Throughout the long evolutionary history of life, species of all kingdoms have undergone staggering diversification and faced countless environmental changes and extinction events. Since the great majority of species that lived on Earth have since gone extinct, it is difficult to infer how biodiversity dynamics unfolded over millions of years and yet crucial to understand the fundamental evolutionary processes driving them. Today, with over a million species threatened with extinction, biodiversity is facing unprecedented challenges, urging the need for conservation policies that maximize its protection and sustain its manifold contributions to people. Here we present a suite of new supervised and unsupervised models with applications in evolutionary and conservation biology. Specifically, we show how deep learning models can be used to infer biodiversity dynamics in deep time, and test hypotheses on past speciation and extinction based on fossil data. We then present a new semi-supervised framework to improve our understanding of the evolutionary relationships among living species based on the analysis of genomic data. Finally, we introduce a einforcement learning framework to optimize biodiversity conservation policies. Coupling artificial intelligence with evolutionary models holds great promise for advancing our understanding of biodiversity and its evolution across time scales, and for improving biodiversity conservation action in a rapidly changing world.