It's Hot Or Not: Deep Neural Network Edition: Scientists at the Computer Vision Laboratory at the Swiss Federal Institute of Technology in Zurich (ETH Zurich) are developing artificial intelligence software that taps into deep neural networks (or hierarchical learning structure based on a set of algorithms) to determine whether someone is considered attractive or not.
In collaboration with the dating app Blinq, the ETH Zurich research team is working on its algorithm with the help of a beta called howhot.io; the beta essentially functions like the throwback site Hot or Not by using data sets rendered from 2D- and 3D-imaging. Users can upload their photos onto the site, and the AI algorithm rates you on a six-point rating scale, from "Hmm ..." to "Godlike." Unlike Hot or Not, however, the use of the beta is not for public consumption, but for users to pick out which photo to use as a main photo for their profile.
How does the beta take into account that there is no such thing as uniform beauty?
"Attractiveness is highly subjective and its perception differs from culture to culture," the scientists wrote on their site. "Our algorithm is trained on the pictures of the Blinq community that is mainly based in Switzerland. In other parts of the world the perception might be very different."
The AI structure also had a plethora of data to help inform it of what constitutes attractiveness.
"We used more than 100,000 images and more than 20 million ratings between users from our database," Blinq co-founder Jan Berchtold said in an interview with Tech Crunch.
The project is part of an initiative to have "difficult, real-world applications drive our research and development," and its partnership with Blinq as an imperative "collaboration with [a settled industry] as an important plus for an engineering lab like ours."
Photo: Gabriel Saldana | Flickr