The art of doing mathematics consists in finding that special case which contains all the germs of generality. Indeed, in many models of deep learning, it’s possible to excite some so-called deep neurons, to create a combination of abstract concepts in the neural network. Some of these neural networks will then be able to imagine which raw data could have led to the excitation of excited neurons. In 2015, Google published images generated by its artificial intelligence DeepDream. These images seem psychedelic. Clouds turn into fish, trees into temples, and tree leaves into birds. In the era of Big Data, another application of Monte Carlo simulations may be the creation of a small representative set of data. To draw a pseudo-random point in a disk, there is in fact a surprisingly simple method. Draw a pseudo-random point uniformly distributed in the square. If this point is outside the disk, discard it and try again.