a meta-model helps to identify interrelations between system parameters in normal operating conditions, while comparing the revealed interrelations to the real-time telemetry data provided by sensors helps to detect anomalies and disorders in the system operation. Thus, a combination of predictive modeling and fast disorder detection methods is instrumental in designing effective predictive maintenance applications.

Predictive maintenance methods ensure fast detection of anomalies and disorders in the operation of mechanisms and complex engineering systems, such as the aircraft’s auxiliary power unit. The core idea behind predictive maintenance is to adjust the operating condition of a mechanism or make a replacement before the detected changes become critical for proper operation of the system or any of its components. In recent years, the intelligent predictive maintenance methods have made significant progress in terms of their diagnostic capabilities thanks to advancement of sensor surveillance technologies and innovative data processing algorithms.

In summary, I am working on the methods that help to deal with the challenges described above and their industry-oriented applications.

— You were among the key people who helped Airbus to speed up their production processes.

— Calling me a key person would be incorrect and unfair towards my colleagues and supervisors; for example, I owe many of my task definitions to Full Member of RAS, Professor Alexander Kuleshov. The truth is that after getting my PhD degree I worked as head of the intellectual data analysis and predictive modeling laboratory at the RAS Institute for Information Transmission Problems (IITP RAS). After we successfully completed a series of projects for Airbus, IITP established its spin-off Datadvance company where, as head of Datadvance’s data analysis group, I introduced a number of surrogate modelling and optimization methods in industrial practice. Later industrial implementations of these methods formed the basis of pSeven Core software library (formerly MACROS) and were used in various engineering projects for IHI, SAFT, Airbus, Astrium and other customers. pSeven was used to deal with the aircraft design modeling and optimization tasks and, according to Airbus engineers, helped to reduce the amount of modeling at the early design phase by 10%.

— How does pSeven work and what can you use it for?

— Imagine that you need to optimize the structure of a composite plate of Formula-1 race car that protects the racer from side crashes. There is a set of parameters which determines the thickness and type of the plate’s various layers. A specific plate design corresponds to a specific set of parameters. In order to measure the plate’s strength, one can manufacture a prototype and conduct tests, but this would take a lot of time and money.

Another way to do it is to perform mathematical modeling for a given plate structure and assess its strength by applying time-consuming calculations which are unlikely to produce a highly accurate description of the real composite-strain-related processes.

The engineer’s task in this case amounts to finding a plate design such that the plate has the smallest possible mass, while its strength meets the specified safety requirements. The biggest challenge is that even an engineer with ample expertise in the subject domain would hardly be able to track the behavior of more than two or three parameters and their impact on the target performances, which are mass and strength in our case. 

If we increase the value of one parameter and decrease the value of another, how will these changes affect the product properties? If the options are too many, an engineer cannot check all of them, because each would require a real experiment or physical modeling which is time-consuming and expensive. It is with the help of a computer and data analysis that an engineer would grope for an effective combination of parameters that corresponds to the best product design.

— How important are the programming skills for your work? Do you do programming?

— I don’t do much programming simply because I don’t have time for it. I did that in the past of course.

— What languages did you program in?

— As a scientifically-minded person, I started off by doing computational experiments in Matlab, but around 2008 I switched to Python in my data analysis research. I’ve got a team with whom I discuss the software functionality and explain the underlying algorithms and the way to implement them.

As regards industrial solutions, we dealt with pSeven Core in two steps: first we performed a huge series of computational experiments and developed a prototype of the algorithm in Python and then a specifically trained software engineer at the spin-off company created its computationally effective implementation in C++. We then worked together with that engineer to embed the algorithm in the library and write the software documentation.

Now I’m back to what I think is the most important and exciting part of my work, that is academic studies rather than applied research, although I still work a lot on industrial applications and collaborate with various companies. On the one hand, they come up with tasks that help me validate my methods, and on the other, industrial applications are a potential source of new problem statements in data analysis, which is in perfect accord with Skoltech’s strategy that strongly encourages cooperation with the industry.

