After the algorithms were demonstrated to exhibit serious bias and mistakes, Microsoft Corp. decided to discontinue providing artificial intelligence-based facial-analysis software products that infer a subject’s emotional state, gender, age, mood, and other personalized features.
The tools’ existing customers can continue utilizing them for a year before their expiry. The company wants to ensure that the technologies satisfy the ethical AI guidelines of Microsoft, and hence are narrowing the utilization of other facial-recognition programs.
New customers must apply for access to facial-recognition features in Microsoft Azure Face API, Computer Vision as well as Video Indexer, whereas the existing customers have a year to renew their access.
The modifications were detailed in blogs authored by Chief Responsible AI Officer Natasha Crampton and Azure AI Product Manager Sarah Bird in conjunction with the issue of the second update to Microsoft’s Responsible AI Standard.
The modifications were introduced two years after Microsoft and Amazon.com Inc., whose cloud unit contends with Azure, halted sales of facial-recognition technology to U.S. police departments after research revealed that it did not perform well on subjects having darker skin. Both the tech companies have their headquarters in Washington, which together with a few states have passed laws regulating the utilization of such products.
Even as some of the world’s largest technology companies abandon the contentious technology, smaller firms such as NEC Corp and Clearview AI continue to thrive by selling facial-recognition tools for use in ways that raise privacy and security concerns, including by law enforcement.
Microsoft isn’t completely abandoning the use of AI to assist in reading human reactions. Other features that make educated guesses about people’s feelings or emotional states are being added by the company. Last week, Microsoft announced a new program for sales representatives that will utilize AI to run sentiment analysis on customer engagements on Microsoft’s Teams teleconferences to analyze how potential clients may be reacting.
