Google I/O 2022: AI-powered real-world searches, 10-shade skin tone scale, and more

Google, owned by Alphabet Inc., announced plans on Wednesday to use artificial intelligence to bridge the gap between the natural world and its digital universe of search, maps, and other services.

Google showcased a product that would let customers record videos of shelves of wine bottles in a shop and ask its system to automatically discover selections from Black-owned vineyards during its annual I/O developer conference.

“It’s like having a supercharged Ctrl+F for the world around you,” Google senior vice president Prabhakar Raghavan said, referring to the keyboard shortcut for locating anything in a document. “You may look around your whole universe and ask questions in whatever manner you choose.”

Google Maps will introduce an immersive view for select major cities later this year, combining Street View and aerial photographs “to build a rich, digital representation of the globe,” according to Google.

Google also revealed a 10-skin-tone palette, which is a step forward in creating devices and applications that better serve people of color.

The company claims that its new Monk Skin Tone Scale replaces the Fitzpatrick Skin Type. This flawed six-color standard had become popular in the tech industry for determining whether smartwatch heart-rate sensors, artificial intelligence systems with facial recognition, and other products show color bias.

Fitzpatrick underrepresented persons with a darker complexion, according to Tech experts. Last year, Reuters claimed that Google was working on an alternative.

The startup teamed up with sociologist Ellis Monk of Harvard University, who studied colorism and felt dehumanized by cameras that failed to recognize his face and reflect his skin tone.

Fitzpatrick, according to Monk, is excellent at recognizing distinctions in lighter skin. However, since most people are darker, he wants a scale that “does a better job for the bulk of the planet,” he said.

Monk picked 10 tones using Photoshop and other digital art tools, a workable quantity for those who train and evaluate AI systems. He and Google polled 3,000 individuals throughout the United States and discovered that a large percentage of them felt a 10-point scale fit their complexion just as well as a 40-shade pallet.

The Monk scale is “a nice compromise between being representative and being tractable,” according to Tulsee Doshi, director of products for Google’s responsible AI team.

Google is already using it. Beauty-related Filtering results based on Monk are now possible in Google Images searches like “bridal makeup looks.” Images with varied skin tones are now shown when searching for “cute infants.”

The Monk scale is also being used to guarantee that a wide spectrum of individuals is happy with Google Photos’ filter choices and that the company’s face-matching tech is not prejudiced.

Still, if firms don’t have enough data on each of the tones, or if the people or tools used to assess others’ skin are prejudiced by lighting variances or personal views, flaws might creep into goods, according to Doshi.