Hasanhodzic, Lo and Viola constructed a “Turing Test” to assess the ability of human subjects to distinguish between actual and randomly generated returns of financial securities. The researchers found that there is significant statistical evidence that humans can consistently distinguish between actual and artificial series and that the belief that financial markets “look random” (i.e. random walks) is hence probably erroneous.
The human eye has indeed a significant advantage over most current computer algorithms on several classification and image recognition tasks and that could probably explain the findings (at least partly). However, would the results have been the same had the researchers used more sophisticated simulations (to mimic stylized facts of financial markets)?
p.s. A link to the test can be found here