A handshake away from Kolmogorov

— Your research supervisor Academician A. Shiryaev studied under Kolmogorov. Do you have a feeling that you are a handshake away from Kolmogorov?

— I communicate a lot with Academician Shiryaev and I visited him more than once in Komarovka (the house in a settlement near Korolyev, former Kolmogorov’s dacha). I read Kolmogorov’s diaries that Shiryaev prepared and had published. I read some background material too. I suppose I was influenced to a certain extent by what I learnt about Kolmogorov who was a truly outstanding personality. He was both a pioneer in and a contributor to various branches of mathematics. Besides, some of his works were of clearly applied nature. Kolmogorov is a brilliant example to be followed.

— In your lecture at PostNauka you talked about Kolmogorov’s complexity.

— It is one of the pivotal concepts that helps you better understand probability and randomness.

— Can a good scientist be a good manager? Is that an innate gift or can you acquire managerial skills by learning?

I would say I have to be a manager whether I like it or not. With some research tasks, you simply work in your office looking for a solution on your own. Luckily or unluckily, I do research that results in algorithms that have to be verified with real data and then applied in real industrial applications. By contrast, to perform an applied project involving construction of a real information system, you do need a research team, so I had no choice other than build one, and manage it too.

I dare to hope that I am coping quite well with my management functions. I don’t believe this is something you can learn: you are either able or unable to do this. People are all different and each person has their own notion of the way things should be done. Of course I have to explain formally to each team member what they should do, but aside from that I sometimes have to be a psychologist and talk to my team about things other than business. A good manager will see right away if something is off about an employee and will help them to step back into the ranks, so that the team could work in harmony and make progress.

— What is the main secret behind good management?

  You should talk to your team and all the team members should perceive you as a leader. If I say the task should be addressed in a certain way but cannot explain why and cannot have the team follow me, the team will fall apart right away. A good leader treats their team as humans with their own strengths and weaknesses and never looks at them as cogs in the wheel. I met and worked with managers of all sorts and, sadly, some take a very formal attitude towards their collaborators, which in the long run results in team breakups.

— Is there anything you don’t like about your work?

— Perhaps the paperwork you have to put together when it comes to applying for grants, writing specifications or reports.

— And what are you most excited about?

— Mathematical research, new challenges and communications with my colleagues. I enjoy going to conferences where I can meet new people from my area of expertise. It’s always exciting to learn new things and share ideas. I find this most enjoyable and stimulating.

 

Skoltech is open to new ideas

— Speaking about Skoltech, what advantages and disadvantages can you see in the project?

— The core idea behind Skoltech is quite sensible.  Skoltech has about a dozen research centers working in various research areas, such as advanced production technologies and materials, bioinformatics, oil and gas, etc. I will speak about the Center for Computational and Data-Intensive Science and Engineering that I represent. We have three key performance indicators (KPI). The first is related to teaching and involves developing and giving lecture courses.

Skoltech runs an MSc program offering advanced courses on a specific topic and not basic disciplines like mathematical analysis. For example, my lecture courses deal with advanced methods of machine learning and Bayesian methods of machine learning. These courses certainly contain a basic component explaining the classical models and methods, but the bigger part is concerned with the latest scientific advances and discoveries, as well as new knowledge and experience obtained in working with the industry. Along with lecturing, I supervise students working on their graduate theses.

The second KPI evaluates the number of papers published in peer-reviewed journals and the number of globally recognized specialized conferences attended.

Finally, the third KPI is used to assess the industrial outreach. Skoltech has a project office that helps us to find new knowledge-intensive industrial projects with a pressing need for advanced mathematics and IT technologies. We do not do “web programming”, but readily get involved in projects which require high-level skills in mathematics, computer science, data analysis and machine learning.

If a project results in a new marketable product, we may consider manufacturing it on an industrial scale.

When it comes to launching industrial production, we should work through application-related aspects to do this, we bring in partner companies which can take care of the engineering and technical part. As a professor of Skoltech, I take responsibility for the scientific and methodological component by developing mathematical algorithms, establishing a problem statement and translating the customer wishes into the language of mathematics and engineering. Once the algorithm is developed and tested, one can start working on the commercial product.

At Skoltech, you can benefit from seed investment to build a product prototype. You can also set up collaboration with other companies, establish a start-up company or have the technology patented. Skoltech strongly encourages this, while the Skolkovo system provides all the tools you need.

— How difficult is it to conduct basic and applied research at the same time? Doesn’t one contradict the other?

— It certainly would if I did nothing but prove theorems. Since in my area of research I can produce a product design on paper and have it put into production in a matter of days, there is no contradiction at all. Of course I need both the engineering resources and the people who know how to draw up a contract properly, taking account of all the legal and economic details.

— It has been claimed for years that Russian science lacks a link between fundamental research and real-life applications. Do you think Skoltech helps to bridge the gap?

— Depends on a discipline. In the case of theoretical mathematics, there may be no applications whatsoever, but theoretical mathematics does prove useful, for it sets a certain standard and produces highly-skilled experts capable of solving intractable problems.  It provides valuable expertise and a powerful impetus for other sciences and applications that would hardly survive without such input.  As for other sciences, like branches of experimental physics, biology, etc. these may have multiple applications. In my opinion, building a complete chain from pure science to a marketable product should be easy enough, if you take the trouble to do it.

— Is it something you can learn? Do you teach your students to do it?

— I personally I don’t, because I have a specified set of disciplines that I am responsible for. As for Skoltech, it offers specialized courses where students are introduced to the fundamentals of innovation and learn to create prototypes, establish start-ups and build a team, develop leadership qualities, draw up project plans and specifications.

I’ve got a wealth of hands-on experience in these matters, but for students such training is indispensable. Every project (and a development project in particular) needs a plan. Once the work starts, the team should get together every week to update the plan, monitor the progress and write an action list. Any attempt to keep everything in your memory will fail, even if you have a handful of people on the team.

Unfortunately, young people aren’t always aware of this and can’t always select the right tools and working procedures. The best undertakings and ideas could go down the drain simply because project leaders have failed to build the right team or process. At Skoltech, students get hands-on knowledge and project management skills.

Of course, students can master this on their own if they wish, but they will dramatically increase their chances of success if they complete a specialized course explaining what tools and methods they can use and how. At Skoltech, we see to it that the educational program is continuously improved to provide students with useful knowledge to take on board.

— Looking at the bigger picture, what will Skoltech’s future be like? How far ahead can your plan your activities? (if it’s true that financing will be available till 2022).

— To be honest, I haven’t really given it much thought, but looking at the salient trends in our professional community, I get the feeling that everything should be OK. We are busy opening new research centers, building new teams, working closely with the industry and actively pursuing academic activities. Scientists are applying for grants and subsidies to secure additional funding for academic work. The most enthusiastic are ready to leap into action.

Evgeny Burnaev. Photo: Skoltech.

— What is your Center working on? How many research groups do you have?

— Generally speaking, our research is focused on two areas: simulation based on ‘the first principles of physics’ and simulation based on data from different sources and making use of applied statistics, data analysis and machine learning methods. CDISE employees, and my group in particular, have gained broad experience in the second area.

As I said, we have worked a lot on engineering tasks and, for example, built models that allow predicting aerodynamic characteristics of the aircraft wing with a new surface geometry using data from various sources, including full-scale wind tunnel tests and computational experiments with physical models. To perform the simulation, we apply advanced machine learning tools, such as transfer learning (a technique that helps to increase the model learning speed and efficiency using the results obtained in other data-based learning cases under similar conditions) and domain adaptation (a way to enhance the model’s prediction accuracy in handling data other than the learning sample).

