Who is to blame for the accident involving an autonomous car? When will we see autonomous KAMAZ on the roads?
Interview with Andrey Chernogorov, CEO of Cognitive technologies
What has changed in the company with his arrival? How is the CEO's day? What innovations are a priority for developers? And also, who is to blame in the case of an accident involving an autonomous car? and when will we see autonomous KAMAZ on the roads? What has changed at Cognitive Technologies since 2015, when you became director?
The company has changed a lot. During this time we have advanced very strongly in the direction of creating artificial intelligence for UAVs. By the time I joined the company, our rather long RnD stage was over and we decided to enter international markets. In fact, all our successes in this direction refer to this period. Today we have already signed 7 contracts with leading western automakers and Tier-1 companies in the automotive market and three agricultural producers, including one from Brazil. From the point of view of the second direction of our business related to procurement solutions, there was also a rather strong reset. The company as a whole has moved from the philosophy of a major IT contractor to the philosophy of the product company. 8 out of the 10 largest state-owned companies conduct electronic trading under our brand:
The cumulative procurement turnover that passes through our decisions now exceeds 10 trillion. rubles per year. Also, a major reorganization of the company was carried out. Changed control system. Introduced new practices. The whole company moved to SCRUM in the middle of last year. Today it is an obvious global trend, it is used by many leading brands, including Google and Apple. New R'n'D centers opened in Innopolis, Kaliningrad and Tyumen.
Please tell us how is the usual day of the CEO of such an innovative company?
Morning, as a rule, begins early with sports. Or boxing training or a long run. At 9 I'm already in the office. By the way, I save a lot of time on moving around Moscow, because I use an electric scooter all the time, almost 7 months a year. It should be understood that the working day of the CEO of such a company as Cognitive does not consist of continuous innovations, there is a lot of quite earthly labor in sales, in marketing, and in the development of client relations. Many operational tasks. In total, we have to interact with all of our subsidiaries, and these are more than 800 people distributed in 6 offices throughout Russia and the Netherlands. During the day, I definitely set aside 15 minutes for the development of emotional intelligence. We use a special set of exercises in abstract painting according to the method of a famous Russian teacher, thinker and artist Elia Belutin for the development of creative consciousness, training of creative possibilities, and in general, context switching. The brilliant head of the New Reality school of the middle of the last century, in his studies, used for emotional shaking up tasks of the type, draw all the colors of the rainbow in one color or depict one person as abstract images, first as an enemy, and then as a friend.
This technique really works. Then my time is about 50% to 50% distributed between work "outside" and work "inside". On the outside - these are business trips, trips, customer relations, sales, and on the inside - very close communication with the team, and here I break the classical principles of management. I do not have “7 people” with whom I communicate - I directly contact the order with 50 people in quite detailed detail, and, in general, about 3 times a week on a variety of issues. And I can not say which of these parts is more important in terms of business development.
Do you yourself participate in the development process?
I am more of a CEO not from technology but from sales. Behind me are business development issues, strategy development. However, every day I participate in technical meetings, support the creative process and solve all the problems that prevent technical teams from moving even faster. But working with talent is mine. I spend it by the method of coaching. We must understand that such people are difficult to teach something. They can not be driven into a certain framework. Therefore, my mission is to help them make a breakthrough that they themselves are 100% capable of.
A question about competencies: here you are conducting hackathons. Recently in September was VisionHack of the 17th year and what is its essence? Why do this? Headhunting new employees, searching for fresh goals or ready-made solutions?
I will say that headhunting is not one of the first tasks. For us, the popularization of the topic and the statement of the place of Russian brains in the revolutionary sphere of AI were more important. And in fact, we showed the world the real level of our technologies. The first 10 places went to the Russian team. And we had 30 teams, including 8 foreign tops, including MIT, Arizona, Cambridge, Chinese and European universities. In addition, we attempted to raise the level of hackathons to a new height altogether. Raise the level of participants, sponsors, mentors and hold a really real tough competition on the mega-relevant topic and with the global goal of improving road safety. In fact, the guys - the participants saved human lives.
As a result, we have the largest hackathon in AI for UAVs. The event was attended by over 150 people. In addition, we have a very high status supervisory board headed by VEB chairman Sergey Gorkov, which gives the participants of the event excellent chances to get not only prestigious work, but also investments for their team. The star expert jury, which was led by Timur Paltashev from AMD, is the global value in the IT market. Now it will become one of the most prestigious annual events in the world.
Cognitive Technologies declares, I have repeatedly found this on the Internet, that it is engaged in import substitution - this is one of the key priorities of your activity. But this label, import substitution, is not something you conformist? What do you mean by this?
