Although the mass loss of massive stars is of prime importance for their evolution, it is still not well understood. Stellar modelling relies on wind prescriptions that are implemented in 1D evolution codes, but many uncertainties remain. In any case, stellar models cannot follow mass loss burst episodes, like the one that Betelgeuse has undergone last winter. They rely on averaged mass-loss rates, but is there such a thing, from an observational point of view? Can we check wether the winds prescriptions are correct or not? How modifying the mass-loss rates during the RSG phase changes the endpoint of massive stars evolution? I will review our current understanding on these questions from the theoretical point of view.