I was putting together a Suggested Reading list for a friend of mine, something for him to chew on while he worked on his next book. His current book (I'll have more to say about that soon) dealt with life after a Big Collapse of sorts, so I yanked out a few things that looked relevant: A Canticle for Leibowitz, and Hiroki Endo's Eden.
One of the books on the list turned out to be Phil Dick's Dr. Bloodmoney (Or How We Got Along After The Bomb). My edition, an older paperback version, has an afterward in which Dick blithely admits that he got it all wrong: he was only too happy to see that nuclear war did not in fact break out between Us and Them in the timeframe he'd anticipated. But then again, as he pointed out, SF isn't meant to be forensically accurate in its visions of the future — it's just meant to see a future, something that serves as a warning or food for thought or something of that nature. Predicting is not really the point.
That was in fact one of the beefs I had with SF for a long time — that it seemed to be about predicting the unpredictable, and that 99% of the time the predictions were dead wrong. But whether or not a story about the future is meant to predict anything or just depict something is really up to us; we're the ones that interpret what we read as being one of those things. And there's any number of stories where the prediction simply doesn't hold up because our sense of what humanity is like is now entirely different. H.G. Wells's Things to Come probably seemed pretty heady and stern and even believable in its day, but today it comes off as naïve and idealistic: does anyone really believe that after a Big Collapse, there will be some utopia of scientifically-enlightened philosopher-kings stretching out a hand to lead the survivors from the wastes? No, because we've had Mad Max and Blade Runner and all the rest of the "dystructopias" since then to serve as more credible counter-examples.
I guess what matters most from any vision of the future is not that it is technically accurate, but that it understands what people do and why, even under the most outlandish circumstances. The only SF writer I can think of who came even remotely close to predicting the information age we have today is John Brunner. Go read The Shockwave Rider for his 1974 take on the Internet; it's not all that far removed from what we have now, including what amounted to Wikileaks. But even if he had gotten all of that dead wrong, the book would still be excellent, because it understands human behavior as transformed by a technological society. And, before all of that, human behavior period.