Chess only has a finite number of moves. Yes, they refined the Algo to the point it can consistently beat Kasparov. I fail to see how this translates to excelling in any other area besides chess.

The fact that lifetime govies think AI is some 72 hour ahead crystal ball pretty much demonstrates the point I already made. What we have is a large, well developed decision tree that incorporates numerous pieces of data and predicts likely outcomes. However, it is still software and it does not "think." All the needs to happen is one single piece of unanticipated data creating an unhandled exception in the code and your AI either breaks or produces an incorrect prediction.

The last data I saw on this said that given certain specific tasks they could increase the accuracy of predictions up to something like 80% accuracy. That still means 1 in 5 predictions is wrong and general purpose usage did significantly worse than that. Remember, at 50% accuracy you may as well flip a coin.

Honestly, in many ways I feel like "AI" is a solution in search of a problem. The idea that we will just submit a bunch of random data into a giant all knowing central brain that will spit put solutions for all of mankind's ills sounds a lot like a 1960s scifi pipe dream to me.