An AI Tool Told Me I’m Pretty… But Then It Said Frankenstein Is, Too.
A bit of a wound to the ego...
Dear reader, The goal of this newsletter is to get you up-to-speed with the rise of AI, because I think you can play a crucial role in its developments. To stay informed, subscribe below.
Recently, my professor and I have been talking about narrow Artificial Intelligence and where it can be found in daily life. We both subscribe to a newsletter that sends a plethora of examples to your inbox daily, which examples range from driverless cars to technology that will help people get to the moon. Those are pretty in-your-face examples, but narrow AI is seriously everywhere. The newsletter is even written by an AI process called “Virtual Peter,” which Craig, my professor, absolutely loves. I think it’s weird.
So, when browsing for more everyday examples of narrow AI, I instantly gravitated towards fashion and beauty-related uses. I came across facial analysis software. I sent a couple of articles about it to Craig with the subject line, “is this actually cool or am I just vain??” (Which, coincidentally, I think would be a perfect name for my memoir if I ever decide to write one.) Fortunately, he agreed that it was, in fact, pretty cool. So I kept looking into it.
I quickly discovered Qoves Studio, a company that uses AI to analyze faces. People typically use it before getting plastic surgery. On their website, they have a software that allows you to upload a photo a photo of your face. Then their AI system picks apart all the flaws. If the flaw is significant enough, it’ll recommend products or procedures to fix it.
Obviously, I had to do this. So I got home from work one day and scrubbed off my makeup to take an adequate photo. I pulled up my hair, set my camera to self-timer, and went to work.
The AI was fairly gentle in its assessment. I had hunches and insecurities about what Qoves’ AI would suggest I change… but it went easy on me. It told me I had dark circles under my eyes and could benefit from lip fillers — two things I was already fully aware of before. This gave it a bit more credibility.
So at this point, I was invested. What else could AI tell me about my face? I wanted to know everything. I found another AI-driven facial analysis tool on “Face++.” The AI program analyzes your photo to provide a “beauty score,” or how your looks stack up among a huge database of other people’s faces. And when I say huge, I really mean it. The Chinese government even uses it for facial recognition and surveillance.
I was pleased with my “Face++” results (above). The percentage was much higher than I thought it would be, especially for such an unflattering photo. According to this AI, I was in the top 10% of attractiveness. I was a little skeptical, but mostly, I was relieved. Maybe my undereye circles weren’t as dark as I thought.
I had my boyfriend try it, and he got around 70%. I called Craig because I wanted him to try it. Even he was impressed with my “beauty score.” But he said we needed to compare it to something else so that we could really test it out. “Put in a picture of Frankenstein and see what it says,” he told me, already laughing. So I did.
And guess what? Frankenstein got just above 70%.
After consoling my boyfriend, I called Craig. Was this AI really trying to say that 7/10 women would give Frankenstein their phone number? That seems unlikely. And my boyfriend is ranked just as attractive as a literal green monster? As if. And,the worst part, Frankenstein’s “beauty score” totally delegitimized mine. Craig laughed and laughed and said he wasn’t impressed anymore. “I mean, he has nice bone structure,” I said defensively.
Obviously, beauty is subjective and AI only comprehends the data it’s been trained on. But if a company worth $4 billion says that Frankenstein is more attractive than 70% of people, what kinds of faces has it been looking at? Is it really trying to tell me that I am just as likely to find Frankenstein as attractive as my boyfriend? This seems significant.
So while “Face++” estimates that my attractiveness would be perceived by women as 92% greater than the rest of the population, and 89 % greater than the rest of the population by men, how is this even decided?
I really believe that there needs to be a person with a liberal arts background involved with the development in these algorithms. Yes, of course, the entire backbone of the process is the person to figure out the technological programming of the algorithms. But there needs to be an equal counterpart: a person who will ask these softer, out-of-left-field questions — because situations like these make it evident that a liberal arts perspective is crucial.
While facial analysis software is everywhere, some are transparent and usable with a quick Google search. Some offer a service (like Qoves Studio disclosing how plastic surgery can make you more “attractive”). It’s been reported that it’s the foundation of many personalized social media algorithms. And if that’s the case, the AI behind your Instagram feed may be prioritizing Frankenstein lookalikes a little higher than you might expect.
Recent reads:
A brief roundup of articles that fascinate me…
Futureloop, the newsletter I mentioned above, is a great aggregator for all things AI…
2. This article has great in-depth reporting on facial analysis software, if you’re looking to dive deeper and get a more technical understanding…
Final thoughts:
I consider myself to be a fairly seasoned aficionado of trash TV, specifically the kinds of reality shows where I get to watch people make huge life decisions on-screen. But I haven’t been able to get my fix since Love Is Blind premiered around a year ago.
The good news: Netflix just released a cringe-worthy, tone-deaf reality series where couples have to choose between purchasing their dream wedding or their dream home. It’s called Marriage or Mortgage, and it is so dumb, and so fun.