What a scam. Nobody is born gay. At most the cumputer can recognized some typical facial gestures that some homosexuals are prone to use, such as a more feminine smile.
There's plenty of evidence that being gay is at least somewhat related to their genetics (which would suggest they are born gay).
I read a very convincing post that homosexuality is a product of evolution.
There's plenty of heterosexual males that have more feminine features.
That's not all that impressive in context. They could have gotten a much more accurate predictor if they'd just had the algorithm say "straight" every single time. Then the accuracy would jump up to around 95-96%.
That assumes that the sample is representative of the population.
If they picked 50 gay men, and 50 straight men, the accuracy of an "always saying straight" algorithm would only be 50%.
This part in their notes suggests that it's not quite that simple:
"Back to the group of 1000 men, including 70 gay men. If one selected 100 random males from this sample, only 7 are expected to be gay: a random draw offers a precision of 7% (7 out of 100 selected men were gay). Let’s turn on the classifier. Among the 100 of individuals with the highest probability of being gay according to the classifier, 47 were gay (precision = 47/100 = 47%). In other words, the classifier provided for a nearly seven-fold improvement in precision over a random selection. There were also 53 “false positives” - straight men classified as gay. Note, however, that as there are only 70 gay men in the examined population, there would be 30 “false positives” even if the classifier was perfect. "