Originally Posted by
DanceswithUnix
Deep Blue was a conventional computer program, not really an AI. It's "cleverness" is programmed in, so it can't be better than it's creators, though it can be the sum of its creators.
The massive leap was in deep learning neural networks. We knew how to make neural networks in the 60's, but we were rubbish at training them. Systems like Alpha Go are trained networks, and these systems often learn by playing against themselves based on a programmed scoring system so they know if they are improving. That means they can learn tricks their creators never dreamed of.
So it's the AI research that is driving this, the game playing AI's are just using the existing tools with better effect as time goes on. In fact, it is already at the point where we should probably be teaching this stuff in schools. "Today we will create an AI that plays Frogger" would be a lot more current than the "Today we will write a calculator in vis-basic" that my kids actually do in lessons. As for the game industry having money for R&D, sadly they have a reputation for poor pay and conditions which doesn't attract the best programmers out there. But I'm sure once a "The Singularity for Dummies" book comes out they will be right on it.
Have you played "The Talos Principle"? A fun puzzle game with a philosophy based storyline, might be right up your street.