28

This article analyses the development of artificial intelligence in engineering and its relationship with more traditional modelling practices, based on the case of research at the École nationale des ponts et chaussées. A bibliometric analysis shows that, after 2015, developments in this sphere focused more specifically on machine learning methods, at a rate comparable to that observed in French research as a whole. A wave of interviews with scientists at the École, at the start of this growth phase, highlights forms of hybridisation between machine learning methods and traditional modelling methods.