The problem with "pie in the sky" research is that for every honest-to-god researcher you attract a dozen charlatans that promise you the world in your hand in ten years. Then the system selects for the best charlatans, the ones that make you believe the breakthrough is around the corner, even though there is no hard evidence for it. The MIT Media Lab blew through billions of dollars in this mode, with little to show for it.
AI has gone through several cycles of this, with the end result that generally weaker students were being attracted to it. One of the reasons AI is more successful today is that the last AI winter was so severe that weak students stopped going to AI, so what was left is honest smart researchers spending all their time in science instead of press releases.
As deep learning becomes more popular we see the return of the weaker minds, with flashy results of little or no practical significance, but plenty of PR value.
You know the type, the people who tell you a slogan and make it seem like they have a solution (ahem Yes we can, Make America great again ahem), but when you stop and think about you realize they were full of "ideas" and bereft of execution.