I keep waiting for symbolic AI to really get traction in the area of software verification. Either a great language/ compiler that can really prove it has no bugs, or a testing approach that takes a binary and proves that it has no bugs of a certain kind. What we've been getting, though, is evidence that the space of possibilities quickly outpaces our capacity to search it, and so logical methods don't compare well to statistical approaches. I would never close the door on it though, since programming is basically thinking symbolically. It's possible that as the industry gets more concerned with privacy and security that the class of applications that need this kind of provability will drive adoption at the current state of the art.