World Cup 2018 results shocks artificial intelligence predictions! What are the reasons?

We are on the threshold of the 2018 World Cup semi-final. ِAmong the temas who are qualified to the semi final there are no Spain, Germany, Brazil, and of course many other major teams, but I mentioned these three teams precisely because they have a special story and it has to do with artificial intelligence.

The story began with the beginning of the World Cup or shortly before, as we have seen many attempts to predict the World Cup champion in this session, and artificial intelligence software and algorithms were the tools used by researchers to make this prediction.

As you know, the field of artificial intelligence and computer learning is almost universally relevant. It is hardly a new day to hear about a new achievement or new experience in this field. So it was logical to meet the path of artificial intelligence with the road of the World Cup at a crossroads, and that is what it was. In fact, there was not one, but three.

The three attempts that attracted everyone's attention had mixed results. First, artificial intelligence predicted Germany as a champion of the world, where another artificial intelligence program expected Spain to be the hero. The third attempt was that the Brazilians were happy with their results, which we had already written about in Aragic.

But as you saw, Germany came out of the first round with a startling surprise, and even finished fourth in a group that everybody thought was within reach. Spain is the other, and the World Cup from the Round of 16 is called by host Russia in a surprise no less than Germany's surprise. Brazil are the last depositors, and in fact, until the last minute of their match with Belgium, there was some hope for artificial intelligence in order to achieve one of their prophecies, but the Belgians had another opinion.

Three different programs based on artificial intelligence, in three different laboratories around the world, resulted in three different predictions of three different champions. However, the final outcome is a total failure of all attempts, and proved to all that there is a clear weakness and flaw that led to that failure.

What are the reasons?

Artificial intelligence programs rely on the concept of neural networks, and although there are many other concepts, neural networks still have the lion's share. The output of the neural network depends heavily on its input, in addition to many other factors, such as the number of layers and the number of neurons in a single layer. This is what is causing the sharp contrast in the final results, which gives us a glimpse to understand why a software program predicted Germany as a world champion while others chose Brazil.

The reason for the failure of all these attempts is the same: output is dependent on income, and therefore no matter how researchers try to provide their neural networks with factual data, there is a severe shortage that can not be remedied. For example, a researcher enters data on the results of the last matches of all teams, Registered and ten other different factors.

But there are hundreds, perhaps even thousands, of other factors that can not be grasped by the many, the difficulty or the impossibility of being represented mathematically, such as the psychological state of the players, the coach's words between the two halves, the voice of the audience or even the length of grass on the court. And certainly a danger to your mind dozens of other factors that are impossible to model.



The conclusion is that it is very difficult, and even impossible, that researchers can soon model the outcome of a football match, because it simply follows the thousands of different factors that are hard to track or measure. But let us leave the door open; for as history has taught us, there is nothing impossible with human ambition. Or do you have another opinion?