The stellar initial mass function (IMF) is one of the fundamental pillars in studies of stellar populations. It is the mass distribution of stars at birth, and it is traditionally assumed to be universal, adopting generic functions constrained by resolved (i.e. nearby) stellar populations (e.g., Salpeter 1955; Kroupa 2001; Chabrier 2003). However, for the vast majority of cases, stars are not resolved in galaxies. Therefore, the interpretation of the photo-spectroscopic observables is complicated by the many degeneracies present between the properties of the unresolved stellar populations, including IMF, age distribution, and chemical composition. The overall good match of the photometric and spectroscopic observations of galaxies with population synthesis models, adopting standard IMF choices, made this issue a relatively unimportant one for a number of years. However, improved models and observations have opened the door to constraints on the IMF in unresolved stellar populations via gravity-sensitive spectral features. At present, there is significant evidence of a non-universal IMF in early-type galaxies (ETGs), with a trend towards a dwarf-enriched distribution in the most massive systems (see, e.g., van Dokkum & Conroy 2010; Ferreras et al. 2013; La Barbera et al. 2013). Dynamical and strong-lensing constraints of the stellar M/L in similar systems give similar results, with heavier M/L in the most massive ETGs (see, e.g., Cappellari et al. 2012; Posacki et al. 2015). Although the interpretation of the results is still open to discussion (e.g., Smith 2014; La Barbera 2015), one should consider the consequences of such a bottom-heavy IMF in massive galaxies.