Inclusivity and diversity are the answer to bias-free AI in smart beauty technology
Two tech and beauty companies have come together in promoting inclusivity and diversity as the answers to bias-free AI in smart beauty technology.
“The limit in diversity does not exist”
“AlgoFace’s mission and goal is to build the most accurate unbiased face AI in the world, and this messaging is all about inclusivity.
“We want our products to work on all skin tones, people of all physical abilities, people of all genders. I like to say that the limit in diversity does not exist,” says Lui.
AI and Augmented Reality (AR) are helping consumers in the beauty industry get the exact shade of makeup they need.
“Certainly, bias is a big issue in this space. We are very aware of this. However, we are a very diverse company in all aspects, be it, gender, race, etc.,” says Alashkar.
“We are aiming that our AI selection be unbiased by ensuring that we fill all the gaps in our data”
AlgoFace, a division of Algo, Inc., and Lui’s beauty tech startup Nudest have announced a merger to bring a smart and trustworthy solution for shade matching.
What AlgoFace Does
Lui’s New York based startup Nudest invented the proprietary AI called Nudemeter, which lets online shoppers match their skin color online to the right shades of makeup for their unique undertones and seasonal tanning preferences.
Nudemeter will be integrating with AlgoFace’s virtual makeup try-on technology.
AlgoFace has launched three core products to date:
- A facial landmark tracking software development kit that detects and tracks the natural movement of the human face across 209 landmark points in real time
- A face attributes analyzer that accurately classifies the human face among 20 different traits
- A virtual makeup try-on solution that allows a user to try on foundation, blush, eye liner, eye shadow, lipstick, and more with AR
“AlgoFace provides skin matching technology that can recognize your skin shade in high resolution and combine the results on our facial makeup trial system. We unlock a lot of opportunity, making the experience of our user much more interesting,” explains Alashkar.
The AI in AlgoFace is basically a computer making recommendations and decisions on behalf of the customer, pointing out products that work for the customer’s skin.
“It’s our constant effort to make our technologies as inclusive and unbiased as possible. That’s our mission, our North Star”
The Machine Learning aspect is teaching the AI how to make those decisions. “To teach our machine learning technology how to take into account a person’s undertone and tanning preferences in order to make a recommendation, over the past three years of building Nudemeter, we have trained our technology on images of people and also tested it out. We had to get a lot of matches wrong before we got it right,” says Lui.
Lui is particularly excited about the AR part, which is a new aspect that the merger gains,
“While the Nudemeter identified products that work for you, now we can also empower the user to see what the matches look like on their face,” she says.
Face Detection Not Face Recognition
Face recognition has been receiving flak in several cases for either displaying bias or being unreliable. AlgoFace claims its technology is not face recognition but face detection.
“When I think of face recognition, I think of being able to identify somebody. Its applications are more likely to be seen in the security industry, where you’re trying to identify a criminal or a Person of Interest,” says Lui.
“Certainly, bias is a big issue in this space. We are very aware of this”
“We’re tracking the face in pretty awesome ways, from a two-dimensional, as well as a three-dimensional perspective, wherein, you are on a live camera, able to move your eyes and lips, shake your head, and track where your features are,” she adds.
Alashkar explains that AlgoFace is determined that their AI remains bias free.
“From day one, we have tried to keep our AI bias free. Bias in the training data set will occur when there is a lack of other skin colors. For example, if the data sets have mostly white skin color, it will help users of that skin color more. However, our sets have all the skin colors,” he says.
“We are collecting diverse sets. We are aiming that our AI selection be unbiased by ensuring that we fill all the gaps in our data,” he adds.
Another aspect of smart tech that worries consumers today is privacy. Lui says that is an aspect that AlgoFace takes into consideration.
“I’m proud of our technology, because it really takes consumer privacy into concern. We are able to allow someone to use our technology without baiting their data unnecessarily. It’s our constant effort to make our technologies as inclusive and unbiased as possible. That’s our mission, our North Star,” she says.
This is totally unacceptable conduct from a Google contractor. It’s why the way AI is built today needs to change. The answer to algorithmic bias is not to target the most vulnerable. https://t.co/VBzC8JTONr
— Jake Snow (@snowjake) October 3, 2019
The AlgoFace team has people from nine countries, 63% of it being women. That can vouch for several perspectives its AI AR face detection can gain from. As Lui says:
“The industry wants shade matching technology for the skin, but the industry also wants the ability to understand what the product would look like on their face. So we’re really excited about this unique combination of inclusion and diversity in skin tone matching and color matching as well as virtual try on.”
Disclosure: This article includes a client of an Espacio portfolio company.