AI is a bloated term---we are facing this since day 1 of the definition stage. There are already quite a few questions going beyond the original (blurry) boundary of the proposal, notably on implementation issues.
But the worst problem seems to be the lack of objectivity in answers, and sometimes in questions too.
I will single out this question at time of reading, but there are already several like this one.
We must avoid too many threads that lack objectivity. I intend to vote down answers that are too subjective (but, well, I cannot down vote infinitely, as you know), and comment as necessary. Scalability issue, even in this private beta.
What would be the best way to proceed?