With the competition in computing and smartphone having reached its peak, the next wave of fight is happening in the digital assistants space. Amazon is the de facto leader here with its Alexa digital assistant and its reach only continues to expand with the launch of new Echo smart speakers. However, these digital assistants powered smart speakers come with an inherent threat to privacy. A number of instances have proved how Alexa and other digital assistants listen to their owners and have at times, shared the audio with a totally different user. Now, Amazon has confirmed that it employs thousands of people around the world to improve Alexa.
In other words, these teams listen to voice recordings captured in the homes and offices of Amazon Echo owners. Bloomberg reports that these recordings are transcribed, annotated and then fed back to the machine learning software that power natural language processing of Alexa. The Seattle-based company says that the process is part of an effort to eliminate gaps in Alexa’s understanding of human speech and help it respond better to voice commands.
The report details human involvement in the evolution of Alexa, which Amazon says “lives in the cloud and is always getting smarter.” The team reportedly comprises of a mix of contractors and full-time Amazon employees working from Boston to Costa Rica, India, and Romania. Each reviewer is said to parse as many as 1,000 audio clips per shift of nine hours a day. These reviewers also occasionally pick up things that Echo owners would rather like to stay private and they use internal chat rooms to share files.
While the work is described as mundane, these reviewers do pick on things at times that can be upsetting or criminal in nature. “Two of the workers said they picked up what they believe was a sexual assault. When something like that happens, they may share the experience in the internal chat room as a way of relieving stress,” Bloomberg reports.
“We take the security and privacy of our customers’ personal information seriously,” an Amazon spokesman told Bloomberg in an emailed statement. “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. All information is treated with high confidentiality and we use multi-factor authentication to restrict access, service encryption and audits of our control environment to protect it.”
Watch: Amazon Echo Spot First Look
Amazon does not explicitly confirm that humans listen to recordings of some of their conversation with Alexa. However, it does confirm that user consent is taken to train Alexa and its speech recognition and natural language understanding systems. Amazon Echo owners have the option to disable the use of their voice recordings for training purposes. In a nutshell, think of the whole exercise as one similar to call centers where the calls are monitored for training employees.
Alexa and other digital assistants powering smart speakers record audio by listening for a wake word. When the wake word is detected, these digital assistants spring to life and start recording and beaming the command to Amazon’s cloud servers for appropriate answers. Employees at Apple and Google also work as human helpers training their digital assistants. Amazon also says that no audio is stored unless Echo detects the wake word is triggered by pressing a button.