However, I respectfully disagree that humans are better at making trading decisions than computers. The world of algo trading can be divided into two sides, a) high frequency traders, who through fast cancel-and-replace limit orders and colo with the market centers, try to act as virtual market-makers (or scalpers, depending on your perspective), b) buy-side institutional fund managers who want to complete their orders, without HFT predators and negative market pressure. Large block orders are spliced into small lots (i.e., VWAP) and sent to the market using intentional limit price over time to hide the movement of a huge buy or sell order on the market.
Human beings might be better than machines at picking single stocks for long term investing (although most people are probably still better off investing in an ETF). But that's not how sell-side traders make money in the first place. Guys like GS/MS/Timberhill make money by having the unfair advantage of faster execution speed, more capital and specialized trading algo's against the small retail investors.