So you believe "incorporate misdirection and trolling" is in the same ballpark as detecting NSFW pixels? Me not, the former is very hard even for humans. For th

So you believe "incorporate misdirection and trolling" is in the same ballpark as detecting NSFW pixels? Me not, the former is very hard even for humans. For that reason, having AI that is trained on it is incredibly dangerous - if it's possible to train AI to avoid trolling, then it's equally possible to train it to troll.