We live on the very threshold of the era of AI or artificial intelligence, but the contours of the future are already visible. About what AI is and how it develops — in the material of Komsomolskaya Pravda
Artificial intelligence. Photo: pixabay
For the first time, scientists started talking about artificial intelligence (AI) in the early 40s of the last century. In English, it is called Artificial intelligence and the abbreviation AI is used. Today, we increasingly meet this phrase in everyday life. Komsomolskaya Pravda set out to explain in simple terms what artificial intelligence is, what is the essence and technology of this system.
What is artificial intelligence
“You can’t see or feel artificial intelligence. It is a program that takes a sequence of data and analyzes it. The result can be face recognition, text recognition, text generation, and much more, to which the characteristic “template approach” is applicable, — says technical director of Chrono.tech Mikhail Savchenko.
- Artificial intelligence in the broadest sense — a set of methods and models that receive data and draw a certain conclusion based on them. This may be assigning a photo to a certain class, grouping texts on similar topics, or predicting the exchange rate. All models are united by the extraction of knowledge from data. Calculation of a single function that can contain millions and even billions of parameters, says the software Ilya Makarov, director of the MADE Big Data Academy at Mail.ru.
>
>
In essence, AI is a program, like some notorious text editor. But it differs in that it can be trained.
- The program itself is a set of code and a knowledge base that it received in the learning process. Both are just a set of data on the hard drive of some computer, notes Evgeny Karpov, AI product manager at Zyfra.
Experts notice that the concept is often speculated, especially in the marketing of various products.
- We distinguish a more intelligent person by behavior, and the ability to find the right solutions for more complex problems. Similarly, an artificial intelligence system is distinguished not by its special appearance, but by the behavior and tasks that it is able to solve. Developers, of course, can “see” artificial intelligence in the text of the algorithm, explains Alexey Dyachenko, Deputy General Director of Open Technologies.
The essence of artificial intelligence
Like any other computer program, AI works like this: it takes data, analyzes it, and produces a result. But the essence of artificial intelligence lies in its features.
“The structure distinguishes it from a regular program: an artificial neural network, which, after being created by a programmer, goes through a training procedure in order to make the results of data analysis as accurate as possible,” says Alexey Dyachenko.
If we simply explain the essence of AI, then the algorithm has two modes: work and training.
- For example, we have an AI that determines whether a person wears a helmet at the factory or not. To make it work, we show 10,000 photos of people in helmets and say that they are people in helmets, then we show 10,000 people without helmets and say that they are without helmets. After that, we transfer it to the operating mode and show a new photo. And our AI can already determine whether a person is wearing a helmet or not. Thanks to this, we can signal a safety violation and save workers from dangerous injuries,” notes Evgeny Karpov.
But this example does not exhaust the essence of artificial intelligence. One of the most important components of technology is self-learning.
- In addition to the initially laid down, the algorithm is able to receive and process new information in order to make more complex and effective decisions based on it in the future. In addition, learning occurs with each new completed task. Artificial intelligence improves on its own experience. This is called machine learning, explains Director for Innovative Solutions at Oberon Evgeny Ovcharov.
>
>
- AI has advantages — it does not get tired and is able to process a huge amount of information in a short time. It differs from human thinking in that it does not yet have emotions and does not depend on them. Whereas the thinking of man is always subject to them, says IT specialist Alexander Serebryakov.
Development of artificial intelligence technologies
How it all started
If you dig very deep, you need to go back to the 17th century, when Wilhelm Schickard created the first mechanical calculating machine — the calculator. And the philosopher and mathematician Descartes presented animals as a thinking mechanism and thus set the task of human civilization to create its own intelligent mechanism.
First ideas
The concept of an artificial neural network was proposed in 1943 by American scientists — the founder of cybernetics McColloch and mathematician Pitts. A year later, their colleague from Princeton (USA), von Neumann, will offer a variant of the architecture that formed the basis of all modern computers. In 1950, mathematician Alan Turing published his famous test to detect strong AI. When a person cannot understand that he is talking to a machine, but thinks that another person is standing behind the screen, then true artificial intelligence has been born.
Development
In the 60s of the XX century, scientific universities of world powers are actively working on the development of technology. For example, in the USSR it was the cybernetics Pospelov and Tsetlin. And already in the next decade, an explanatory dictionary, reference books and voluminous scientific works devoted to the topic of artificial intelligence as an independent phenomenon were published. In the West, in 1965, they create a robot, Eliza, who could speak English. Today we would call it a chatbot. Four years later, the android Sheki appears, who also knew how to move.
