It is looking at clothes, including necklines - that's in your link. The researchers also mention "transient facial characteristics, such as grooming" and "expression" rather than the biometry that you imply... The writers stress several times that these are not all biometric attributes but transient attributes such as grooming, expression and clothing. They note that lesbians wear less make up, have higher necklines and tend to wear baseball caps (in their results, that is) and that gay men have lighter skin and less facial hair. From there the algorithm decided who was more feminine/masculine than the gender of the image and partly based its decision on that
Yea, I stopped short of saying that it's entirely looking at biometrics. One article I read on it did suggest that it weighed items that can be changed easily less than items which could not. I guess you got me that baseball caps and necklines are clothing. It's not what I meant, but you're technically correct.
Most of what it has to work with is the face itself. Even if the results were entirely based on grooming and expression (which they aren't), that can still be used to support some of the conclusions I drew earlier. The point is that it may be feasible to train a machine learning algorithm on a known dataset to find gay people from within a population. Whatever cues it uses, and they wouldn't even necessarily be the same from one culture to another, it can determine what to look for.
But the biological component of this is really what's interesting to me. It could fuel a bunch more research into embryo development, brain chemistry, and change the way portions of the population and world governments think about homosexuality.
The pre-experiment projection was 6/10 (male) 5.5/10 (women) in the 2009 figures with a higher accuracy achieved in 2013 figures. Those two accuracy measurements were not from the same set of images. On single images the regression logarithm scored 8.1 (male) and 9.4 (female), it was more accurate when given five images of each subject.
....at which point it obtained 91% accuracy for men.
So... why the strong negative reaction? This is interesting stuff.
If you inherited it the bloodline wouldn't go very far...
I wish genetics worked that way exclusively. There are plenty of genetic disorders such as down syndrome that arise simply from mutations. That's not to say down syndrome can't be propagated to children of down syndrome suffers, that's to say that people with down syndrome aren't out there having tons of kids. There are so many syndromes resulting from genetic disorders that are not carried by the parents, but which result in children with those genes. It's staggering and depressing when you really get into it.
I don't mean to pick on you, but people have a false sense of security when it comes to the assumed health of their offspring (for many reasons) that is based on the faulty assumption that the parents are "normal" so the children will be. That assumption discourages early testing in pregnancy, and testing prior to conception, to everyone's detriment.
Being gay is obviously not a genetic disorder (or any kind of disorder), but it could still be genetically transmitted, provided that people who carried it had enough offspring. I think the research suggests this is not the case, but just pointing out that it is possible to have more than one kid and thereby to propagate genes despite one of your offspring not having biological children. And of course, gay women can have biological children.
Edit:
Aw heck, I know a gay man who has biological children. Why did I not include that?
