Researchers from the Institute for the Future of Humanity at Oxford University in the UK and the Political Science Department of Yale University in the US have published findings from a survey of 350 AI experts around the world.
The report says that artificial text translators will surpass humans in 2024, smart trucks will get rid of drivers in 2027, and the need for writers who specialize in fiction will disappear in 2049. Surgery will also only be performed by robots, which will completely replace humans in 2053. But the probability that artificial intelligence will completely replace humans in 45 years is estimated by experts at 50%.
We invited Alexander Kaplan, an expert in natural intelligence, the most famous specialist in the study of the brain in Russia, head of the laboratory of neurophysiology and neurointerfaces of the biological faculty of Lomonosov Moscow State University, to evaluate these conclusions.
Alexander Kaplan / ria.ru
Alexander Yakovlevich! How do you assess the time frame within which artificial intelligence, according to the world's leading experts, will surpass human capabilities?
“I think they are not wrong about translators, trucks and fiction writers. As far as surgeons are concerned, I think quite a lot of surgical procedures will be automated over the next 20 years. All suture procedures, which are now done by the delicate hands of master surgeons, may well be automated. It is possible to measure the thickness of blood vessels, the thickness of the wall of an organ, softness, elasticity - all this is digitized and, naturally, can be transferred to such a mechanical automaton. But where a creative approach is required, when you do not know in advance how you will conduct the operation, how exactly the vessels lie there, it will most likely never be possible for robots to master. In difficult situations, surgeons act intuitively, without understanding which algorithm they are following. Intuitively, it means that they use their own generalized experience. It is impossible to convey such generalizations to the program.
Demonstration of a robotic surgeon at the Cancer Institute of Saint-Herblanc in the west of France / AFP 2017 / Frank Perry
What do experts mean when they say "artificial intelligence"?
Promotional video:
“They mean an algorithm that will automate some of the intellectual actions of a person. When the autopilot is steering the plane, it means that some human actions are automated, but this does not mean that the plane is controlled by full-fledged artificial intelligence. It's just a set of programs that describe how to fly the plane in different situations.
In the same way, many can automate other human actions, for example, a game of chess. But it is impossible to put into the program exactly the way a person does it. However, imitation is obtained. As you can see, on one side there is a person, and on the other side there is a computer in which an algorithm, devoid of creative impulses, operates. At first, the computer program memorized a lot of successful games of the world championships, but now it looks at how the game begins and immediately selects options that develop in a direction favorable to it. A person cannot keep so many games in his head - they simply do not fit into his brain!
Why hasn't nature provided man with a great memory?
- It's a matter of optimization.
It is clear that the body will not last long in this mode. The brain is a very "expensive" organ! And it doesn't matter whose brain - a stupid or an intelligent person … Nerve cells work, spending a lot of energy!
How does a person generally resist a machine that remembers thousands of chess games? After all, sometimes chess players beat the computer ?
“This is the secret of natural intelligence. The fundamental difference between human intelligence and any other artificial system is that the intelligence uses the model of this world, where it is necessary to make decisions. Of course, we cannot put into the brain everything that we see around us, but in our brain we have concepts about everything. We think only in concepts. It is clear that our model of the external world is not complete, since it describes only what we ourselves saw - the part where we had experience. And we make all decisions according to this created model.
Can artificial intelligence grow to the point that it will have its own model of the world?
- In theory, yes. But they have to grow. Get your own experience. Get a model of the surrounding reality.
Why can't you just download Wikipedia?
- “Wikipedia” is a disorganized model, it is just a collection of facts. There is no experience - how does it all add up to a whole.
Can't you describe it?
- Simple models are possible. For example, the Titanic model with all its cabins. But this description is incomparable even with the experience of a 5-year-old child! Not only informational content is immediately added, but emotional contact, experience of personal communication, experience of bodily communication … It will be impossible to describe all this! Even the difference between “fool” and “smart” cannot be described.
Intelligence is a system that contains a model of the external environment. The ant has its own model, we have ours. But a person has his own special model - it is woven from concepts, since a person has acquired a new way of putting information in his head - words.
What is the correct name for the very artificial intelligence that world experts have in mind?
- "Automated assistants for making intelligent decisions." And they are very far from the real intelligence that, perhaps, will ever be grown.
Anna Urmantseva