Center for Strategic Assessment and forecasts

Autonomous non-profit organization

Home / Science and Society / Analytical work: the experience of Russian and foreign experts / Other
Almost human: artificial intelligence in factories and fields
Material posted: Publication date: 04-06-2018
The subject of artificial intelligence (AI) in business in recent years has become one of the most debated: experts say the rapid growth of technological applications, especially in Finance and services, and forecast an even more active investment in this direction. Gradually emerging projects and the use of artificial intelligence in industry.

First steps

Examples of creating computational techniques can be found in the XVII century. For example, the German astronomer Wilhelm Shickard have created a clock for calculating astronomical tables, allowing you to add and subtract six digit numbers. Later the French scientist Blaise Pascal made a mechanical device that produced the addition of numbers using spinning wheels, and closer to the end of the seventeenth century Gottfried Leibniz invented the mechanical calculator, capable of subtract, multiply and divide. In the 1830-ies, the London mathematician Charles Babbage built a mechanical model of the difference machine with the familiar rollers and gears along the way, developed the project of a universal digital computing machine, and detail the design and principles of its work, including the data warehouse, the prototype of the future of the processor and operations management with the help of punched cards. According to some, such a revolutionary development Babbage prompted of the incident a few years earlier, meeting with chess "computer", which mathematician lost both games. In the "computer" really was a man, a chess player of sufficiently high class, but I know that Babbage and whether such a meeting actually, the story of an unambiguous answer does not give.

Anyway, these devices can be considered the first, albeit very distant, types of AI. But in the middle of the XX century another English mathematician, Alan Turing, has aimed to find out whether machine to think independently. The result was the famous Turing test suggests the existence of such machines, in communication with which "blind" test can't distinguish them from human. And although the introduction of such a misconception required certain conditions (for example, then the machine could not match the speed of the responses), the idea of creating artificial intelligence has excited the minds of scientists. The entire concept of the possibility of self-learning computer programs, including proposed by the American neuroscientist Frank Rosenblatt a model of an artificial neural network capable of solving problems to learn from the experience and improve.

Enthusiastic expectations, however, quickly gave way to disappointment due to poor capabilities of computer technology of the time and not capable of complex calculations. And only at the beginning of the XXI century, with the advent of huge databases and powerful processors that can handle the data, recent fiction has become reality and artificial intelligence quite successfully replaced the human in many industries and fields.

Weak and strong

Today AI is divided into two conventional types: weak and strong. "Strong" must be able to perform any task, simulating processes of higher nervous activity of man, including emotions. But while this is only a theory. "Weak" is limited to the list of special problems, to solve which allows a specified person algorithm. Although given the booming in recent machine learning increasingly, we are talking about the so-called average intelligence, when the accumulation of data, the program learns to classify them, recognizing images, texts, audio files and any other objects, each time improving and performing the task better. And given the capabilities of the computer to handle huge dataset the application of AI is increasingly much more efficient than human labor and usually cheaper.

First of all, the application of artificial intelligence found in the work with mass consumers: banks, telecoms, trading companies. AI in the form of chat-bot can communicate with the client, answering queries and offering services to gather information and on the basis of its analysis, to identify patterns and preferences, and then to forecast demand or create new offers. For example, Amazon thanks to a similar recommendation engine provides about 35% of sales. Decisions on granting loans, according to representatives of Bank structures in 70% of cases are also accepted on the recommendations of a computer.

Artificial intelligence have long been used to ensure the security of online transactions, successfully identifying application fraud schemes
Photo: GettyImages.com/ visualspace

AI can perform the function of security as PayPal, which compares millions of transactions and identify suspicious. In land transport the artificial intelligence used by, for example, to analyze the current situation on the roads, routing, or predict the time of bus arrival at a specific stop. New technologies also allow companies, regardless of their profile to automate the processing of documentation, hiring staff, analyzing questionnaires, monitoring infrastructure, etc. In the end, with many similar programs, mass clients can meet directly at your own smartphone, including personally interact with an artificial intelligence voice assistant: Siri, OK Google or "Alice".

Infographics: Tatiana Udalova

Data from a survey of representatives of business circles, commissioned by Teradata Corporation, confirm the importance of the development of AI for business: about 80% of large companies around the world are investing in the development of appropriate technologies. According to expert forecasts, the return on each invested today in this area the dollar in the next five years will be just $1.99, and after ten years — $2,87. Maximum effect from investments in AI the majority of respondents (59%) expect in the IT and communications 43% in the commercial and professional services, 32% — in the financial sector and the service sector. Only in the last two years, according to the Russian system integrator "infosistemy Dzhet" and analytical center TAdviser, the number of projects related to artificial intelligence and machine learning, was raised several times: in 2015, large firms reported 17 such projects by mid-2017 — about 162.

