Technology giant Microsoft says it will remove face detection, analysis and recognition from its own AI service Face API. They will be unavailable to new users this week. For cash, access will be suspended for one year. This is reported by The New York Times.
The changes were the company’s response to concerns from activists who argued that facial analysis software capable of identifying a person’s age, gender and emotional state could be harmful and should not be sold.
Earlier, Microsoft took a course to tighten control over artificial intelligence systems. After two years of research, its team has created the Responsible AI Standard, a 27-page document that sets requirements for AI systems so that they do not have a detrimental effect on society. Among other requirements, systems must provide “up-to-date solutions to the problems they are designed to solve” and “similar quality of services for different demographic groups, including marginalized ones.”
Microsoft was particularly concerned about an emotion recognition tool that labeled facial expressions as anger, contempt, disgust, fear, happiness, surprise, and more.
“There are a huge number of cultural, geographical and individual differences in how we express ourselves. This raises concerns about reliability (technology performance), and also raises more questions about whether facial expressions are a reliable indicator of inner emotional state,” said Microsoft’s chief executive officer of AI.
An age and gender analysis tool and a means of recognizing attributes such as facial hair or a smile will also be unavailable. They could be useful for the visually impaired, but the company decided not to make these tools available at all.