News By Tag
News By Location
Beauty.AI 3.0 contest judged by robots postponed due to ethical concerns
Plans to form a task force to prevent discrimination by Artificial Intelligence.
For many people perception of their appearance is often more important than the many other issues related to their individual performance, contributions to society, well-being and even survival. The Beauty.AI competition was originally envisioned as an entertaining and engaging research project to develop a framework for the impartial analysis of human appearance and allowing algorithm developers worldwide to test their approaches. The first Beauty.AI competition was launched in November 2015 and concluded in January 2016 with several thousand participants worldwide. Beauty.AI 2.0 launched in April 2016 and concluded in August with over 6,000 humans evaluated by a panel of five algorithms based on different theories on how human beauty may be evaluated utilizing various machine learning techniques.
After the winners had been announced, Sam Levin, a reporter for "The Guardian" pointed out the perceived racial bias of the contest, since few of the winners had dark skin. The story rapidly went viral and resonated worldwide resulting in hundreds of critical emails to developers.
"Our team is part of a collective of computer scientists and biologists religiously committed to extending human longevity and preserving humans in the youthful, healthy state for as long as possible. But surprisingly, many people care more about their looks than the age-related pathologies they develop and we decided to focus on appearance and beauty instead to get a mass market appeal and developed a platform, where human appearance could be evaluated using various techniques with frequent updates. This is how RYNKL and Beauty.AI were developed. But we did not expect that anyone would take the results of our beauty competition seriously, and we made it clear that as the organizers we do not agree with these results and would like to welcome more algorithmic approaches to join the effort. But journalists quickly noticed the lack of darker people among the winners and it created a media frenzy resulting in negative publicity and warning letters from collaborators, partners and even companies that got associated with the contest by mistake", said Anastasia Georgievskaya, head of beauty research at Youth Laboratories.
Several algorithms used in Beauty.AI 2.0 competition utilized deep learning, a popular biologically-
"The media attention received by our experimental beauty contest in the context of racial discrimination demonstrates the blessing and a curse of contemporary journalism. From one perspective, these claims are not justified, but from another perspective, they pointed out the major issue associated with machine learning and potential for discrimination. And while many AI safety activists are concerned about machines wiping us out, there are very few initiatives focused on ensuring diversity, balance and equal opportunity for humans in the eyes of AI", said Konstantin Kiselev, CTO of Youth Laboratories.
The collective of scientists involved with the Beauty.AI contest will form a task force to develop a set of guidelines to prevent or reduce the chance of racial, gender, age, disability and other discrimination by the artificially-
"Discrimination by AI is likely to be a much greater threat shortly than AI going berserk and eliminating us. We already see it in other projects in healthcare, like Aging.AI, where the insufficient representation of older people in the training sets leads to the poor accuracy of age prediction and the need to exclude the oldest old from experiments. And if you search for "white person" + keyword and "black person" + keyword in any image search engine, you are likely to see very different results", said Alex Zhavoronkov, Ph.D., CSO of Youth Laboratories and CEO of Insilico Medicine.
"We thought we took every precaution to ensure that there is no discrimination by the algorithms and prevent gender and age bias and one of the official partners of the contest is Imagene Labs from Singapore, which is one of the leaders in facial analysis research focusing on the Asian population. While we were not discriminating anyone in Beauty.AI, the press opened our eyes on a very important issue, and we would like to ensure that we take extra precautions before launching the next contest or hackathon", said Alexey Shevtsov, CEO of Youth Laboratories.
About Youth Laboratories
Youth Laboratories is a research and development company applying for advances in machine vision and sensing and artificial intelligence to improve human health, longevity and appearance. It also strives to explore new levels of human-machine interaction by developing intelligent algorithms for analyzing human attractiveness and emulate the emotional response ("feelings") to the various features related to human appearance. The collective is dedicated to helping people retain youthful state for as long as possible using advances in machine vision and artificial intelligence. The company develops series of mobile applications that track age-related changes, wrinkles, pimples, dark spots and other parameters affecting the perception of beauty, health, and youthfulness and help evaluate the effectiveness of multiple interventions. The company's first application is called RYNKL, developed using the funds crowdsourced via Kickstarter.