Interview with Eliza. Photo: wikipedia.org
Breakthrough in the development of artificial intelligence in the 80–90s
In the 80s, a qualitatively different era in the development of AI began. There are even more advanced robot consultants that could solve mathematical and some everyday problems, learned even better to keep up a conversation. At the end of the decade, they develop the Deep Thought program, which beats Grandmaster Larsen. Then the chess player Garry Kasparov will say that he does not believe in the capabilities of the machine. They say if they can beat the best chess players, then they will compose the best music and write the best books. The grandmaster announced that on behalf of the human race he would fight in a chess duel with AI. Such an opportunity will be presented to him in 1996 and 1997. Kasparov won the first match and lost the second.
Nowadays
Starting from the “zero” they begin to introduce the development of a smart home. Artificial intelligence is finding applications in the growing PC gaming industry. There are a number of complaints against him, they say he still acts according to patterns, and is not guided by the situation. But over time, developers eliminate this problem, and the machine gradually learns by itself.
AI is actively beginning to be used in urban services. The same definition of faces through cameras. And with the mass character of smartphones and the ability to easily install applications on each mobile phone, artificial intelligence is somehow being introduced into programs.
To understand the scale of the seriousness of the attitude of top government officials to the topic of AI, just look at the decree of the President of Russia, which refers to the creation of a national strategy for the development of artificial intelligence in the country. The plan is up to 2030.
Artificial intelligence control systems
Not all experts accept this formulation. Some believe that instead of control systems, it is more appropriate to speak of algorithms and artificial intelligence services. They are divided into two types.
The special one is well able to solve a narrow area of template tasks. Each needs its own algorithm. This AI will never be able to evolve into a full-fledged assistant or competitor to a person. In robotics, algorithms are only suitable if the machine will perform a narrow task. For example, open and close the door.
General purpose (AGI Artificial general intelligence) — in the future they will acquire cognitive (cognitive) abilities, emotions, the ability to learn, make decisions for a wide range of areas, self-awareness, etc.
>
>
Who develops artificial intelligence
Google
It would be strange if the technology giant who invented the most popular search engine and the most massive mobile OS did not develop artificial intelligence. After all, they can immediately implement their developments in products, which means they will receive profit from this. Different Google APIs — that is, tools for developing programs, are trained in speech recognition, translation from foreign languages, maintaining a template conversation, for example, in a call center. Of course, these are the most understandable developments for a simple user. Although the company has hundreds of AI projects in one area or another. Most, of course, are related to online sales.
opencog
The company created a cognitive architecture, in simple terms, pieces of computer code that are aimed at creating artificial intelligence. The program is open, that is, it can be used by any developer. The company believes that by joint efforts the community will quickly come to the creation of a full-fledged AI that will be strong not only in its field, but in general in all.
Chinese Institute of Artificial Brain
Scientists from the Celestial Empire have its laboratory to create a “brain-like” artificial intelligence. Officially, they are engaged in neural networks capable of recognizing objects, human messages or creating AI, which, in a second after an accident, signals an incident to emergency services. But one of the most ambitious developments of Asian scientists is the system of social trust.
Microsoft
The most popular AI from an American corporation is called Azure. His firm sells to corporations that develop software. The neural network understands speech, is able to make predictions and imitates other human intellectual abilities. The company makes a special emphasis on AI, which is trained to recognize errors in the code. And this means that soon AI will be taught to create programs on its own.
How to connect phone to TV
The most trouble-free ways to connect your phone to your TV when you have a USB cable at hand
Russian players
— In Russia, there are no normally functioning AI centers of excellence integrated with business. There are a number of companies that are successfully promoting AI-based solutions in their business models, such as Sberbank and Mail.ru Group, Yandex and Kaspersky Lab, Sibur Corporation and Severstal. But the solutions are not unique in many ways. The representation of Russian scientists and AI developers at leading international conferences is extremely small compared to scientists from the USA, China and Europe, Ilya Makarov believes.
Popular questions and answers
Why are more companies using artificial intelligence?
- Most analysts predict that the introduction of AI will increase the efficiency of the company by tens of percent. In business, a huge number of mistakes happen due to the human factor. AI will increase the accuracy of operational processes, including when working with documents and when entering data. There are already and are actively used neural networks that recognize, for example, passport data and enter data into the corresponding cells of the system. Artificial intelligence is becoming a key element in increasing the competitiveness of both companies and entire states in the era of industrial revolution 4.0, says Director of the IT Institute of the Synergy University Stanislav Kosarev.
Where is artificial intelligence applied?
At the request of KP, ABBYY Russia CEO Dmitry Shushkin told in what areas AI is used today.
in banks. Technologies allow opening accounts for people and companies with little or no participation of employees, including remotely. Based on information about the borrower, determine the size and suitable terms of the loan, analyze news about the bank’s partners and identify risks, and effectively trade on the stock exchange.
In retail. They help to understand the demographic characteristics of the buyer and offer the most suitable products. Control the filling of shelves, understand the reasons for queues in the store, better organize delivery and purchase from suppliers, simplify the work of accounting.
