Astra and Fable may learn less than humans given repeat tries, a user says, citing Epoch
A user expects future language models to learn on the job much better than Astra and Fable do now.
TLDR
Citing Epoch, a user says Astra and Fable don't seem to learn as much as humans allowed multiple attempts at a task. The user sees room for improvement, predicting that future language models will learn on the job much better than these models do.
Combined views
59.8K
2 Sources, first seen ago