AI predicts whether gay or straight from photo (81% accuracy)

Discussion in 'Off-Topic Discussions' started by me again, Sep 8, 2017.

Loading...
  1. me again

    me again Well-Known Member

    Full story:
    https://www.theguardian.com/technology/2017/sep/07/new-artificial-intelligence-can-tell-whether-youre-gay-or-straight-from-a-photograph
     
  2. SteveFoerster

    SteveFoerster Resident Gadfly Staff Member

    So, that's extremely strong evidence for nature rather than nurture. Thanks for sharing!
     
  3. Maniac Craniac

    Maniac Craniac Moderator Staff Member

    I would disagree with this as I could possibly come up with at least half a dozen reasonable social science and statistics based arguments against the assertion. In fact, it would be very easy to spin the perspective to the equally viable assertion that this is extremely strong evidence for nurture. :scratchchin:

    I would agree that it is very meaningful data, if accurate, which I won't spend time trying to verify as I lay in bed typing this on my laptop, eagerly waiting to count the last sheep. :biggthumpup:
     
  4. SteveFoerster

    SteveFoerster Resident Gadfly Staff Member

    You think that gay men tend to have narrower jaws, longer noses and larger foreheads because... of the way they were raised?
     
  5. cookderosa

    cookderosa Resident Chef

    I'm going to get in and get out before this thread turns into a fire storm- but here's a thought - people need to stay the *bleep* out of other people's lives. Seriously.
     
  6. decimon

    decimon Well-Known Member

  7. me again

    me again Well-Known Member

    Researchers and students at Stanford University are the ones who have delved into the topic using a combination of algorithmic fashion design (nurture) and algorithmic physical design (nature).

    Styles come and go. If specific styles are currently associated with gaydom, then those styles can be inputted into the algorithm.

    If people are born to look a certain stereotypical way, then it could become a psychological self-fulfilling prophecy for some (not all) of them.
     
  8. 03310151

    03310151 Active Member

    Article says 91%. But in reading the article it looks like, apparently, they pulled the pictures from dating profiles of people who self-identified as "gay".


    Active steps to engage in sex with someone of your own sex would be a definite "tell".
     
  9. me again

    me again Well-Known Member

    That's an excellent observation. Instead of pulling photos from dating profiles (where people identify as being gay), it would be interesting to pull photos from:
    - a drivers license database
    - or from a high school yearbook
    - or from some other neutral database.

    The ability of AI to have such high probability statistics would invariably be significantly reduced if a neutral database was used for their experiment. However, it would be near impossible to use a drivers license database, due to privacy issues.
     
  10. Maniac Craniac

    Maniac Craniac Moderator Staff Member

    What I think is entirely irrelevant to the question.

    Also, for one of the major propositions of the nurture argument, you have it exactly backwards.
     
  11. Johann

    Johann Well-Known Member

    My son once had a cat that loved the company of gay men. Remarkable animal. I'm told he "knew" once or twice, before the man did!

    J.
     
  12. Kizmet

    Kizmet Moderator

    I haven't really read the article so forgive me if these factors are discussed in there somewhere but our conversation has raised two questions in my mind. The first is that it would almost certainly be interesting to see the predictive power of the tool if a global population were used as the sample. If, for example, they used a collection of people from Mongolia, or Iceland or Japan, etc. would they get statistically identical/similar results? Secondly, I'd want to point out that the gay/straight binary is a myth. There are a number of points on that sexuality continuum, the most obvious being people who are bisexual. So did they sort people into "strictly straight" v. "anything other than strictly straight" categories or something other than that?
     
  13. cookderosa

    cookderosa Resident Chef

    But then the researchers would have to contact the participants, disclose to the participants that they are attempting to identify their sexual preference, get informed consent, the whole ball of wax.
    In the cases where they found people who self-identify on a website, I'm guessing there was no contact between the researchers and subjects necessary - certainly takes out a lot of red tape.
     
  14. Abner

    Abner Well-Known Member

    Ay Johann! He he! :smile:
     
  15. Johann

    Johann Well-Known Member

    (1) Kudos to Kizmet, for the above.
    (2) Kudos to decimon, for his reference to now-debunked phrenology. Very much à propos.
    (3) Kudos to Jennifer for her comment about staying the *bleep* out of people's lives. I sometimes forget this.

    J.
     
  16. Bruce

    Bruce Moderator

    There's a novel concept!!

    If someone wants me to know their sexual orientation (not that I even care), they'll tell me. Otherwise, it's none of my (or anyone else's) business.

    What next, AI can tell if you're liberal or conservative?
     
  17. decimon

    decimon Well-Known Member

  18. Bruce

    Bruce Moderator

  19. me again

    me again Well-Known Member

    While researchers may want to delve into the issue of sex, it remains, nonetheless, a very personal issue for many people, as is demonstrated by this extremely sad story:

    https://www.washingtonpost.com/national/suicide-of-teen-who-made-sex-video-shows-dilemma-for-schools/2017/09/12/5d4f419c-97d1-11e7-af6a-6555caaeb8dc_story.html?utm_term=.b9c0049bdecb
     
  20. cookderosa

    cookderosa Resident Chef

Share This Page