In telecom. Predicts the outflow of subscribers, effectively distributes calls to call centers.
In logistics. It helps to build routes for freight transportation, control fuel costs, extract and analyze data from transport documents.
Music services. They use machine learning methods to analyze the user’s musical tastes.
In medicine. Canadian company BlueDot uses AI to track the spread of infectious diseases. They warned about pneumonia in Wuhan a week before the announcement of the coronavirus epidemic. AI trained by Sberbank has learned to detect characteristic blackouts in the lungs caused by coronavirus infection. Scientists from the leading American university MIT used AI to create new antibiotics.
How much does artificial intelligence cost?
- It is impossible to give a specific answer to this question: it all depends on the type of business, the scope of the tasks being solved. In short, it’s expensive. So far, only large businesses can afford it. As AI-related solutions develop, as new models are trained and data is enriched, the cost of implementation and use will decrease,” says Stanislav Kosarev.
There are many freely distributed libraries that implement artificial intelligence algorithms that a programmer can simply download, adapt to his data and use in his software product. There are also ready-made products, for example, the Comma One project is an open source autopilot for a car that you can simply download. Of course, there are also closed systems. So, Google annually spends billions of dollars annually on the development of artificial intelligence systems, while, if the company protects search algorithms, then some secondary libraries are published in the public domain.
When will AI fully enter our lives?
- If we are talking about replacing a person with a formed and intelligent person who can become a friend, then this should not be expected in the foreseeable future. Otherwise, daily life, services and communication — AI has already penetrated tightly into these areas and regularly helps us out. Algorithms work where we no longer notice them: from building the optimal route by a navigator to personal recommendations from a music service, says Evgeny Ovcharov.
Can you recommend books, films, games that most plausibly show artificial intelligence?
- All the Terminator series, where the Skynet system, which has an independent life, acts as a source of danger, and a person is already superfluous in this world and is fighting for salvation. I also recommend the trilogy “The Matrix”, “Zoey”, “Blade Runner. 2049”. The best movie in my opinion is Lucy. This is science fiction, where the main emphasis is not on the power of artificial intelligence, but on human intelligence, on its unlimited abilities. This is like a signal that humanity should develop itself, and not machines that will replace them together and in everything,” says innovative technology expertVioletta Galician.
— There are few books in Russian on AI. I would recommend the work of Sergey Nikolenko “Deep Learning”. But to be in the subject, you need to actively read articles in English from leading conferences. Well, if you want to immerse yourself in the world of fantasy related to AI and its perception, I advise you to check out the video games from the Mass Effect and Deus Ex series, as well as the films Artificial Intelligence, I, Robot, and others. For those who want to become a part of history and create AI themselves — the MADE Academy of Big Data, advises Ilya Makarov.
- The best option is a course from Yandex and MIPT on AI technologies. If you pass it, you will understand what AI is very well. Movies, books and games are full of fantasies that are very far from reality, notes Evgeny Karpov.
Can AI pose a danger to humans?
- A significant part of the worst fears about artificial intelligence is unfounded. Its development depends on the person and the criteria set by him for assessing the surrounding world and events. That is, a scenario in which AI independently “acquires” the desire to harm people is simply impossible. By itself, the machine will not even understand what harm is and why it is necessary to inflict it, AI does not have such a component as motivation and desire,” says Managing Director of Applied Intelligence Accenture Practice in RussiaLarisa Malkova.
According to the expert, another category of fears — “robots and programs will replace everyone” — is also greatly overestimated. The labor market will change, but not dramatically. The interaction model “Man + Machine” will become even wider.
What is the difference between artificial intelligence and neural networks?
- AI solutions can work on different algorithms, approximately how a car can work on a gasoline, diesel or electric engine. One of these algorithms is neural networks. Now they have become very popular, explains Evgeny Karpov.
- In a separate category, we can single out the use of AI based on neural networks for video image analysis. Such intelligent video analytics is able to solve the most complex tasks. The number of potential applications of this technology for security and smart video surveillance is truly enormous. From the detection by video analytics systems of unusual or aggressive behavior aimed at harming people, to the recognition of individuals even in large crowds. And even the automatic detection of the notorious “bomb in a suitcase” — the implementation of baggage screening without the need for X‑ray installations, — adds DSSL Leading Specialist Mikhail Sharubin.
What is machine learning?
“This is the principle of the evolution of artificial intelligence, which is based on self-learning algorithms,” says Violetta Galitskaya. — The necessary information and goals are loaded into the machine’s memory. There are three types of machine learning:
1) with the participation of a person when he corrects the goal;
2) without human intervention. Robotic technology independently, by comparing patterns, is looking for a more optimal solution, thereby honing analysis skills;
3) mixed learning process. The main feature: the processing of a large amount of information with the participation of neural networks.