All our activity related to import substitution concerns the direction of automation of procurement. And here we are constantly confronted with the blind worship of Western suppliers. Today, the sector of near-state automation largely depends on 3 companies.
SAP, Oracle and Microsoft. This is not a secret, this statistics shows, we did our research, and relied on public data. Such a cult of cargo has developed in Russia - if any large structure is being automated, then these systems are purchased. In general, the share of decisions of foreign companies in the Russian IT market is excessively high - in some cases it reaches 90%. This gives them opportunities for the most powerful lobbying of their own commercial interests on many levels of the vertical of power and society. It is clear that we, like other Russian companies, are competing openly in this regard. But at the same time it is one thing to compete in private facilities, where the purchase of a product is, roughly speaking, the personal decision of a certain shareholder. And another thing is when the Russian budget chooses between giving a contract to a Russian enterprise or a Western one. This is a very disturbing situation. In fact, our and you money that we pay to the budget in the form of taxes and the next step: they are not returned to the economy of the country, giving to Russian enterprises, but, as it were, exported, are taken abroad. By the way, according to ISEM, every year the main foreign players of Oracle, SAP and Microsoft earn about 180 billion rubles from the public sector. And in terms of the level of innovation, Russian solutions are not inferior to their Western counterparts, and in terms of the price factor - an obvious advantage. SAP and Microsoft earn about 180 billion rubles from the public sector. And in terms of the level of innovation, Russian solutions are not inferior to their Western counterparts, and in terms of the price factor - an obvious advantage. SAP and Microsoft earn about 180 billion rubles from the public sector. And in terms of the level of innovation, Russian solutions are not inferior to their Western counterparts, and in terms of the price factor - an obvious advantage.
Tell us about the balance of interests of your products. What products are the most promising for you?
Today, the 4th Industrial Revolution, which is in full swing around the world, is shaping new markets that are being launched by fundamentally new products based on AI, robotics, blockchain technology, big data and others. Therefore, we focus on solutions where we have a unique competence and where we can occupy the market, our niche. In the first place for us is software for unmanned vehicles. This is a market with $ 100 billion in potential and a three-digit growth rate. A huge vacuum formed here after Intel took over the previous leader (MobilEye) for $ 15 billion this year. Now the market is actively looking for a new leader.
What was the priority when choosing the automotive market?
This is a global market. There is such a thing as a market with two zeros, that is, 100+ billions are vehicles. The markets of similar capacity are B2C - the markets of smartphones, computers, and TVs. That is, where a lot of users. In addition, it was a challenge for us to try our technologies on critical tasks, the solution of which will allow humanity to save 1 million 200 thousand human lives per year.
You talked about contracts with automakers and Tier-1. Can you name with whom?
I can not. Contracts under the NDA. Automotive is a conservative industry, where the contracting period reaches 5 years. This means that the decision that the line of cars is completed with this or that product is made 1 time in 5 years. Perhaps from here in the industry a very large number of phobias. I first saw NDA on 200 or more pages. Prior to this, in our application sectors this was not the case. They all said, yes, we work on these, we work with these. Since here it is a zone of know-how, the very fact of working on a specific solution and a specific company is already information for competitors. I can say that these are all global companies with $ 10 + billion turnover.
What about your activity in automotive in Russia?
In Russia, there are no world brands. The country in the global automotive industry, unfortunately, is much less than 1%. Most industry decisions are made in Germany. I do not say that these are German companies, I say that decisions are made there. Because historically, RnD centers and suppliers and manufacturers are located in this place. And so the focus is at this point. In addition, we do not have a legal framework that would allow testing autorobots on public roads. Therefore, we work only on landfills, and in the United States, Japan and Europe on urban highways. But in Russia, we are working with Rostselmash, the only agricultural producer with a strong position in world markets, to create an unmanned combine and with KAMAZ, which is also known for its victories in international rallies, to create an unmanned truck.
Since you yourself have spoken to the account of KAMAZ, then I will continue. You planned to test the unmanned KAMAZ in 2016 on public roads, now the date has shifted to 2018. What is the reason for the delay?
There is no delay. We have never made such a statement. As far as I remember, Arkady Dvorkovich said that drones would be used at the 2018 World Cup. And we have the technological capabilities for this. However, to solve the problem, it is also necessary to conduct detailed testing of the prototype of the autobot in real conditions or as close as possible to real ones. But we know nothing about it. Nobody addressed us.
And 300 million. Ministry of Education, what's with them?
