Discussion about this post

User's avatar
Jurgen Gravestein's avatar

Thanks for this in-depth analysis. It surprises me how many people are rushing to conclusions without even having read a single page of that paper.

Expand full comment
Visar Berisha's avatar

The authors of the manuscript could have framed this in terms of repeatability, or lack thereof. It’s difficult to use instruments that’s aren’t repeatable for high stakes applications or for scientific discovery. And, as suspected, LLM outputs don’t appear to be repeatable.

Expand full comment
11 more comments...

No posts