Amazon recently announced that it is restricting police use of its facial recognition tech, Rekognition. This comes at a time when incidents of excessive use of force by the police have been highlighted by mainstream and social media. As well as an increase in the level of support for Black Lives Matter, in response to the death of George Floyd.
Recently, Amazon has been an active participant in finding ways to limit the spread of COVID-19. For example, the company has proposed new technology to track the spread of coronavirus, while also delaying Prime Day until later in the year. Amazon has also found itself in hot water for attempting to silence employees sounding off on the company's alleged mistreatment of workers during the pandemic. However, now Amazon has turned its attention to policing conduct and how its technology might be used by police.
Rekognition is a machine learning technology similar to artificial intelligence with its main two components focusing on facial analysis and facial recognition. Facial analysis can only find generic features such as facial hair or gender, and not features that make a face unique, like the shape of a nose or the size of lips. Facial recognition uses those unique facial features to identify possible matches. Both aspects of computer vision can help fight crime and the police can use the technology to quickly identify suspects. However, there are concerns that facial recognition technology can be used to discriminate and violate civil rights. In light of the current situation in the US, Amazon has now decided to place the police's use of Rekognition on hold for one year.
Amazon's Police Ban Could Mean Stronger Regulations
The use of facial recognition technology has been used for unlocking phones and tagging friends on social media. Yet, its use in law enforcement is relatively new and therefore, it's constantly subject to change or improvements. With recent instances of excessive police force and mounting evidence of racial profiling with facial recognition technology, Amazon seems to believe it must implement more guidelines for the use of the technology by law enforcement, in the hope of limiting the chance of the technology being abused or subjugating civilians to unfair discrimination. As with artificial intelligence in general, there are concerns of racial bias, but unlike AI, facial recognition technology's bias not only stems from the algorithms or the people creating them, but also from people actively using the technology in the real world.
Facial recognition and facial analysis are not exact sciences, but sciences of approximation. Each provide percentages of confidence or confidence thresholds, and require integrity among users to understand the importance of the percentage, and formulate their own conclusions. For example, Amazon's guidelines for Rekognition state users should only use confidence thresholds of 99% or higher to avoid errors and false positives. Furthermore, using facial recognition and facial analysis together, or to gauge the confidence of the facial recognition with facial analysis would be directly against Amazon's guidelines, considering both use different algorithms and may produce inaccurate results. The subjective element - the fact law enforcement officers must use their own intuition to deduce their own conclusions instead of relying so heavily on the technology and the data - is where the added concerns come in. Although Amazon wants change to arise, the company has yet to disclose specifics on the type of regulations it now expects.
Source: Amazon.