On: ML is useful for many things, but not for predicti...

[blitzar]

> Earlier this year, a paper in the widely read PNAS journal raised the possibility of detecting non-replicable findings using machine learning (ML).

I wonder if they faked their paper too.

[godelski → blitzar]

I wouldn't doubt it. As I said in a longer comment. I would be incredibly impressed if any statistical model could perform this. It isn't something even humans can perform. Human reviewers are only capable of determining if a paper is invalid or indeterminate, but not if a paper is valid.

The question really is if they intentionally faked the paper or if they tricked themselves.