Beauty pageants have never exactly been a platform for diversity and it looks like introducing AI doesn’t fix the problem.
The company Beauty A.I. recently hosted its second-ever international beauty pageant judged by AI.
The five robot judges scored more than 3,000 male and female contestants based on five categories: wrinkles, skin clarity and pigmentation, their similarity with models within their racial group, symmetry and chronological versus perceived age. Not altogether different from metrics used by traditional beauty pageants (though there was no talent component), the AI judged beauty based on qualitative metrics versus human biases.
Or so reasearchers thought. The contestants represented 100 countries, yet only one of the 44 winners was of a presumably darker skin color. The rest of the winners were predominantly caucasian or Asian.
“What matters in beauty is perception. Perception is how you and other people see you, and this perception is almost always biased. Still, healthy people look more attractive despite their age and nationality,” Beauty AI states on its website.
Youth Laboratories, the company behind the contest, claims its mission is to use deep learning and massive semantic analysis to “find effective ways to slow down aging and help people look healthy and beautiful.”
“If you have not that many people of color within the dataset, then you might actually have biased results,” Alex Zhavoronkov, chief science officer of Beauty.ai told The Guardian. “When you’re training an algorithm to recognize certain patterns … you might not have enough data, or the data might be biased.”
Like with most cases of discrimination by AI, the racist pageant reflects human trends. There was recently a racist backlash in Japan when a half-Indian woman won Miss Japan. In the pageant’s nearly 100-year history, only eight black women have been crowned Miss America.
Beauty.AI boasts partners like Nvidia and Microsoft, whose best known AI was the racist and sexist chatbot Tay.