So I think the way I see it is we kind of gave up general AI in the 80s. Instead the focus was on machine learning, which tackles the learning problem in a formal manner.
The thing is, in the last maybe 10 or 20 years we got really good at machine learning. I mean algorithms using support vector machines take what was thought to be a intractable non-linear learning problems and make them tractable. This sort of thing didn't exist for the AI researchers of old.
Deep learning is the hot new thing now, and it basically is attempting to take all we learned in ML and apply it to the AI problem. And you know, it's exploding. And there are some promising results already.
So I think it might not be too far off that we crack the general AI problem. I'm betting on maybe late 2020s. Yeah it's stupid to make predictions but whatever.