Artificial Intelligence (AI) is a field in rapid development with major societal and economic issues. The most common technical solution is based on deep neural networks. Unfortunately, today there is no mathematical formalism of their operation to explain their performance.
To this end we are moving towards an interface approach between Machine Learning (ML) and theoretical physics. This is motivated by the fact that the ML problem, aiming at extracting intelligible information from a large number of data, finds an echo in statistical physics. Indeed, the latter is aimed at extracting an effective description in terms of a restricted set of parameters of physical systems having an arbitrarily large number of degrees of freedom.
This day will aim to promote this alternative approach and discuss its potential in terms of interdisciplinarity.