Hacker News Reader

Noob question - how do we know that these autoencoders aren't hallucinating and really are mapping/clustering what they should be?

Hmm the hallucination would happen in the auto labelling, but we review and test our labels and they seem correct!