@trochee@dair-community.social
Post #2908085
2025-06-18 05:07 UTC
The thing is, LLMs DON'T ACTUALLY WORK for knowledge or computation
they provide outputs that *resemble* someone thinking about it
But don't actually bring reasoning to bear
any problems that require actual mental models of the world or a specific area of expertise fall apart at the first or second step
Replies (3)
-
@trochee@dair-community.social 2025-06-18 05:12
Just to do level setting: I would bet ten bucks to one that an LLM can't get out of an escape room rated for a caffeinated tween birthday party
-
@trochee@dair-community.social 2025-07-04 15:55
☝️ > DON'T ACTUALLY WORK https://eev.ee/blog/2025/07/03/the-rise-of-whatever/
-
@CatsWhoCode@twoot.site 2025-06-20 10:25
@trochee@dair-community.social This is why I was cursing out OpenAI a day or two ago. Not that it will stop them, but some people are treating ChatGPT as some sort of "all-knowing" god and it's literally just stolen data and images. It's not even correct most of the time 🙄