And for that reason, I think your version exposes the flaw even more thoroughly: you can't reasonably merge a data set of stats on people's opinions on AI with a data set of stats on people's opinions on IP in the way that you're proposing.
To throw out random numbers as an example of the flaw: If 55% of people on HN believe that IP protection for media should not exist and 55% believe that GitHub stole code, it's entirely possible that TFA's condemnation only applies to 10% of the total HN population that holds the supposedly conflicting belief even though HN "statistically" believes both things.
And that's before we get into the question of whether there's actually a conflict (there's not) and the question of whether anyone is accurately measuring the sentiment of the median HN user by dropping into various threads populated by what are often totally disjoint sets of users.