Why Maths behind Artificial Intelligence isn't optional? The Fukushima disaster: an example of overfitting

In most of the popularization articles about Artificial Intelligence, it is easy to guess a kind of secret hope (at least for non scientist people) that all the related tools (neural network, deep learning etc...) could "magically" bring the solution to almost everything in the future. Easy and elegant, just train it on data and it works!

This fantastic marketing slogan might be true. But despite all its power, Artificial Intelligence should never become a toolbox for everyone's hands (and even more a miraculous black box for non scientist people as it already tends to be the case). There is no magic bullet in Sciences: wrong Maths could lead into real disasters. And unfortunately the basics are relevant for Artificial Intelligence models as well.

Here is just below a good tutorial about the "overfitting" Machine Learning general problem and its terrible consequence for the Fukushima power plant.

[Machine Learning Course #4 : The Bias-Variance Delemma]

Extract: "The overfitted model predicted one earthquake of at least magnitude 9 about every 13000 years while the correct model predicted one earthquake of at least magnitude 9 just about every 300 years. And because of this, the Fukushima Nuclear Power Plant was built only to withstand an earthquake of magnitude 8.6. The 2011 earthquake that devastated the plant was of magnitude 9 (about 2.5 times stronger than a magnitude 8.6 earthquake)."