Join Nostr
2025-01-11 13:59:02 UTC
in reply to

Space Hobo on Nostr: I always frame #LLMs in that they reason through "dream logic". This is especially ...

I always frame #LLMs in that they reason through "dream logic". This is especially resonant for literature students or art students, as it gives them a lens through which they can express the ways in which the tools have been helpful. But it is also useful in harder research or engineering disciplines to show how the results cannot be taken as fully reasoned or researched.

So from an academic standpoint, sure: go have a chat with an AI chatbot to work through your ideas about that essay, but then close the tab and write the thing yourself. Look up citations for all the points from your conversation—Can't defend it? Shucks, on to the next one!

And yeah, maybe explicitly LLM-generated material is useful in a presentation on Surrealism, or Jungian ideas, or... I dunno, I don't work in those fields. You know where dreams and dream-logic is appropriate in your discipline, and how much disclosure is appropriate, and that's exactly where #LLM materials belong.

I had one of the Official Fisher Space Pens, and it came with a slip of paper where the inventor talked about having trouble making the design work in the early days. He said his father came to him in a dream and told him to put a bead of rosin in a particular spot. Now, rosin was absolutely the wrong material for this, but *resin* worked perfectly.

He didn't start up the factory with a million units containing rosin just because of his prophetic dream: he used it as inspiration to *test* new avenues that he may not have been consciously considering.