Pages

Google is working on promoting MST scale to make AI skincare products work much better

Google is known to unveil a partnership with Harvard’s research associate professor, Ellis Monk, who has developed a more comprehensive and defined method of characterizing the various skin tones on a scale. Google aims to make the Artificial Intelligence-based skincare apps less biased by improving its search algorithm for people with darker skin tones.

Over the coming months, Google will launch a 10-step Dr. Monk skin tone scale, providing diversified images and different color shades in the search results. Google tries to replace the previous, musty sort of skin scales that misguide people who have olive skin.

Tech products are incorporated with outdated image-based software with a Fitzpatrick phototype scale for human skin colorations. The problem is that the Fitzpatrick scale was developed in 1975, having six categories for skin tones, and it was designed for burns and tan skins. It does not provide much room for darker skin tones. Due to these reasons, Google plans to train machine vision programming for up to ten skin tones only available on MST scale.

To overcome the AI bias and help brands design inclusive products and cater to a large number of people with tan skin. Furthermore, Google seems to develop a systematic method to label digital webs to enable companies to publish their stuff showing hair color and skin texture precisely. In this way, the search giant will understand imagery with ease. Along with that, Google will modify its AI apps to enhance the performance of models designed to detect features in images.

Google encountered a massive failure when implementing a feature like this before. For instance, when black people searched using Google Photos, they would encounter irrelevant results and the algorithm would relate them with chimpanzees and gorillas, which was highly offensive. It’s one of the reasons why features like this are being implemented so late. This time, to improve diversification across various skin tones and races, Google is taking steps towards improving the AI.

MST Scale will help make the search results more accurate to what the user searched for and improve user experience on the search engine. It will remove the bias lighter skin tone standards that were set by the 6 tone Fitzpatrick scale. While all kinds of bias in face AIs remains unchanged, we feel that a step towards more diversified and accurate skin and hair tones is a step in the direction.

These sorts of feature always come with a social risk of upsetting any minorities or groups which is why it’s so rare to see them in action. Especially, a massive corporation like Google taking this step is a big surprise seeing as their reputation may go down if these updates are received poorly. Nonetheless, we can still judge the upcoming feature and keep in mind that it’s based on good intent, and for that, we feel that Google deserves some credit.


Read next: Google Prioritizes Security With Launch Of Features That Keep User Safety In Mind

No comments: