Review for NeurIPS paper: Learning Object-Centric Representations of Multi-Object Scenes from Multiple Views