29 points surprisetalk 3 days ago 7 comments

groby_b 1 hour ago | parent

The existence of attractors (e.g. "stable at iteration 31") for AI models is one of the weirdest things in AI land. They are a wonderful space to explore, fascinating and full of horrors.

Meanwhile, my hat's off to the (currently) last comment on the page: "I like how the shear-sheathed legs still earned their keep — probably load-bearing for Mrs. Regan"

amelius 1 hour ago | parent

The same happens with humans, in e.g. chain stories, cumulative stories, etc.

Not unique to AI.

Muromec 57 minutes ago | parent

It doesn't merely stabilise, it almost converges on itself

lowbloodsugar 56 minutes ago | parent

>I don’t write about generative AI much. Everyone else does; not much of consequence ever gets said.

I see he chose not to break with tradition.

MarkusQ 49 minutes ago | parent

It would have been interesting to try something closer to the movie poster example; say have the AI rewrite the passage with the room toggling between the initial light-mode decor and a dark-mode decor, then back, repeatedly. Or even change the woman's name to "Mrs. Smythe" and back.

layer8 17 minutes ago | parent

As the article explains: “[…] In other words, if you instruct a model to change a single word in a paragraph of text, it can almost always handle the task with no collateral damage.“

pattt 2 minutes ago | parent

It would be interesting how many iterations of LLM-based PR reviews and fixes it takes to turn the code into horror. It’s already started happening.