The limits of artificial intelligence

Basilio Sierra is a professor at the Faculty of Informatics of the UPV/EHU and a member of the Board of Directors of the Spanish Association of Artificial Intelligence. it reminds us that what we call “artificial intelligence” is only data analysis.

Computers have only one advantage over the human brain: they are much faster in performing calculations. It’s a unique advantage, but very important. In fact, thanks to this, artificial intelligence is more than just the ability to process. Machines can’t learn the way we learn, but they can learn statistically. Through artificial intelligence, the strategy can be programmed.

And so computers, "dumb" computers, today can do several things that have an "appearance" of intelligence: diagnose diseases, for example, or play chess - and win the great masters.

BASILIO SIERRA; UPV/EHU: In artificial intelligence, in general and especially in all the programs used for games, there is a strategy, and above all there is the so-called "search": what might be the movements of the enemy and, as a consequence, what will be our best movements. However, it should be noted that most chess programs are based on large databases and the speed of computers. The computer called Deep Blue, which beat Kasparov, became very famous, but that, as it is said in computer science, was "purely by force". He didn't beat him for his intelligence.

The power of artificial intelligence becomes more evident in cases where it is necessary to use large amounts of data at the same time. This is very difficult for humans. But on the other hand, the problem posed to the machine must always be very specific. Machines don’t know how to use ambiguous information like we do.

That is why, for the moment, the most sold robots are those used to perform repetitive tasks; tasks that are boring for the human being, as well as tasks that are dangerous, fast and/or require a lot of effort.

However, the main challenge of artificial intelligence in the field of robots today is to achieve an increasingly autonomous movement.

BASILIO SIERRA; UPV/EHU: It is true that much work is being done in Europe, especially, and in Japan, in social robotics or social robotics, especially to help older people. These machines - the robot in this case - have to decide if a person wants a coffee when they enter the house - with or without asking. And then maybe he'll give you a coffee without asking for it. Or he sees that he's getting tired, and he's going to bring his chair closer. By analyzing the expression of the person to make such decisions, this is quite close.

They will make our lives easier, yes; but the challenges of achieving this are not too small: giving machines mobility like ours or creating a vision like ours are very difficult goals. In fact, much more difficult than originally thought.

BASILIO SIERRA; UPV/EHU: We don’t know how our mind works. We know that in mathematics computers are faster than we are at making calculations. But we always tell the computers what calculations they need to do. However, it is true that it has not yet been possible to make an analysis of the images that are received by a camera through a computer - although very fast and very good today. In what is known as computer vision, much remains to be done to reach a child’s level of perception. It's missing a lot.

Machines receive and use huge amounts of data very quickly to achieve a goal, but they don’t know what they’re working for, and they don’t even know they’re working. Machines are not autonomous.

Research lines for the creation of autonomous robots have been launched, but they are far from the goal. And if a robot could ever develop intelligence, it would not be of the same kind as our intelligence.

BASILIO SIERRA; UPV/EHU: It has recently been published in a very famous magazine that an expert has said that 50 years from now humans will marry robots, for example. I don't think so, and I'd be sad if that happened. What you won't get is for the robot to feel emotions. Maybe to say yes, as the actors do: express their satisfaction, even though they are sad. That's another matter. But not real emotions. That's not gonna happen.

Emotions arose as a result of evolution in order to develop interactions between the viscera, senses and movement in organisms of the complex nervous system.

So if one day it were possible to create a machine with emotion, after all, it would be to create a living one. And that, more than robotics, is another matter.