Amazon employs "thousands" to listen in on customer exchanges with AlexaMarie Donlon | April 12, 2019
According to a new report from Bloomberg, Amazon employs thousands of people to listen in on customer conversations with its virtual assistant Alexa.
In a process known as supervised learning, Amazon employees reportedly listen in on the recordings of user exchanges with Alexa, appropriately labeling data and feeding it back into the system to improve Alexa outcomes. The report said thousands of Amazon employees parse the Alexa recordings in an effort to train the system to recognize differences in regional slang and dialects, among other characteristics.
Reportedly, this practice is detailed in Amazon’s product and service terms. But privacy advocates are expressing concern that conversation contents could reveal the identities of those in the recordings or that the content could be stolen by third parties or misused by employees. Similarly, it is unclear how long the recorded exchanges are held by Amazon.
Addressing the concerns, Amazon explained in a statement:
“We only annotate an extremely small sample of Alexa voice recordings in order [sic] 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.”
The company also insists that it has “strict technical and operational safeguards, and have a zero tolerance policy for the abuse of our system.” Amazon claims that employees are unaware of the identity of the persons engaging with Alexa, and any information of that variety is “treated with high confidentiality,” and protected by “multi-factor authentication to restrict access, service encryption, and audits of our control environment.”
Although Amazon recently expressed a desire to move away from the supervised learning it employs to improve Alexa’s operation, it admits that it will still need to employ actual people to train the system for now.