I don't really buy this whole, "the correlations have gone to 1" argument. First of all Felix states that correlation is impossible to measure and cites a wired article about the Gaussian Copula. This is a bit of a straw man. Correlation isn't hard to estimate at all. Felix's second point is that correlation is backward looking and therefore no good. I agree that it is a backward looking measure, but I wouldn't through the baby out with the bath water quite yet.
You have to consider what are the alternatives there are to asset allocation. Felix states:
In investing, nothing lasts forever. And the era of asset allocation is in its waning years. The problem, of course, is that no one has a clue what might replace it.
Asset allocation is not a trading or market timing strategy which might loose its ability to generate returns over time. Asset allocation is a method of managing risk. It will reduce portfolio risk overall, but as we have seen, in severe events, correlations do not necessarily go to 1, instead the impact of the market component on diverse assets becomes far more important. Nothing except being in cash would have saved you in the recent market drop, but for most time periods, asset allocation has and will continue to reduce portfolio risk. Put another way, if you are driving a car a seat belt (asset allocation) will help a lot, most of the time. But there are times when it won't (i.e. you drive off a cliff). That still doesn't mean that you shouldn't wear a seat belt.