[daveguy]

The worst part is the "p-values" it completely made up.

Verification that the numbers were legitimate -- "What's a p-value."

At least it got the values right .9912 (twice), .8813 (twice). It estimated 88-99% chance of bullshit... To 4 significant digits... Repeating the exact values.

Turns out, this is an excellent example of how poor these models are with accurate knowledge vs confident bullshitting.