Football – La France wins against Germany and AI from Microsoft – ZDNet France

Predict and predict: this is a dangerous game. It is indeed taking the risk of error and to see blame afterwards. Microsoft France had nevertheless played the game during the Euro 2016 football.

And to promote Bing and its predictive technology based on the artificial intelligence. The AI ​​has not failed to cheek by predicting a German victory over France. A win by a short head though. The likelihood that Germany wins was 52%.

Microsoft admits “not defeated so far”

But as often , enter in the equations of the human, and therefore unpredictable, and your statistical model may produce an incorrect prediction. This applies to football. Thus France beat Germany two goals to nil.

What does it mean? That Microsoft’s IA should return to its calculations and make amends? This underlines especially that what is probable is not certain, especially at 52%. On Twitter, Microsoft France therefore reacts with humor (true or fictional, with each judge).

“A defeat our #IA but a victory for #FRA for our delight! appointment for the final! “

We expect to present the publisher’s prognosis for Sunday’s final between France and Portugal. Come on, give Portugal winner. Why ? Like this, just to make the evil spirit and tease football supporter can be a bit touchy and indifferent to the magic of statistics.

AI and predictive, it already works

with regard to artificial intelligence, if it can go wrong in such areas as sports, there are others where it already works effectively as can attest to different companies .

for the targeting of its newsletters, and thus to sell more, Fnac operates a SaaS solution for predictive analysis based on machine learning (Tinyclues). Using the mass of data available to the sign, it is possible to score the probability that a given person buys a product. A percentage is provided. A human then trigger an action and then measure the results.

The operators are also users of predictive solutions, especially to identify customers on departure. This way they can reduce the “churn”. But these companies are also interested in these technologies to evaluate the responsiveness of a client to an offer.

Let’s leave the last word to an Orange expert, Sébastien Boutin, this in Paris at the show Big Data 2016: “it’s not a magic box, the machine Learning”. “90% of the work remain in the data preparation” he added. The Microsoft AI is not magic, but to err is human.