There are generally arbitration opertunity's that offer short term rewards well beyond the standard risk reward curve. Which is how many algorithmic traders consistently return high double digit returns with ought leverage. The simplest being having a mix of buy and sell orders that straddle the average.
Which is exactly what the hypothesis says. Assume that others are using arbitrage to reduce your opportunities, leaving you with an efficient market.
Except arbitration requires more than just money + information, people doing arbitration must have low fees / latency for example so it's not available to all players in the market. Net result real world markets are not efficient due to various access levels.
Also: Empirical analyses have consistently found problems with the efficient-market hypothesis, the most consistent being that stocks with low price to earnings (and similarly, low price to cash-flow or book value) outperform other stocks. Which is presumably due to cognitive bias. http://en.wikipedia.org/wiki/Efficient-market_hypothesis
PS: It's a reasonable simplification that's useful for the average investor, but not policy makers for example.
This is what people mean when they say "efficient market" in this context:
> In finance, the efficient-market hypothesis (EMH) asserts that financial markets are "informationally efficient". In consequence of this, one cannot consistently achieve returns in excess of average market returns on a risk-adjusted basis, given the information available at the time the investment is made.
http://en.wikipedia.org/wiki/Efficient-market_hypothesis
There are variations of this hypothesis, so saying "it's dead" is meaningless unless you specify which one is "dead" and how it has failed.