Yea that sounds pretty out there. I don't see why data should be linked to human brain power when you can generate gobs of data and use computers to search through it for the few bits of interesting stuff.
Quite frankly, I'm shocked this was published anywhere. What sort of a distribution would they have expected? A normal distribution where the average filesize was 5GB? And to then infer that this has to do with mental constraints... this will be a laugh with my colleagues.