I enjoyed and learned so much from this video that I'm viewing it a second time and studying it. At 13:15 the top two graphs under "Hierarchical feature examples" are labeled "GloVe PC2" and "GloVe PC1". What do those quantities mean? Postscript: Okay. Dr. Gwilliams explained that they are "proxies for semantic representations using GloVe word embeddings, which are an unsupervised method of embedding word vector representations." Oh well. --