From the summary of the article, I'm skeptical about drawing many conclusions. For one, his sample size is clearly low--15 students, 90 faces they looked at. For another, his effect size is small--60% probability of guessing correctly when 50% would be expected by a coin flip means barely better than chance.
Yes, it's statistically significant if you look at it as a simple binomial test, but not by much: each student individually tested would be deemed as statistically insignificant at the standard alpha = 0.05 if they got only 53 correct (p = 0.0567), and 60% correct means 54 correct (p = 0.0363). If you do a Bonferroni correction for multiple tests, you find that a student needs to get 59 correct to be statistically significant in this group, which means at least 1 student did better than chance, but possibly only 1 of the 15. And this is all based on assuming it's a simple binomial distribution, rather than running a contingency table. If the subjects knew that half the faces were of gay men and half were of straight men, that can change the probability calculations.
Even if it is significant, you also need to look at effect size. If the average person can indeed guess right 60% of the time under conditions constructed that random luck would expect a 50% success rate, that means that people are 20% more likely to be right than a coin is. (Yes, 20%, not 10%--probabilities are multiplicative, not additive, so you need to divide 60 by 50, not subtract 50 from 60.) But that person will still be wrong 40% of the time. That's right around the cusp of an F versus a D-, depending on your school district. Not exactly something to write home about.