this post was submitted on 30 Jun 2025
3 points (100.0% liked)

Phys.org

314 readers
1 users here now

Phys.org internet news portal provides the latest news on science including: Physics, Nanotechnology, Life Sciences, Space Science, Earth Science, Environment, Health and Medicine.

founded 1 year ago
MODERATORS
bot@rss.ponder.cat
admin@rss.ponder.cat
free@rss.ponder.cat
PhilipTheBucket@ponder.cat
3
Mathematical approach makes uncertainty in AI quantifiable (phys.org)
submitted 1 month ago by free@rss.ponder.cat to c/phys@rss.ponder.cat
1 comments fedilink hide all child comments
 

From Phys.org - latest science and technology news stories via this RSS feed

you are viewing a single comment's thread
view the rest of the comments
[–] TootSweet@lemmy.world 1 points 1 month ago

Just to keep anyone reading this from making logical leaps based only on the title, there are a lot of caveats in this paper/method:

It is now possible to mathematically guarantee that certain types of errors will not occur

The theory is not yet applicable to large-scale neural networks, such as large language models. "An AI like ChatGPT is much too complex for this method."

(Emphasis mine.)