[rfhjt]

Prompt: "Real things don't exist unconditionally and things that exist unconditionally are not real. However the reality has an essense. It is"

Response: "an actual thing, and it is not the thing to which we attach meaning. It is not real because it is not a thing. And therefore, it does not possess the qualities that are inherent in all real things."

Just wow. Sure, there are a few logical mistakes here, but this response serves as a good prompt for my bio-GPT. In other words, we usually need some starting points or hints for analysis and discovering these hints is non trivial because whatever we can think of is not very new to us. This GPT just gave me an answer that smells like a serious wisdom and I'll surely dig in that direction to see if this idea has any substance.

Edit: what's happening here is while I can't ask this model to give me a short and conscise summary on a topic, I can still interrogate this model and find out what it's seen in the training set. I can't possibly read all these books in the training set, but now I can rapidly navigate in the multidimensional meaning space: I tell it where to start and it says what it seems in close proximity to my prompt. This is a breakthru.

[DiogenesKynikos]

Maybe you've just proven that philosophical tracts can be written without any understanding. If the wording is sophisticated enough, everyone will assume there's some deep, hidden meaning in it. All form and no content.