>>16830954
woah...
>b-but it's higher than blind chance
that could have been impressive if had been n=1000 or if the n=100 experiment was repeated dozens of times with different models, as it is this is basically reporting that they rolled a dice 100 times and a 6 came up 1/5th of the time.
that's without even getting into the alternative explanations for how it could guess better than chance (which the authors helpfully list)
