Are we even at point #3 for anything besides structured games like Go or Chess? Not that those tasks aren't valuable but there is a difference between a rigidly structured and scored task like Chess and something free-form like "fold this towel" or "write this program".
Are we even at that in large problem space games like go? Alpha go lost to amateurs making a really big wall that took too many steps to calculate until it was hand patched by humans.
Perfect Plagiarism is a hell of a handicap