You are on page 1of 2

A Chink in The Armor: Microsoft, Phasing out Emotion Identification AI

AI tech is perhaps one of the best technological innovations in the 21st century. Its specialized
application to different industries procures benefits beyond any other tech. It is an efficient piece
of technology that can potentially revolutionize various fields, scientific or industrial. 

However, it remains flawed. While it is potent, AI still lacks flexibility compared to human
cognition. That is perhaps the biggest flaw tech giants recognize, which affects their decision-
making process when introducing an AI tech to the public. 

That was the motivation behind Microsoft's decision to phase out the public access to their AI-
powered facial analysis products. 

Microsoft's AI Facial Recognition Tool: Azure Face API

Microsoft's Azure Face API offers efficiency and accuracy. It delivers a low-friction, state-of-
the-art facial recognition system that perceives facial features and attributes, even accessories
such as face masks, glasses, and face location. It can recognize someone's facial blueprint by
matching the ID via a private repository or photo ID. 

While that is common in a facial recognition tech, it possesses another imperative element;
emotions. One of Microsoft's AI-powered facial analysis products claims to be capable of
determining emotions. They purport that it can dictate emotions based on videos and images of
participants. However, that single element that can potentially revolutionize facial recognition is
also the reason for its fall. The reason is: that it's not flexible and accurate. 

Issues of the AI-powered Facial Analysis Product

Being able to recognize an emotion through AI is potent. It opens different applications for both
marketing and security. However, the element has become the most significant flaw of the
product. That is because a universal facial expression is simply non-existent. 

Emotional manifestation through the face differs across various demographic. Plus, it is
unscientific to equate exterior displays of emotion with inside feelings. While the tech can
recognize a 'frown,' it remains dissimilar from recognizing "fury." 

"Experts inside and outside the company have highlighted the lack of scientific consensus on the
definition of ‘emotions,’ the challenges in how inferences generalise across use cases, regions,
and demographics, and the heightened privacy concerns around this type of capability," Natasha
Crampton wrote, the Microsoft’s chief responsible AI officer.

The lack of universal emotion causes the tech to misidentify feelings. Thus, casting doubt on its
efficiency. 

So, Microsoft has no choice but to discontinue this specific part of the product. Interestingly, it
coincides with a major revision of Microsoft's AI ethics and guidelines. 
Phasing Out Process
Microsoft deems to take the process methodically. They will initially remove some parts of the
facial recognition services from "Azure Face" and employ a restriction on other users. They will
also phase out Azure Face's capabilities in identifying attributes such as gender, smile, age, hair,
facial hair, and makeup.

As per Microsoft's announcement, existing users will lose access to the features starting the half
of next year, June 30, 2023. However, new customers can no longer utilize them starting June
21. 

The tech remains open to new users. However, they need to file an application and provide
specific information regarding their deployment plans in Microsoft. While they can't access the
features mentioned above, elements like automated face blurring in pictures and videos remain
open to the public. 

Interestingly, discontinuation of the feature will only apply to the public. Microsoft's Seeing AI,
an app for the optically handicapped, will continue to utilize the feature. 

You might also like