This paper develops a Graph Convolutional Network that works in hyperbolic space. The development is a relatively straightforward analog of other neural network models that have been adapted to hyperbolic space, but all reviewers agree the experimental results are interesting. There were some mistakes in the original submission and a lack of clarity about whether the theoretical results were being claimed as novel. The authors have clarified that the key mistake was a typo and the correct setting was used in the experiments, which satisfied R1. The authors also clarified that they are not claiming novelty of the theoretical results. This needs to be made more clear in the camera ready. Having said this, all reviewers agree the experimental contributions are interesting. R2 and R3 supported acceptance, and R1 said they were ok with that outcome.