For years, activists and academics have been concerned that facial analysis software that claims to be able to identify a person’s age, gender, and emotional state may be biased, unreliable, or invasive. should be sold.
Acknowledging some of these criticisms, Microsoft said Tuesday that it plans to remove these features from its artificial intelligence service to detect, analyze, and recognize faces. They will no longer be available to new users this week and will be phased out for existing users during the year.
The changes are part of a push by Microsoft for stricter control of its artificial intelligence products. After a two-year review, a Microsoft team has developed a “Responsible AI Standard,” a 27-page document that sets out the requirements for AI systems to ensure that they do not have a harmful impact on society.
Requirements include ensuring that systems provide “valid solutions to problems that are designed to solve” and “a similar quality of service for identified demographic groups, including marginalized groups.”
Prior to publication, technologies that would be used to make important decisions about a person’s access to employment, education, health, financial services, or a life opportunity are subject to review. part of a team led by Natasha Crampton, Microsoft’s head of AI. .
There was a growing concern at Microsoft for the Emotion Recognition Tool, which labeled someone’s expression as anger, contempt, disgust, fear, happiness, neutrality, sadness, or surprise.
“There’s a lot of cultural, geographical, and individual variation in the way we express ourselves,” Ms. Crampton. This caused reliability issues, along with bigger questions about whether “facial expression is a reliable indicator of your internal emotional state,” he said.
Read more about Artificial Intelligence
Removed age and gender analysis tools, along with other tools for detecting facial attributes such as hair and smile, could be useful in interpreting visual images for blind or partially sighted people, for example, but the company decided it was problematic to do so. Profiling tools generally available to the public, Ms. Crampton.
In particular, he added, the so-called gender classifier of the system was binary, “and this is not consistent with our values.”
Microsoft will also put new controls on its facial recognition feature, which can be used to perform identity checks or search for a particular person. Uber, for example, uses its application software to verify that a driver’s face matches the identifier in that driver’s account file. Software developers who want to use the Microsoft facial recognition tool will need to request access and explain how they plan to implement it.
Users will also need to request and explain how they will use other potentially abusive AI systems, such as Custom Neural Voice. The service can create a human voice impression, based on a sample of someone’s speech, so authors, for example, can create synthetic versions of their voice to read their audiobooks in non-spoken languages. .
Due to the possible misuse of the tool, to create the impression that people have said things they have not said, speakers need to follow a series of steps to confirm that their voice is authorized and recordings include watermarks detectable by Microsoft. .
“We are taking concrete steps to comply with our AI principles,” Ms. Crampton, who has worked as a lawyer at Microsoft for 11 years and joined the ethical AI group in 2018. “It’s going to be a great journey.”
Microsoft, like other technology companies, has had issues with its artificially intelligent products. In 2016, she launched a Twitter chat bot called Tay, which was designed to learn “conversational understanding” from the users she was interacting with. The bot quickly started sending out racist and offensive tweets, and Microsoft had to withdraw it.
In 2020, researchers found that text-to-speech tools developed by Microsoft, Apple, Google, IBM, and Amazon worked less well for blacks. Microsoft’s system was the best in the group, but it misidentified 15% of words for whites, compared to 27% for blacks.
The company had collected various speech data to train its AI system, but had not understood the extent to which the language could be diverse. So he hired an expert in sociolinguistics from the University of Washington to explain the linguistic varieties that Microsoft needed to know. It went beyond demographics and regional variety in how people speak in formal and informal settings.
“Thinking about race as a determining factor in how someone speaks is really a little misleading,” Ms. Crampton. “What we have learned in consultation with the expert is that a wide variety of factors actually affect linguistic variety.”
Ms Crampton said the trip to address this disparity from voice to text had helped inform the guidelines set out in the company’s new standards.
“This is a critical period for setting standards for artificial intelligence,” he said, noting that the proposed European regulation lays down rules and limits on the use of artificial intelligence. “We look forward to using our standard to try to contribute to the bright and necessary discussion we need to have about the standards that technology companies need to meet.”
There has been a vibrant debate for years about the potential harms of artificial intelligence in the technology community, fueled by mistakes and errors that have real consequences on people’s lives, such as algorithms that determine whether whether or not people receive welfare benefits. The Dutch tax authorities mistakenly withdrew childcare benefits to needy families when a faulty algorithm penalized people with dual nationality.
Automated software for recognizing and analyzing faces has been especially controversial. Last year, Facebook shut down its system a decade ago to identify people in photos. The company’s vice president of artificial intelligence cited “many concerns about the place of facial recognition technology in society.”
Several black men have been unjustly arrested after faulty facial recognition matches. And in 2020, at the same time as the Black Lives Matter protests following the police assassination of George Floyd in Minneapolis, Amazon and Microsoft issued moratoriums on the use of their facial recognition products by police in the United States. United States, saying clearer laws about its products. they needed use.
Since then, Washington and Massachusetts have passed regulations that require, among other things, judicial oversight over the use of police facial recognition tools.
Ms. Crampton said Microsoft had considered starting to make its software available to police in states with books laws, but had so far decided not to do so. He said that could change as the legal landscape changes.
Arvind Narayanan, a computer science professor at Princeton and a leading expert on artificial intelligence, said companies may be withdrawing from face-scanning technologies because they were “more visceral, unlike many other types of AI. which may be dubious, but we don’t necessarily feel it in our bones. “
Companies may also realize that, at least for now, some of these systems are not as commercially valuable, he said. Microsoft could not tell how many users it had for the facial scanning features it was getting rid of. Mr. Narayanan predicted that companies would be less likely to abandon other invasive technologies, such as targeted advertising, which profiles people choosing the best ads to show them, because they were a “cow of money.”