Machine learning in physics – philosophical perspectives on new tools

In most fields of physics, machine learning (ML) is all the rage. Physicists use ML algorithms to analyze data, to cast predictions or to emulate computer simulations. But what exactly is machine learning? How trustworthy are its results? What are possible pitfalls? And how does ML impact on the methodology of physics? The aim of this article is to discuss initial answers to these questions on the basis of recent philosophical work