Another good example is the prediction of crop yields using Earth remote sensing data which is scarce for Russia but exists in abundance for Canada where the climate is very similar to ours; what I mean of course is detailed information about crops, the fields they grow on, soil condition, etc., and not just satellite imagery.  This data can be used to create a predictive model and calibrate it using a small set of Russian data so that the model could be used to predict crop yields in Russia. That’s something our agriculture really needs.

We can see some interesting tasks in the oil and gas sector which offers broad opportunities in terms of efficiency-enhancing applications.

Our center also looks into various aspects of the Internet of Things, develops data analysis approaches in relation of soil condition analysis, looks for new chemical compounds and their potential applications in medicine, and other issues of immediate concern.

Soil data analysis projects are the responsibility of Ivan Oseledets’s team who are working on a joint project with Rusagro company. The soil scientists go out to test fields, pick soil samples and look at how fertilizers can change the soil composition and properties. Their research findings are of high value for raising the land fertility and productivity.

New chemical compounds are designed so as to optimize their solubility and toxic level and tailor them for a specific purpose. Such tasks lie at the crossroads of biology and medicine, on the one hand, and computational mathematics, on the other hand. Researchers have to process huge databases containing both real experimental data and computation results obtained using labor-intensive physical models which are used to assess the properties of chemical compounds. Machine learning is quite instrumental in making this assessment much faster.

— Do you communicate with other teams? Do you organize joint conferences? And how do you exchange information?

— We work at the same center and have offices on the same floor. I can drop by a colleague’s office any time to discuss various tasks and approaches. We also have a few collaboration projects. For example, we work with Ivan Oseledets on kernel methods of machine learning: Ivan is a recognized expert in computational maths who knows how to accelerate matrix calculations by applying various approximations, while I understand how to adapt such approaches in order to construct more effective kernel methods.

We work with CDISE Director Maxim Fedorov on chemoinformatics; we also collaborate with Victor Lempitsky, a renowned expert in computer vision, to apply generative models for image segmentation and 3D-image processing a research area of high importance for medicine, image recognition, design of self-driving cars and so on.

I personally work a lot on methods that facilitate fast detection of changes in the properties of big data flows. These methods can be used to deal with cyber security and hardware failure prediction tasks, amongst other things. Experts from the oil and gas industry have come to realize the importance of this kind of methods, so we are closely cooperating with them now.

President of RAS Alexander Sergeev (left) and winner of the Moscow Government award for young scientists Evgeny Burnaev (right) pictured during the awards ceremony, February 5, 2018.

— How do you decide who can be admitted to Skoltech’s PhD and MSc programs and who cannot? What is your choice driven by?

— Applicants are supposed to pass exams in mathematics and English and an interview. With this filtering, we select the best candidates. If a postgraduate student asks if I could be their supervisor, I look at their graduate paper and ask them about their research interests and short-term plans.

I look for good MSc and PhD students. “Good” means having solid mathematical background that you would get at leading universities, such as MIPT or MSU Mechanics&Mathematics and Physics departments. They should not only be proficient in maths but also capable of setting up a computational experiment in the right way. Globally, I first look at what the candidate is capable of and not where they studied before.

In my research group, PhD students got their bachelor’s and/or master’s degrees from universities based all over Russia. One of my PhD students graduated with a bachelor’s degree from Saratov State University. He has been demonstrating excellent research results. Recently, he has presented his research paper at the front-line SIGIR conference in the USA.

Students enter PhD programs for different reasons. Some wish to pursue scientific research where a scientific degree is a must. Others choose to acquire solid knowledge and hands-on skills in the selected specialty, learn to use machine learning methods, develop and solve new non-routine tasks and eventually make a good career in industry or business.

— Where do your students do an internship?

— My colleagues and I are now actively engaging PhD students and postdocs in the Center’s applied projects which serve as a testing ground where they discover what real applied projects are like and what they should do to solve an industrial task.

