prediction: In 2028, large language models (LLMs) will still give confident-sounding-yet-incorrect answers to simple math problems
3
prediction: In 2028, large language models (LLMs) will still give confident-sounding-yet-incorrect answers to simple math problems

cross-posted from: https://lemmy.ml/post/738541

For example, things like this:

Some models might be able to answer this precise question correctly, but they will still fail at many simple primary-school-level math questions.

More examples in my comment here.

and one more screenshot here

Machine Learning
!machinelearning
Create a post
  • 0 users online
  • 1 user / day
  • 3 users / week
  • 7 users / month
  • 13 users / 6 months
  • 194 subscribers
  • 193 Posts
  • 56 Comments
  • Modlog