He knows many tasks better than humans. Meet the smartest cat in the whole world.
Deepmind, a Google company specializing in artificial intelligence, recently presented its model called Gato (cat in Portuguese). This so-called generalist AI would be able to perform 600 different tasks. And on most of these tasks, the AI would perform better than a human.
First of all, Deepmind would have built the first generalist artificial intelligence, that is to say capable of learning several tasks at the same time, where most AIs are trained for a specific purpose. Since the American company presented its new work, it has caused reactions from computer scientists around the world. What are the limits of this virtual chat?
The cat learns everything
The Cat is an artificial intelligence model capable of performing no less than 604 different tasks. He learns several different tasks at the same time. And according to the American company, it would be better than human experts for 450 of these tasks.
He is able to play Atari video games, caption a picture, converse with a human via chat, and stack blocks from a robotic arm that he can control at will. All the strength of the Cat lies in one point: he does not forget what he has been taught. In recent years, many models have started to mix different skills. We can cite “DALL-E” or “Imagen”, capable of generating images from a simple textual description.
Recently, the French artificial intelligence “NooK” succeeded in winning several world bridge champions. “AlphaZero”, another model already built by Deepmind has learned to play go, chess or even shogi. With one difference: “AlphaZero” could only learn one task at a time. After learning to play a strategy game, he had to forget what he had learned to move on to the next game.
Towards a super intelligence?
The cat learns several different tasks at the same time. He can therefore easily switch from one skill to another without having to forget it. This method is a significant step towards a mythical quest, that of generalist artificial intelligence.
At the beginning of the computer age, many theorists defined this intelligence as the ultimate goal. She would be able to think, to learn, to reason, to show logic, in short, to be similar to humans in their way of thinking. So are we near the end?
In fact, in a technological world accustomed to advertising effects, opinions are divided. First, the reliability of the model is sometimes strained on basic issues and makes mistakes. For example, during a conversation with a person,
Namely: the Cat would have replied that Marseille was the capital of France. In turn, Deepmind claims that in 450 of the 604 tasks that Gato would be able to perform, his model performs better than an expert in more than half of the cases. However, for some researchers and experts in the field, the Cat would still be far from the performance of a human.
