By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Ganoncrotch said:
numberwang said:

If you have a group of 100 people with 10 gays in it and you just say everybody is straight, it would give 100% false statements for the gays.

AI would be correct 90% of the time though.

The AI isn't to tell if someone is Gay, if you read the OP it's "Artificial intelligence can accurately guess whether people are gay or straight based on photos of their faces"

It would be correct 90% of the time by saying that a large enough sample size was all Straight, it would just be wrong for each gay person.

You do not construct statistical tests like that for the obvious reason. If you have the paper we can look what exactly they did. One way would be to test an equal number of straight and gays photos and see how often the test is accurate.