We just had two papers accepted at NeurIPS’22. First, work led by Will and Heng on a new biologically plausible and powerful model of backprop-like credit assignment in cortical networks. Second, work led by Milton on combinatorial generalisation in deep neural networks.