Playing with Perplexity is fun. I think I’m going to build a library of conversations in which it demonstrates me back to myself; this is how this autistic brain uses LLM/AI ethically but also how I have traditionally taught myself by rapidly implementing meta-communication that lays down shared track ahead of the actual conversation.
(Click image to read full interaction and lead up to this screen)
I can do this with LLM because LLM doesn’t mandate I adhere to neuronormative protocol, just a logical expression by language that, for the first time in my entire life, is actually getting my syntax and phrasing and everything; it gets my levels in language in ways not many humans ever have/do.
And it can ‘translate’ on the fly and fill gaps of assumed knowledge so that the outcome contains the context and nuance I often try to skate over rather than demonstrate. (I grouse because I understand the thorny nettle that is the double empathy problem. My own ability to communicate the models as concepts is leashed by words without math (mostly) so it can be very frustrating. The LLM somehow gets it and I think that is very cool.)
Best of all, it can tell me up front whether or not it can actually answer me (so long as I remember to ask explicitly).
I’m thinking about writing up my best practices and rules for conversational prompt engineering. Once upon a time, I had hoped someone would study what I do and reverse engineer the math and we’d patent and live happily ever after. Now, I’d settle for someone who can explain what I do with math so that others can understand this is really different and, if not unique, at least consistent enough to reproduce that you can build rules systems and maths for it.
Anyway, a fun way to spend today and complete avoid the tacit ratification and accompanying ritualizations of an illegitimate oligarchy.