April, 2019

AI Shoplifting Prediction Program

A recent Bloomberg article described an artificial intelligence software – developed by a Japanese company, Vaak, and currently testing in stores in the Tokyo area – that attempts to predict and avert potential shoplifting. The software analyzes security footage looking for signs of nervousness or other body language that may be considered suspicious. When suspicious behavior is found, the software alerts staff through a smartphone app. Shoplifting impacts the retail industry at a cost of about $34 billion annually; companies are predicted to invest about $200 billion in security technology this year. Vaak hopes to have its product in about 100,000 stores in Japan in the next three years.

The company hopes to expand the application of the technology to monitor how customers interact with products. The company says that the technology can be also utilized in public areas, such as train stations, to monitor for suspicious behavior and potential suicides.

Sources:  Lisa Du and Ayaka Maki, “The AI Cameras That Can Spot Shoplifters Even Before They Steal,” bloombergquint.com, March 4, 2019:
https://www.bloombergquint.com/technology/the-ai-cameras-that-can-spot-shoplifters-even-before-they-steal#gs.x56Yu8NC

Amazon Workers
Listening in on Alexa

Recent reports indicate that Amazon employs thousands of people globally to listen to voice recordings made through use of the Alexa digital assistant, or smart speaker device. The workers make transcriptions, annotate the transcriptions to provide better context, and the data is then input into the software to facilitate the Alexa program’s algorithm development. The workers, in 9-hour shifts per day, can review up to 1,000 audio clips per shift.

If the workers hear something potentially criminal – in one example cited, a suspected sexual assault – they share the information in internal work-chatrooms. An Amazon spokesperson was quoted as saying, “We take the security and privacy of our customers’ personal information seriously … We only annotate an extremely small sample of Alexa voice recordings in order [to] improve the customer experience. For example, this information helps us train our speech recognition and natural language understanding systems, so Alexa can better understand your requests, and ensure the service works well for everyone … We have strict technical and operational safeguards, and have a zero tolerance policy for the abuse of our system. Employees do not have direct access to information that can identify the person or account as part of this workflow.” Florian Schaub, a professor at the University of Michigan, was quoted as saying, “You don’t necessarily think of another human listening to what you’re telling your smart speaker in the intimacy of your home.”

Smart speakers are a growing market. Amazon’s Echo was launched in 2014, and other companies, including Apple Inc. (the HomePod) and Alphabet Inc. (the Google Home), have marketed their own devices. Globally, consumers purchased about 78 million smart speaker devices in 2018.

Human assistants are also used at Apple and Google. Apple reports that the recordings reviewed by the workers contain no personally identifiable information, and the data is kept for six months. Similarly, the Google workers do not get personally identifiable information.

Sources:  Matthew Day, Giles Turner, and Natalia Drozdiak, “Amazon Workers Are Listening to What You Tell Alexa,” Bloomberg.com, April 10, 2019:
https://www.bloombergquint.com/business/is-anyone-listening-to-you-on-alexa-a-global-team-reviews-audio

Amazon’s Ring Doorbell May Use
Facial Recognition Technology to
Identify Suspicious People

Amazon has applied for patents for systems to use with its Ring doorbell systems; the program would – when a suspicious person was identified – automatically alert law enforcement. Google’s doorbell system, Nest Hello, also uses facial recognition technology, but does not automatically alert law enforcement.

According to Ring, the doorbell system would work as follows:
A video may be analyzed by an A/V recording and communication device that recorded the video (and/or by one or more backend servers) to determine whether the video contains a known criminal (e.g., convicted felon, sex offender, person on a “most wanted” list, etc.) or a suspicious person. Some of the present embodiments may automatically submit such a video stream to the law enforcement agencies.

Jacob Snow, of the ACLU, was quoted as saying, “Amazon is dreaming of a dangerous future with its technology at the center of a massive decentralized surveillance network, running real-time facial recognition on members of the public using cameras installed in people’s doorbells.”

Sources:  Ben Fox Rubin, “Amazon’s Ring takes heat for considering facial recognition for its video doorbells,” cnet.com, December 14, 2018:
https://www.cnet.com/news/amazons-ring-takes-heat-for-considering-facial-recognition-for-its-video-doorbells/

by Neil Leithauser
Associate Editor