This is cool stuff: http://www.nature.com/nature/journal/vaop/ncurrent/full/nature12600.html Of course, this means that brains are that much more difficult to simulate. Technology prognosticators always underestimate paradigm shifts in innovation, constantly forecasting the end of moore's law without accounting for breakthrough technologies that make yesterday's physical systems irrelevant. Progress marches on. Amusingly, this is (of course not actually) something like a biological corollary. Things are more complex than you might have thought. Evolved systems use things in surprising and unpredictable ways. Kurzweil made his predictions based on the number of neurons present in a brain and how interconnected they are. But now there is more too it than that. How many transistors do you need now? How many more permutations and interconnections and variables? I'm neither a neurologist nor a hardware architect, but adding sub-processing and signalling at every dendrite has got to be more than just a geometric increase in complexity and processing capability. From a lay perspective, it seems like it must be exponential. Admittedly, an exponential sequence is just another kind of geometric sequence, so that should tell you just how very lay I am. Regardless of that, how far out does this push the "human brain equivalent computation" areas of singularity prediction? I'm not one of these "Long Now" type of folks, but I do have to imagine that we might be another hundred years out, instead of 2030 or whatever it was?