“We create a monster with our own hands.” What will happen when AI becomes smarter than us?

On May 3, 1997, in New York, on the 35th floor of the Equitable Building, a chess match began between world champion Garry Kasparovand the Deep Blue supercomputer. It was their second duel, the grandmaster won the first one. But this time the car turned out to be stronger: the match, which consisted of six games, ended on May 11 with a score of 3½ : 2½ in favor of Deep Blue, whose name, by the way, can be translated as “deep sadness”.

It was a historic event. And not so much in the world of chess as in technological development. Exactly a quarter of a century ago, a computer program first defeated a person, and from that moment it became clear that the capabilities of artificial intelligence will only grow, and fears that sooner or later it will get out of control and seize power over people , got louder and louder.

A key industry 4.0 technology

We didn't notice it, but artificial intelligence (AI) has already become an integral part of our lives. This happened with the advent of the era of smartphones. Mobile devices have become our second self, without which we cannot exist, and many of the services installed in work with AI support. This is a navigator that plots the best route, and voice assistant, and "smart" camera, and recognition system.

With every year, there will be more and more such services, and not on smartphones, but and in our life in general. Financial transactions, shopping, security — all this will be automated and filled with algorithms. After all, humanity is entering the era of the fourth industrial revolution (it is also called industry 4.0), and artificial intelligence is one of its key technologies.

There is a canonical definition of AI: these are programs and hardware components that are able to learn and are based on complex mathematical solutions. But if we interpret the concept more broadly, then AI — this is everything that thinks instead of a person, freeing his neurons from mental activity and implementing a similar process on a bit-based basis inside computing systems. Experts predict that by the beginning of the next decade, the world will become almost completely automated and controlled by intelligent machines — they will track our every move, create and reveal social prejudices and much more. And the ubiquitous AI will be present in virtually         — self-driving cars, personalized medicine, retail and insurance…

The manner of the game was described as alien

But back to the Deep Blue supercomputer. After him, the AlphaGo program picked up the baton of rivalry with a person in intellectual games. It was written specifically to teach the game of Go, which is considered more difficult than chess. Only in unlike Deep Blue the program did not use the algorithms and suggested in advance and evaluation functions that help to win, but just relied on the rules of the game and its theory.

Analyzing played by professionals, AlphaGo was constantly learning and reached such a high level that it was able to defeat the three-time European champion Fan Hui, and one of the strongest players in the world (in go there is no single official ranking ) Lee Sedolya. This happened in March 2016 

“When the computer first beat the world chess champion, everyone shuddered: it turns out that the machine defeated one of the strongest human minds, — Tatyana Chernigovskaya, a well-known neurolinguist, director of the Institute for Cognitive Research at St. Petersburg State University, shared her impressionsspeaking at last year's marathon “New Knowledge”. — Then they began to say: chess — this is, of course, a difficult game, but here the game is more difficult, it will not succumb to artificial intelligence. But and she succumbed! Lee Sedol, the go champion, was smashed by the AlphaGo computer program.

“height=”425″ width=”640″ style=”display: block;” />

Fan Hui. Photo: Commons.wikimedia.org

And        the case did not end there. Since then, intelligent self-learning programs like AlphaGo and AlphaZero have taken almost all frontiers — in addition to chess and go, they learned to play shogi (Japanese board game. — Note. ed.) and even in poker. It's not the game, it is the fact that they are advancing on our territory. Those things that, as it used to be thought, only a person can do, it turns out that a machine can do.

I was deeply shocked that the AlphaZero program beat another program due to its “deep intuition”. She was able to go through far fewer positions per second than her opponent, and still won, because she used the so-called holistic strategy and   artificial intuition“. That is, it turned out to be smarter and smarter, despite the fact that it worked more slowly. The creators of this program were amazed. They described her playing style as alien: such moves had never been seen in human or other programs.

Thus, computer programs bridged the gap between logic and intuition, jumped over this barrier. Intuition has always been considered a trait unique to humans. But it turned out that the machine can beat us on this field».

Man's biggest enemy

Continuous learning — one of the  principles of artificial intelligence. The system analyzes data arrays, due to which it grows and develops. This can be compared with ontogenesis (individual development of an organism), but scientists go even further and offer something similar to phylogenesis, that is the development of a species.

The University of Edinburgh has developed an AI that can evolve in accordance with Darwin's theory. He knows how to adapt to changing conditions and transmit new skills (or knowledge) to next generations, selecting those “descendants” who are more likely to survive. It's about robots. AI will “reproduce” by producing them without any human intervention — with the 3D printer it is connected to. The developers want to achieve complete autonomy in the life cycle of such robots, and suggest that they will find applications where a person cannot reach — on other planets and on the ocean floor.

Now experts are talking about that the appearance of the so-called "strong" AI. In fact, he will be able to think and be aware of himself as a separate person. And          exceeds man in the ability to to extreme conditions, but    in intelligence.

“Whom has Homo sapiens struggled with” throughout its history? With intelligent systems! With predators, enemies external and internal, intruders, criminals, ill-wishers, etc. etc., — arguesa specialist in the field of adaptive control and behavior, chief researcher at the A. Lebedev Institute of Fine Mechanics and Computer Engineering of the Russian Academy of Sciences Alexander Zhdanov. — The smarter they are, the scarier and more dangerous. The biggest enemy of a man, with whom he fights all his life — intelligent systems!

Until a person sees around him no one who would be much smarter than him. Therefore, he feels like the master of the world. But imagine that an intelligent being appears, which is a thousand, million times smarter! It is known that a person at one moment is able to control about seven parameters (plus or minus two), and this creature will control, say, 70 thousand parameters. And at the same time, have access to all personal information of each person and make decisions based on them. Tell me, will you have at least one chance to deceive him, to defeat him? None. This system will always outplay you. And now we are creating such a monster with our own hands».

Machine morality

There is another important point. When people are at war with each other, they can always hope for indulgence, no matter how evil and cruel the enemy may be. But       the case with           it is not necessary to count on this at all. After all, he does not have the usual human morality, he has it machine-made. It is not known by what criteria he will make decisions and whether they will meet our ideas of good and evil.

British expert in artificial intelligence Stuart Russell gives such an example. As is known, in recent years, AI has been widely used for the diagnosis and treatment of oncological diseases. “Don't expect him to be about to solve these deadly diseases.” On the contrary, AI will most likely find a way to transplant cancer cells to all of humanity in order to perform millions of experiments at the same time, using all of us for its training as guinea pigs, — says the expert.   A all because he will strive to solve the problem that we set before him, but at the same time, we forgot to clarify that you cannot conduct experiments on people».

Rate the material

Источник aif.ru

Leave a Reply

Your email address will not be published. Required fields are marked *