How 10 skin tones will reshape Google’s approach to AI

For many years, technology companies have relied on something called the Fitzpatrick scale to classify skin tones for their computer vision algorithms. Originally designed for dermatologists in the 1970s, the system only covers six skin tones, a factor that may contribute to AI’s fully documented failures in identifying people of color. Now, Google is starting to incorporate a standard 10 skin tones on its products, called Monk Skin Tone (MST) scale, from Google Image Search to Google Photos and more. This development has the potential to reduce bias in data sets used to train AI in everything from healthcare to content moderation.

For the first time, Google signaled plans to go beyond Fitzpatrick last year. Internally, the project began with an effort in the summer of 2020 by four Black women at Google to make AI “work better for people of color,” according to a report. Twitter thread from Xango Eyeé, a responsible AI product manager at the company. Today Google I/O Hội Conference, the company detailed how the new system will affect many of its products. Google will also open-source MST, which means it could replace Fitzpatrick as the industry standard for judging the fairness of cameras and computer vision systems.

“Think of wherever images of people’s faces are used, where we need to check the fairness of the algorithm,” says Eyeé.

The Monk Skin Color Scale is named after Ellis Monk, a Harvard University sociologist who spent decade studies the impact of colorism on the lives of Blacks in the United States. Monk created the scale in 2019 and worked with Google engineers and researchers to incorporate it into the company’s product development process.

“The reality is life chances, chances, all of these things are very much tied to your phenotype,” Monk said in prepared remarks in a video shown at I/O.” We are able to eliminate these biases in our technology at an early stage and make sure the technology we have works equally well on all skin tones. I think this is a huge step forward.”

An initial analysis by research scientists from Monk and Google last year found that participants felt MST performed better than on the Fitzpatrick scale. In one Frequently asked questions announced Wednesday, Google says that there are more than 10 skin tones that can add complexity with no added value, unlike industries like makeup, where companies like Rihanna’s Fenty Beauty supply level more than 40 shades. Google is continuing to work on validating the Monk’s Skin Color scale in places like Brazil, India, Mexico and Nigeria, according to a source familiar with the matter. More details are expected in an academic research paper soon.

The company will now expand the use of MST. Google Images will provide an option to sort makeup-related search results by skin tone based on a scale, and filters for people with more skin pigmentation will appear on Google Photos next year. at the end of this month. If Google adopts a 10-skin color scale across its product lines, it could have implications for the fairness evaluation algorithms used in Google search results, Pixel smartphones, YouTube classification math, Waymo self-driving car, etc

Source link


News7g: Update the world's latest breaking news online of the day, breaking news, politics, society today, international mainstream news .Updated news 24/7: Entertainment, the World everyday world. Hot news, images, video clips that are updated quickly and reliably

Related Articles

Back to top button