We made a driver assistance system with ADAS 2 functionality, which provides for the operation of collision warning systems, the intersection of the marking line, and the detection of road signs with quality that ensures the necessary traffic safety. We have developed a prototype of a truck that performs a full range of autonomous movement functions and performs a certain amount, if I am not mistaken about 30 exercises at the training ground. Today, our relations with KAMAZ have moved from the RnD stage to the industrial stage, and now we are waiting for what KAMAZ will decide on in terms of sales and marketing policies for these products. In our plans - the creation by 2021-2022 of the version of the driver assistance system corresponding to the level 3 ADS and providing for autonomous movement in certain conditions (in traffic jams, on the motorway, etc.). We are counting on a completely unmanned prototype of KAMAZ, as stated earlier, to be prepared by 2025. But how things will go in the future will depend on the position of the KAMAZ leadership and the position of the state, including the allocation of funds to support the project.
You are talking about a level 3 driver assistance system. Explain what such levels mean? How many are there?
Now, according to the new international classification SAE Inernational, the driver assistance system or the ADS (automated driving system) is already a three-dimensional matrix. This is not just a level 0, 1, 2, 3, 4, 5, but a level under certain restrictions. With certain conditions. For traffic on closed territories, some rules apply, for traffic jams, others, for highways, a third, etc. It clearly states that the level of ADS 0 and 1 is generally a squeaker, hazard warning system, manual control. Level 2 is also not autonomy. The system interferes with management, for example, it brakes in front of an obstacle, keeps the runway. Level 3 is a very important stage. This is autonomy in some conditions. For example, traffic on the highway or in traffic. That is, the driver at speeds up to 20 km / h, while driving in a traffic jam, the car lights the lights, ready to drive itself, the driver can conditionally remove hands from a wheel. But it provides for the possibility of human intervention in a critical situation.
In fact, I must say that no one has yet seen level 3. It will appear as legislation changes. The fourth level is the autonomous movement of a car in certain modes, and the system will be able to control even in critical situations when a person will not be able to interfere with control. Roughly speaking, it does not require interception of the driver’s control, but may pull over. Go to safe mode without the driver. The fifth level is completely autonomous movement. In the car, there is simply no steering as such.
We know firsthand about the advantages of domestic software, but are there any Russian hardware solutions that you can use in your business, say, in robotics?
Please, our video cameras, which we currently use in our test cars, are Russian designs. They also have the best price-quality ratio. In addition, we coordinate the development of the domestic radar of a new generation and assume that in the very near future it may create a bomb-effect on the automotive market.
Will we see these solutions in Russia? What is the problem to move to a greater degree of localization?
There is no mass customer in Russia. Everyone makes only hints.
Beaten question to pain, but you, probably, like no one else, he is known. In the event of an autonomous car hit in an accident. Who is responsible: software developer, hardware developer or driver?
The automaker and solution integrator will be responsible. And if the problem is in the software, we will be responsible to it. It is very important. But in reality, the issue of responsibility, as is customary in Europe, America will be decided at the level of insurance companies.
Hacker Geohot , offered everyone to download the application and collect all the stream that was directed to the training of the neural network. And you use neural networks too. How do you solve this issue?
We use neural networks to select objects. But here we must understand that even if we recognize every object in the road scene, we will not get the meaning of the whole situation. It will not give the sum of the results of object recognition. To learn to understand the meaning, we have developed a truly revolutionary Cognitive low level data fusion technology. We collect data from sensors and from all systems at the bottom level. The man also acts. We kind of collect a neural network over a neural network, which determines the meaning of what is happening on the current video cut. For example, if a tennis ball rolls out onto the road, then this absolutely does not mean that the control system should ignore a small object. With a ball with a high probability a child may jump out. Therefore, you need to carefully "look" in this direction for additional data slow down or take a maneuver. We plan to show this technology at CES in Las Vegas in early 2018. As for the data collection model, the principle of crowdfunding really works here. Two years ago, we threw a call to all motorists who are interested in participating in the creation of a Russian drone and volunteers from across the country, from their DVRs sent thousands of kilometers of real video data, and most importantly, rare emergency situations that we later used to train our neural networks.
And you do not think about working in the field of service robotics?
No, we are trying not to lose focus, we have a unique global competence in the automation of ground moving vehicles. We are on the quality of computer vision system №2 in the world market. Last week, experts along with Google and DeepMind included us in the TOP-10 trendsetters of 2018. And our AI applications are either cars or combines, tractors.
Competition. For the best gift comment from Cognitive Technologies
In the comments write your answer to the question:
- What proofs of reliability and safety of an unmanned vehicle should developers bring you so that you can get into such a car (without a driver) and drive on public roads today?