mwatkins comments on Mapping the semantic void: Strange goings-on in GPT embedding spaces

[mwatkins]

Thanks! I’m starting to get the picture (insofar as that’s possible).