Amazon rejects revolt over its facial recognition technology
Its Rekognition technology is said to bear racial and gender bias, but it's a problem for lawmakers and not the company, it said
Amazon's shareholders have overturned an internal revolt over the sale of its Rekognition technology to the police by an overwhelming 97% majority.
There has been sustained criticism of facial recognition technology of late, particularly in the way it can show bias towards ethnic minorities and women. Civil rights groups have called for the technology to be reconsidered for use by the police where a mismatch could lead to a wrongful arrest.
"Amazon has faced increasing criticism over its relationships with key constituencies such as its workers and the communities in which it operates," the company said in a supporting statement.
"Amazon's surveillance technology has provoked an outcry from civil rights organizations and may have damaged our company's brand."
The company said that it would back and offer support to a government-led legislative framework, indicating that the onus shouldn't be placed on the company itself.
"To the extent there may be ambiguities or uncertainties in how existing laws should apply to facial recognition technology, we have and will continue to offer our support to policymakers and legislators in identifying areas to develop guidance or legislation to clarify the proper application of those laws," it said.
"We also support the calls for an appropriate national legislative framework that protects individual civil rights and ensures that governments are transparent in their use of facial recognition technology."
The vote regarding the sale of Rekognition to government agencies was held using a one share-one vote system. There were 8.3 million votes to ban the sale of Rekognition but these were eclipsed by the 327 million votes that opposed the proposal - just 2.4% of all votes were in support when more than 50% was needed for it to pass.
Votes were also taken on the proposal to commission an independent study into the risks presented by the bias exhibited in the technology to the disproportionate surveillance of people of colour, immigrants and activists.
This vote received greater support with 27.5% of votes in favour of commissioning the study, but this still fell well below the
The criticism of facial recognition isn't unfounded, many reports from different studies and bodies suggest that there are flaws with it in most cases where it's used.
Focusing on Amazon's Rekognition specifically, MIT researchers found that in 31% of cases, Rekognition mistakenly identified individuals through gender and ethnically biased algorithms.
"In internal accuracy tests of Amazon Rekognition's facial recognition features, AWS evaluated photos from a publicly available dataset of 1 million face images and found zero false positive matches at a 99% confidence level," said Amazon.
"AWS plans to work with industry and academic groups that specialise in computer vision to help establish additional standardised tests and benchmarks for cloud-based facial recognition technology."
It was also reported earlier this month that the New York Police Department (NYPD) had been clumsily abusing facial recognition technology, following a report by Georgetown Law's Centre on Privacy and Technology (CPT).
The reports showed that the NYPD would take a witness statement and then use that information to edit existing images the police had on file to see if it matched any person using facial recognition technology.
In one alarming case, the NYPD saw an alleged perpetrator that looked like the actor Woody Harrelson so just plugged an image of the actor's face into facial recognition software which ultimately led to an arrest. It's believed the lookalike wasn't charged.
"The stakes are too high in criminal investigations to rely on unreliable-or wrong-inputs," said Clare Garvie, author of the Georgetown Law report Garbage In, Garbage Out.