As a forecast of tech-to-come, it seems like science fiction hasn't managed to really predict anything radically different than it was predicting in back in 1990 or so. The accuracy has gone up because everything is now "five minutes in the future" rather than "in the year 5000." Even if we look further ahead, today's far future is just like yesterday's far future: filled either with spaceships (albeit carrying ems, ala Accelerando or Diaspora) or grey goo, and with not much room for real retroactively-falsifiable speculation, which I think is the key to good sci-fi.
The accuracy has gone up because everything is now "five minutes in the future" rather than "in the year 5000."
Er, this seems exactly backwards to me, unless you're including near-future thrillers in "SF". Of stuff on the SF shelf at the bookstore, the majority appears to be fantasy (and much of that contemporary fantasy, now), and of the ones that are not explicitly fantasy, many are far future space opera -- Drake, Weber, etc -- which has the same characteristics as fantasy. There's a whole subgenre that's now missing: the medium-term, realistic future. I think it's become clear even to people who shout down the singularity concept that it's become impossible to put together a detailed-yet-believable future of fifty years from now.