No doubt the next "breakthrough in computational efficiency for generative AI" will be an algorithm that receives a prompt, passes it unchanged to an image search website, and returns one of the pre-existing image from the third row of the results.


You must log in to comment.

in reply to @belarius's post:

I briefly considered writing a program that jumps between different scales of Markov process (by letter/word, watching history of different lengths, etc.), connect it to some semantic-analysis APIs, pretend that it's run by massive neural networks and "maybe showing emotions," and seeing how much VC money I could grab...