Full simulation of the human neocortex is expected to begin in about 10 years. It seems reasonable to peg the Blue Brain Project as a guide post for AGI, however the key to building a practical human level AI (one that normal people might actually be able to play with or go to the mall to see) requires the positioning of many other technologies which are advancing in parallel and in conjunction with ol' Blue. It seems probable that these other technologies will be in a position to approach (not necessarily reach) human level AI in the next 15-25 years, but in terms of alternatives to the Blue Brain.
While understanding the brain is crucial to building an AI that humans can recognize and relate to, humans seem to disregard too easily the notion that human level intelligence, and more specifically human-like intelligence and more generally biological intelligence are a certain kind of selected evolutionary solution to biological pressures. That is to say that biologically inspired processes may be a good place to begin to understand artificial intelligence, however it does not strictly require that our kind of intelligence is the only kind or the most effective kind, but it is perhaps the simplest, sexiest and/or most fuel efficient solution that has evolved to do what we need to do to survive here on earth - probably not much more or much different than that.
Genetic algorithms already produce novel solutions to specific problems that humans have no other way of generating in a comparable amount of time. Research into mimicing human social cues is producing very convincing interactive robotics which seem to posess a level of intelligence that we recognize, but not the kind that can answer all of our questions. Wolphram Alpha is attempting to address that score. Fortunately tracking software which learns to predict our needs is near on the horizon. Ultimately all of these things must merge in order to complete the loop of human level intelligence, but that is only software. The hardware is coming of age too. Advances in self-assembling networks and advanced memory and processing architectures are all moving in concert to make this revolution an inevitability.
The only remaining question now is when. The best guess seems to be sooner rather than later. That's not to be hasty or overly optimistic. Most would agree that understanding human intelligence is a difficult problem from the top down so science is trying to build one from the bottom up with the Blue Brain which is currently slated to turn the corner on human brain sumilation at the beginning of the potential curve for AGI. Tracing this event as a 5 mile marker for human level AI places the overall trend squarely in the next generation or two and on par with Kurzweil's predictions. 2050 would be a conservative mark. But the immediate best course in any case is further dialog and debate. The general public certainly wouldn't want to wind up with this technology a couple of years or months away without knowing what to expect or even how to consider what will be whether it looks good or not so.