Post by @OrionKidder@mas.to

2026-09-13 20:13 UTC

@pluralistic@mamot.fr The new insight here, to me, is that the math problems that LLMs are solving are a) actually real, and b) of a specific kind that that technology can solve--ie, the ones that can be "brute forced"--because LLMs are suited to brute force attacks. #AI #LLM #hype #AIBubble

Replies (3)

@OrionKidder@mas.to 2026-09-13 20:16

@pluralistic@mamot.fr Once again, LLMs are a *normal technology*, one that has use cases, but it is being pushed by a really dangerous industry. Not because it's going to become sentient or become a god or take over all our jobs. But rather... #AI #LLM #AIBubble

Open ##5422300
@m3tti@functional.cafe 2026-09-14 05:27

@OrionKidder@mas.to @pluralistic@mamot.fr this is so true!!! Bruteforcable problems are really well suited for random word generators.

Open ##5422304
@troberts@theblower.au 2026-09-14 05:29

@OrionKidder@mas.to @pluralistic@mamot.fr I did like that point, how with a clear success criteria *and* being susceptible to brute force approaches, it can be successful. It’s important because only a certain subset of real world problems meet that criteria. E.g. a lot of problems you only get the first chance to do it well, you can’t just keep trying over and over.

Open ##5422305