This is, scientifically speaking, fraud.
Higher dimensional problems require monstrously larger sample.

It is similar to the racist cranial theories [with even worse math]. https://twitter.com/baumard_nicolas/status/1308715609795039232

2) The paper is a plain case of fooled by randomness. Look how noisy the pictures, how nonsensical are the regression lines.
In higher dimensions, errors get worse.

3) Now compare the regression of that pseudo paper above to the ones coming from random points #FooledbyRandomness

4) Managed to derive the distribution of the #FooledbyRandomness
regression line. I don't know if such a result has been published.