[catlifeonmars]

Why would the embeddings be higher dimensionally than the data? I imagine the embeddings would contain relatively higher entropy (and thus lower redundancy) than many types of source data.

[cm228]

depends on the chunk-size used to create the embedding.