AI for refrigerators

The industry traditionally lags behind the more "light" business areas, and yet only a few large projects with the participation of AI belong to the manufacturing sector. That is, appropriate technologies, of course, are used here, but rather in supporting roles. To integrate the AI directly in the production process is solved a few, primarily because of the price issue: the costs of implementation are very significant, and the imperfection of the technology can result in serious losses. So often we are talking more about the remote control processes, where technology is only relative independence. For example, recently, the American Caterpillar presented a draft for the remote control of machinery in mines and quarries, where the work of drivers is associated with increased risk. Instead, the work of bulldozers and trucks will be to follow the operators in the future, with distances of several thousand kilometers, and in case of emergency you can stop the technique. Although it is already extreme case in that the vehicles are equipped with systems with AI that can recognize obstacles and avoid collisions with people and other equipment.

More credible artificial intelligence has demonstrated the South Korean LG Electronics, which last fall announced the construction of a factory of kitchen appliances worth $525 million it is Assumed that all stages of production, from manufacturing and procurement to quality control of finished products, should be managed by a single system based on AI, which in this case will also constantly to optimize the production process. The plant will occupy an area of 336 thousand sq. m by 2023 will produce up to 3 million units per year.

World leader in number of deployments of technologies of artificial intelligence and machine learning, according "information systems jet" and TAdviser, USA. They are followed by United Kingdom that use AI mostly in major investment banks, as well as India, to supply these technologies to foreign customers. The volume of Russian market of artificial intelligence and machine learning in 2017, experts estimate that 700 million rubles, and large industrial projects in which AI can be counted on the fingers.

From oil to steel

One of the already implemented projects on the use of AI in the Russian industry — the introduction of artificial intelligence technologies for steel production in basic oxygen furnace shop of the Magnitogorsk metallurgical combine. Because remelted scrap is typically inhomogeneous in composition, to bring the steel to the required standard it is necessary to introduce it in the smelting process of ferro-alloys and other special additives. Developed Yandex Data Factory service accepts data on the initial composition and mass of the charge (original mix materials loaded into the melting furnace) and taking into account the target parameters of the finished steel gives the operator real-time guidance on the use of additives. Consumption of the latter in the course of experimental heats with the application of new technologies decreased by 5%, and given the relatively high cost of ferro-alloys steelmakers expect to save up to 23 million rubles a month.

The technological leader in oil and gas industry, the company "Gazprom Neft", is implementing several projects involving AI. One of them, "Cognitive geologist" implies the creation of a learning model of geological object. The fact that key decisions on development of deposits it is necessary to take at an early stage of development, and the error made in the beginning of the process in the future to correct is almost impossible. Geologists are trying to restore the data to get reliable picture of the structure of the subsoil and to answer the main question: how cost-effective would be the prey? It takes a year or two, the confidence in the correctness of the answer still does not exceed 60%. "Cognitive geologist" will be mathematically processing the original information to estimate the probability of correctness of the responses and make recommendations on methods of development or the need for additional research. According to specialists of "Gazprom oil" and the interpretation of geological data due to the work of AI will be reduced to six times, and the number of extracted useful information will increase by 30%.

Another project of "Gazprom oil" involves the use of AI in drilling difficult wells. A typical example: oil companies on the basis of a geological model is necessary at a depth of several kilometers to get to a thickness of only two or three meters and to use it as a well over a kilometer, responding quickly to change the configuration of the productive horizon, which are tracked using sensors mounted on the drilling tool. However, the sensors are located at 17 m from the bit, so the experts remotely control the situation from the control Center drilling "Geonavigator" in Saint-Petersburg, learn about the outlet hole of the productive horizon with a delay of 20-30 minutes. During this time, the trajectory of the drilling can go from a three-meter reservoir at a considerable distance.

Control center production company "Gazprom Neft" in Khanty-Mansiysk
Photo courtesy of PJSC "Gazprom Neft"

The solution was found in the trained model, which in real time will draw conclusions about changing conditions at the farthest point wells based on the parameters such as load on the drilling tool, resistance, temperature, vibration and speed of penetration. This will enable the specialists of "Geonavigator" quickly adjust the trajectory of the drilling to Refine the geological model of the Deposit, simultaneously forming additional data for further study "smart" drill. In the future, a mathematical model of the drilling will enable indirect data to proactively predict potential emergencies and to establish the optimal regimes of equipment operation and even real-time to determine the formation productivity, while evaluating the economic efficiency of drilling concrete horizon.

Despite the fact that many of the projects with the AI, especially in industry, are still experimental, analysts believe in the great prospects of this direction. "Infosistemy Dzhet" and TAdviser in their submissions, the forecast growth of the Russian market of artificial intelligence and machine learning to 28 billion rubles by 2020. On a global scale, according to a recent PwC forecast, the use of AI by 2030, will boost world GDP by 14%, or $15.7 trillion. And businesses that ignore these technologies today risk tomorrow to be simply uncompetitive. Moreover, if you believe the predictions of some futurists, in the area 2030-2050 years, the expected breakthrough in the field of creation of "strong" artificial intelligence which has exactly nothing to inferior to the human.

Katerina Ovsyanko


RELATED MATERIALS: Science and Society
Возрастное ограничение