The Facial Rec Tech Wreck
Facial recognition technology has been taking it on the chin lately (pardon the pun). Earlier this week, the BBC reported that a UK court ruled the use of the technology by British police violated human rights and data protection laws in that country. A week before that, a team of researchers at the University of Chicago unveiled Fawkes, an algorithm and software tool that makes pixel-level changes to your image that are invisible to the human eye, but effectively mask you from the current crop of facial recognition applications. And back in July, Amazon, Google, and Microsoft were sued over claims they used photos of individuals to train their facial recognition software without getting prior consent, which violated an Illinois biometric privacy statute. (Facebook had already settled a class-action claim that it also violated that law.)
Remember when we were all thrilled that we could open our phones with our faces?
One of the most important things to keep in mind as you're thinking about the future of the facial recognition technology industry, Gartner analyst Nick Ingelbrecht told me, is that it's not a single monolithic market today, and it never has been.
"It's made up of lots of different segments and use cases," he said, "ranging from 1:1 verification of customers' identities to border control screening; mobile payment verification to password replacement; one-to-many face matching for building access control to the many-to-many systems the police use."
Ingelbrecht is a research director with Gartner's Technology and Service Provider Research group. He focuses on computer vision, emerging trends and technologies, video and image analytics, and physical security. I asked him about recent developments in the facial recognition technology market, and announcements by Amazon, Microsoft, and IBM that they would be curtailing their efforts around this technology.
"The large US technology companies have been backing away from marketing facial recognition products for some years now," he said. "The recent announcements are just the latest in a series of retrenchments by large US tech firms to avoid the reputational risks and potential legal exposures associated with misapplication of biometric technologies. These decisions tend to slow commercial adoption generally, as well as the investment return on research. That said, facial recognition is not going away. Technologies will continue to mature, and research continues outside the US, especially in the People's Republic of China, where there is a very large internal market for facial recognition products."
Amazon has said it will stop selling facial recognition technology to police forces for a year. Microsoft, which doesn't currently sell facial rec tech to U.S. law enforcement, said it won't do so until the federal government passes a law regulating its use. And in a letter sent to Congress in June, IBM's CEO Arvind Krishna said his company has sunset its general-purpose facial recognition and analysis software products, and he called on Congress to regulate the use of the controversial technology by police.
Ironically, the most recent lawsuit against the three tech giants cited their use of IBM's Diversity in Faces Dataset, which the company developed to reduce racial and gender inaccuracies and biases in the technology. Big Blue released the dataset in January 2019 to the global research community "to advance the study of fairness and accuracy in facial recognition technology."
That effort was a response, at least in part, to conclusions by researchers from MIT and Stanford University in 2018, who found that commercially available facial-analysis programs from major technology companies demonstrated both skin-type and gender bias. Facial recognition is a type of image recognition technology that detects faces in captured images, and then quantifies the features of the image to match against a templates stored in a database. In the researchers' experiments, facial recognition algorithm errors in three leading solutions were 49 times more likely for dark-skinned women than white men. These results raised serious questions about how neural networks, which learn to perform computational tasks by looking for patterns in huge data sets, are trained and evaluated.
Clearly, facial recognition technology has come a long way since Woodrow Wilson Bledsoe began his pioneering work in a field he all but created back in the 1960s. It's now part of a class of biometric tech in increasingly widespread use, primarily in security applications, that includes fingerprint, iris, speech, and gate recognition. It's also worth a lot of money--billions, according to industry watchers. The global facial recognition market was valued at $3.4 billion in 2019 (according to a Grand View Research report) and is anticipated to expand at a CAGR of 14.5% from 2020 to 2027.
Biometrics continue to be used extensively across a range of security applications--primarily access control and attendance tracking. And recent headlines notwithstanding, the technology also continues to improve, evolve, and expand at an explosive rate. Advancements in artificial intelligence and machine learning have been applied to biometric technology, leading to increased accuracy and accessibility.
Despite recent controversies and what could be called growing pains, facial recognition technology isn't going away any time soon, Ingelbrecht said. However, the market is already changing.
"The current arguments about facial recognition in the US will put further pressure on small vendors and accelerate consolidation, aggregation, or exit at a time when the industry is suffering from a slump in buying activity, severe cash constraints, and supply chain difficulties," he said. "Outside the US, we expect facial recognition technologies to evolve at a rapid pace. It is ironic that in Europe, the GDPR [General Data Protection Regulation] has made commercial deployments of facial recognition very difficult, while governments and law enforcement there enjoy exemptions. In the US, there is extensive commercial use of facial recognition, especially in the retail and hospitality sector, but constraints are largely targeted at law enforcement."
Ingelbrecht's advice to the facial rec tech vendors during this volatile time: "They need to be very focused on their target market segments and differentiate themselves clearly via the business value they deliver to customers."
Posted by John K. Waters on August 13, 2020