In the first part of the paper the study of rocks from the late century up to now is briefly outlined. In particular, it is shown that classic petrology is passing through a crisis. In order to resolve this crisis, a completely different type of approach is proposed. In the second part of the paper, by introducing new principles in petrology, it is shown that it is possible to study a rock at the same molecular scale of an aqueous solution, partial melt, or magma from which it originated. Furthermore, because statistical physics builds a bridge between molecular properties and macroscopic phenomena, it is possible to understand the mesoscopic behaviour of outcrops. The route traced out is that of the ''reductionism and complexity''. In the final part of the paper, the future direction of research is indicated. Special emphasis was placed on the ''liquid → solid interface'' and on the role of the ''colloidal state'' during petrogenesis. It is suggested that petrologic systems result from two different effects: one deterministic and the other probabilistic.