Between their first and second year, students spend two summer months working with companies on science-intensive projects. We are expanding the range of interested companies which can offer real tasks that students can work on and that pertain to their field of study. Skoltech pays a monetary allowance to its students during their two-month industrial immersion so that they could fully concentrate on their project. Of course we closely monitor their internship, making sure that they are working on real knowledge-intensive assignments and are not just pushing papers.

— Do you encourage your graduates to do PhD studies at Western universities and is there actually any need for that? And where do they write their PhD theses?

— They write their theses in Russia under supervision of a Skoltech center’s professor and in some cases may have a co-supervisor from a Russian or Western university. Recently one of my PhD graduates did an internship at Philips where he worked on object detection with the use of thermal cameras.

In his project, regular and thermal cameras were placed in a room. In poor lighting, data from the thermal camera can be helpful in increasing the detection accuracy. My student was supposed to develop an algorithm that could improve detection accuracy using data from both cameras. He ended up producing an excellent paper and a useful industrial application.

Another intern who worked at another unit of that company developed algorithms that allow detecting anomalies in the data coming from users of the company’s products. He did a good job and is now preparing a research paper for publication.

Skoltech encourages such industrial contacts which help students in building a broader perspective.

— Have you ever thought of going to work abroad?

— At different points in my life I had different ideas. Luckily, I have an exciting job in Russia that gives me an opportunity to travel worldwide. Recently, I went to the USA twice in a month’s time    first as a speaker at the Artificial Intelligence and Statistics conference (AISTATS)  and then to attend a workshop dedicated to the memory of E. Braverman, one of the fathers of machine learning and an eminent scientist who had worked for the RAS Institute of Control Sciences.

Cyberman fantasy or reality?

— Is there a chance of creating a cyberman or artificial brain with the same or higher intelligence level than human intellect?

— I don’t think so. At least not at this point. Interestingly, the term “artificial intelligence” means different things in English and Russian. In Russian it commonly signifies alien intelligence, like Skynet or Terminator, whereas in English it is used to refer to computer programs that perform a limited set of intellectual functions following a given algorithm.

There are a lot of applications requiring fast actions beyond the human speed limit. In this sense, software programs offer an excellent replacement for human operators. I guess we will soon use intelligent systems to a much broader extent than today: we’ll have more AI-powered assistants to take care of routine operations and give us useful advice.

Imagine that by the end of the day a physician gets too tired and misses a shadow in the patient’s lung. It’s in situations like this that the software can give a prompt. Clearly, these functions are very useful but they are not what we mean by the term “artificial intelligence”. This type of systems will definitely continue to develop.  Whether an anthropomorphic robot indistinguishable from a human being becomes a reality is a big question. But you never know…

— Do you believe in that a digital system can be embedded in the human brain?

— Maybe. But again, you’ll have to grapple with a bunch of problems from legislative and ethical issues to purely technical obstacles: how should this digital piece be designed to be implanted in a human head? Are you sure you will find a lot of people eager to have something sewn into their head? Let me take another example today we’ve got all we need in terms of technology and engineering to build an unmanned airplane. In a manned aircraft, piloting as such takes about a minute and could easily be done by the autopilot. However, I don’t believe passengers will get excited about travelling on an airplane with no one in the cockpit.  Having a piece of steel in your head doesn’t seem like a lot of fun either. At least not until people surmount some mental barrier… I find it hard to make any predictions.

— Have you fulfilled yourself as a scientist in Russia? Do you think you could have done better if you had worked in Silicon Valley?

— You are asking about what my life could have been like if… But there are no ifs in history. Right now I don’t have any regrets about having missed something. If I had worked abroad, possibly, the ratio between my academic activities and industrial projects would have been different. Anyway, in Russia, there are a lot of opportunities for growth and fulfilment which, curiously enough, are not so easy to find in the well-established Western system.

Original article By Nataliya Demina.

 

Contact information:
Skoltech Communications
+7 (495) 280 14 81

Tweet about this on Twitter0Share on Facebook0Pin on Pinterest0Share on Tumblr0Share on VK