Perhaps it is perhaps it isn't. But we can now equal and sometimes beat human perception on some standard machine learning datasets. Perception is one very larg

Perhaps it is perhaps it isn't. But we can now equal and sometimes beat human perception on some standard machine learning datasets. Perception is one very large piece of the problem. The other large part shows some signs of cracking, consider the recent success in go and poker. AI can now beat professional players in both of these games. Agreed these are a different class of problem to realtime learning in a dynamic environment, but there is interesting progress here. I'd wager that we're less than 20 years away from a useful general intelligence that could say, earn money on mechanical Turk. And the recent succes with transformer networks understanding text. Incredible.