Is the danger of robotic progress true or myth?

2024 Author: Seth Attwood | [email protected]. Last modified: 2023-12-16 15:55

When we say that robots will not replace people, because there is nothing human in them, we do not mean at all the exceptional ability of a person to create or act illogically. Someday robots will be able to do that too. But to be afraid of them is simply pointless. Why - explains Andrey Sebrant, Director of Strategic Marketing at Yandex.

How the Tin Woodman became the Terminator

The great writer Arthur Clarke formulated three laws, one of which reads: "Any sufficiently advanced technology is indistinguishable from magic." This formulation accurately describes our attitude towards high technology. But in the media age, with television and Facebook, it becomes more and more difficult to be a magician.

An example of quite a artificial intelligence is the Tin Woodman, with whom Ellie (or Dorothy) was friends and had sweet conversations. At what point and why did he suddenly turn into the Terminator? This is a purely media story: fear sells well - so much so that the issue related to robots has to be included in the title of the lecture.

And this really reflects what is happening in the public mind. Recently, the HSE conducted a survey that shows that the greater the subjectivity of the robot, the more people are afraid that it, as a subject, will do something bad to them. When a robot is just doing some household chores or bringing goods from the store, no one is afraid of it. But when it comes to nurses, medics, educators and self-driving cars, most people argue that they would be extremely uncomfortable with their surroundings. Meanwhile, accident statistics for one million kilometers traveled show that drones are less likely to get into car accidents. Of course, all the same, people will get into road accidents, but they will die less often - 300 thousand people instead of one and a half million. And the million will live on, because the driver was not a drunk person, but an imperfect autopilot.

Why you shouldn't demand explanations from robots

Nobel Prize winner in physics Richard Feynman said that no physicist understands quantum physics. Unfortunately or fortunately, today there are many other areas in which something is happening that a person cannot explain.

It is useless to demand interpretation from robots (why such a decision was made, why the car slowed down, etc.). Moreover, if you look back at our history, it is completely illogical.

For example, acetylsalicylic acid, synthesized in 1853 and registered under the Aspirin trademark at the end of the 19th century, is consumed today in huge quantities - about 120 billion tablets a year. However, its action, associated, for example, with the use in heart disease, was more or less explained only 70 years after it began to be widely used in medicine.

Modern pharmacologists say that no one knows how sophisticated modern medicines for serious illnesses work. I wonder how many people who are afraid to get into a self-driving car will refuse treatment with a drug that saves in 90% of cases, but we know almost nothing about the mechanisms of its action?

So, even in everyday life, we do not understand everything that is happening around us. And it is extremely naive to require robots to explain their actions before widely implementing machine learning. As long as we strive to achieve this from the current algorithms, quantum computers will come, and there will be no hope for understanding at all. Therefore, it is best to learn to accept what you cannot understand. This is not an answer to the question of what the robots will do to us. This is the answer to the question of how not to spend everything you earn on psychoanalysts if robots are next to you.

How to create with artificial intelligence

The next story about coexistence with robots is devoted to an idea that any creative person can understand - how difficult it is to find someone with whom it will be great to create together. The famous Russian artist and art theorist Dmitry Bulatov formulates this in a more harsh form: "The new norm is this: if we want to infect the world with art, we must put an end to our protein chauvinism."

We (at Yandex. - T&P note) began to have fun with the music written by neural networks back in 2017, - the music we created was recognized as an original composer and expert on Scriabin's creativity Maria Chernova. As Ivan Yamshchikov noted, what if the neural network likes to play the same note for four minutes? I think it will not cause anything but laughter ("the script stuck"). And if we assume that this was invented by a person, then a huge number of interpreters will immediately come running, who will begin to explain that this is a deep thought, expressing the idea of a monstrous stagnation in which we live, etc. This is a question of interpreting not the work itself, but the context given to us.

Today, even in the introduction to the mathematical article Music Generation with Variational Recurrent Autoencoder Supported by History, its authors write that tasks involving an intuitive or creative approach have long been considered purely human, but now more and more algorithms are becoming available, and music is just one example. such tasks.

Two years later, we wrote music for the largest musician Yuri Bashmet (a neural network created by Yandex created a piece for viola and orchestra in collaboration with the composer Kuzma Bodrov. - T&P note). When you tell people about this event, they react like this: “Oh, we get it! They say that neural networks cope well with routine tasks, so the composer creates the very melody, the brilliant idea of the piece, and the neural network probably learned to do the rest of the orchestration work”. The opposite is true. Composer Kuzma Bodrov claims that the neural network became its full-fledged co-author and it was she who generated the most difficult thing, the original, which later turned into something more. I would like to always have such a co-author, capable of creating something new and unexpected, without getting tired and without falling into depression.

Neural networks and physicality

In the book of the Strugatskys "Monday starts on Saturday", entities are described that are called doubles: dictation, but who knows how to do it well. […] Real masters can create very complex, multi-program, self-learning takes. " One of the heroes of the novel sent such a take by car instead of another hero. The double led the Moskvich superbly, "swore when he was bitten by mosquitoes and sang with pleasure in chorus." Our "Alice" is not doing this yet, but one more hackathon will start. Smart adaptive systems were described in 1965. Now they already really exist - as duplicates, which are better at sorting out pieces of paper, coming up with new melodies, doing media planning, etc. And this is just the beginning.

In Kevin Kelly's book Inevitably, there is a beautiful phrase: "The most important thinking machines will not be those that can think faster and better than humans, but those that will learn to think in ways that humans can never." It's as if we have been implementing the idea of flight all our lives, creating and improving a bird with wings, only making them larger and using modern materials. The idea of a rocket that will take us through space where wings are useless would simply not have appeared, because it is completely different from where it all started. And this is yet to come - in the meantime, we have great co-authors.

When we talk about artificial intelligence and are afraid that the machine will replace us, we all the time believe that man and intelligence are almost synonyms, some kind of interchangeable essences. This is not true. I will again quote the Strugatskikh: "I am still a man, and the whole animal is not alien to me." Even when, with the help of neural networks, we learn to dance beautifully on screens, this will not make us people who can get a real thrill from dancing. Physicality is as important as intelligence. And so far we do not understand at all how to make an algorithmic something, which, just like us, would not be alien to the whole animal.