I recently finished Nick Bostrom's book Superintelligence: Paths, Dangers, Strategies. I highly recommend it. If you don't buy Kurzeil's predictions, check out Bostrom's survey of all the experts in the field, which I mentioned here. Exponential growth is something we're really bad at because we didn't encounter it in nature in order to evolve hardware to handle it. We might all be surprised with how quickly things happen. Kurzweil's prediction accuracy over multiple decades up to this point is, IMO, astounding.