This site may earn affiliate commissions from the links on this page. Terms of Use.

When it comes to pioneers in the field of color equality, Google is not at the top of many lists. But there is a contingent in the company that is trying to change that. At its I / O 2022 conference, Google unveiled a tool it intends to use to improve color equality through presentation. This is a set of ten color patterns that match the tones of human skin, covering the entire range from very light to very dark. And Google made it on the spot with open source.

Justice is a major problem in machine learning. It is already difficult enough to reduce human values ​​to an algorithm. But there are different ones kinds in fairness, twenty-one or more, according to one researcher. Statistical justice is not the same as procedural justice, which is not the same as distribution justice. What do we do when different definitions of justice are mutually exclusive? Instead of trying to write a formula that governs them all, Google has taken a different approach: “Start where you are.”

Where we are is in a state of desperately unequal digital performance. Google has been the largest search provider on the planet for a long time. Do an incognito search on Google Images for “CEO” and what you will get is a sea of ​​white male faces, two of whom are Elon Musk. Look for a “woman” and it is absolutely true that the results distort young, slender, white, healthy. But one of the people the search returned was a deep falsification of a pale young woman generated by NVidia’s StyleGAN. I wrote about this specific dip before in another article, so it surprised me to see her face again. I had to check if I was incognito – but I was.

There are seven billion people on this planet, and most of them are people of color. There is some poetry in the idea that Google’s search algorithm, instead of showing a brown or black man, would prefer to bring back a woman who doesn’t really exist.

Introducing the monk skin tone scale

The ten-tone scale was developed by Harvard sociologist and ethics specialist Dr. Ellis Monk in collaboration with Google. “In our study, we found that most of the time people feel they are grouped into racial categories, but there is all this heterogeneity with ethnic and racial categories,” Dr. Monk said. statement. “And many categorization methods, including scales for past skin tones, ignore this diversity. This is where a lack of representation can happen … we need to refine the way we measure things so that people feel represented. “

The monk’s skin tone scale. Image: Google / Dr. Alice Monk

Google has announced that it will use the Monk Skin Tone Scale (MST) to improve racial and color performance in search results. There, the scale will make access much easier, for example to information about the colors and textures of black hair. And the colors are not named – none coffee with milk or chocolate comparison. (Are you listening, Pantone?) But the company is also embedding it in Google Photos, where it will be part of “a new set of real-tone filters that are designed to work well in skin tones and rated using the MST scale.”

In addition to using the MST scale to improve color evenness, Search Titan outlines plans for a “standardized way to label web content. Creators, brands and publishers will be able to use this new inclusive scheme to label their content with attributes such as skin tone, hair color and hair texture.

Google intends to introduce MST / Real Tone features in Android, iOS and web services in the next few months.

Now read:



https://www.extremetech.com/extreme/335462-google-introduces-monk-skin-tone-scale-to-improve-ai-color-equity

Previous articleBAE Systems Australia to supply sub-nodes for the ESSM program
Next articleIBM has struck a deal with Amazon Web Services to expand Big Blue’s hybrid cloud offering