Technically, artificial intelligence was born in 1956 at a conference held in Dartmouth. That dates back to 60 years ago! At the time, researchers gathered with a common will to create computers that would be as intelligent as human beings. We'll come back to that later. The notion of intelligence is not well defined, but the point of reference they choose to reach their goal is the example of a game of chess. One of their goals is to build a program able to play chess as well as a human being and eventually beat them. They will be interested in other fields such as automatic translations. From that conference onwards, quite a few researchers will join the movement and start rather serious studies. Many new results will appear. Remember this was 60 years ago, it was a different time. These results are absolutely prodigious at the time. One of the very first artificial intelligence that was created was called theoretical logician or theoretician logician. It was a program that was able to demonstrate 38 of the Principal Mathematica theories. For the first time, there was a program capable of reasoning and demonstrating things. As soon as AI was born, there are two trends that emerged, one of which is called the symbolic trend and the other, the digital trend. These two trends will be distinguished with what will be stated later, on the one hand, they wanted to create a mind, that's the symbolic trend, on the other hand, they wanted to model the brain. Creating a mind started with a rather simple idea that was to say that the brain is a machine that processes symbols What the researcher did, was to find new algorithms based on the theory that the brain processes everything as symbols. It is true that some problems, such as a game of chess will help to express the power of symbols. When handling symbols, algorithms are found, they are extremely astute and effective. I can't list all the algorithms that were found, but there really was a golden age for AI between the 60s and the end of the 70s - a lot of discoveries were made. On the other side, there was this digital trend-led in particular by Franck Rosenblatt, where the idea is to model the brain. The symbol theory is rejected in favor of a more digital approach. And that gave way to the artificial neuron network trend, for example. Both of these new trends will give new results. For Rosenblatt, it is the case with the Perceptron. What happened was that although they gathered a lot of results, the original promises concerning AI were not really kept. There was a lot of progress but the general problem about intelligence was not solved. For instance, at the end of the 70s, the chess game didn't work very well. They are clearly still not able to beat the world champion of chess. Based on all the results, those who dole out the money said, "that's enough, as long as you don't have this general intelligence, we won't fund you anymore." What happened is that we had the AI Winter that is to say that from one day to the next, a lot of research was stopped because of the scarcity of funds to support research. They will have had to wait a few more years, until the mid 90s to make rather serious discoveries again. In 97, Garry Kasparov was beaten by the Deep Blue program. The world champion was beaten by a machine. That was a thunderclap because no one thought it could be possible, but it was. It made lots of algorithms progress but again, the general intelligence problem was still not solved. Nowadays, there is a lot of research again concerning AI. For instance, in the same way, the chess player was beaten, the go world champion was beaten a few months ago. That was also a thunderclap because that result was not expected before a few years, because the go game is so complex with an order of magnitude higher than chess. It was thought to be a really arduous problem. Does it mean that the problem of general intelligence has been solved? No, not yet because the problem remains the same. It hasn't changed. because The definition of intelligence can't really be given. Artificial intelligence can be developed, they can be extremely good in one particular sub-area, for example the go game, the game of chess, or even more specific problems where algorithms will be found to help solve these problems. But today, we can't create an intelligence capable of answering every problem even the simple ones . Today, we can't do that yet. The other adjacent problem is, at one point, to define what "intelligence" means. That's the researcher's original sin. There isn't just one intelligence but several intelligences. That's why we prefer to speak about artificial intelligences, in a plural form, the term of artificial intelligence will even be put aside in favor of other fields such as: machine learning, artificial neuron networks, expert systems and lots of other fields. At the same time, in 56, related to this general intelligence problem, Alan Turing suggested a test, that is called the Turing test, its aim was to detect whether or not a machine was intelligent or could be considered as intelligent. This test is interesting because it is what is called an imitation game: the game for the machine was to pretend to be a human and give rather convincing answers. his test has always been very criticised but it has been referred to because it is entirely based on language which is something very complicated, and that it is only valid for humans as speech must be used. For example, animals couldn't take the test, neither could babies. Nowadays, with research we have on animals and child development, we know that intelligence is not only based on language, and can embody other forms. One of the biggest artificial problems today is creating a football player. It could seem surprising as a chess player has been created. Why is the issue a football player? Because if you take a humanized robot, it has to move, to play - alone and with others, so it will have to interpret emotions, and others' intentions. All of which are extremely difficult problems yet to be solved and are at the heart of a lot of research. This joins up with a more general problem about this symbolic theory, that was that when people started rejecting this symbolist theory, this time, they started considering the body. There was a new movement called embodiment, of which an emblematic text is a paper by Rodney Brooks, a rather famous roboticist who wrote: "Elephants Don't Play Chess". The title is quite surprising, but what he meant was that elephants live their life of elephants in the savannah or in the jungle. An elephant has never heard of chess. We can guess it doesn't need symbols but has its own elephant intelligence. That was to reject this simplifying theory. Today, the idea of embodiment is to say that intelligence is linked to a body which grows and interacts with its environment. So, if I want to test an object's flexibility, I will need a body to experiment this object's flexibility. That's one of research's the main directions in artificial intelligence and in robotics too, in cognitive robotics.