The proof of the pudding: in praise of a culture of real-world validation for medical artificial intelligence

Artificial Intelligence (AI) in healthcare has become a quasi-normal subject (1). In the last few years, there has been an impressive increase in the number of publications concerning the application of machine learning (ML), a set of techniques and models for building data-driven AI systems, to medical tasks, such as the diagnosis, prognosis and anticipation of treatment effects and complications (2).