When you talk to an Amazon Echo device, most of the time you only hear Amazon's speech recognition software. But sometimes, Bloomberg reports, a copy of the audio is sent to a human examiner in one of the many Amazon offices around the world. The human being listens to the audio clip, transcribing it, and adding annotations to help Amazon's algorithms get better.
"We take the security and protection of our customers' personal information seriously," said an Amazon spokesman in a statement emailed to Bloomberg. "We're just commenting on a very small sample of Alexa voice recordings to improve the customer experience."
Bloomberg points to a significant workforce doing this kind of work Boston, Costa Rica, India and Romania: Employees interpret up to 1
It's not hard to see why Amazon wants to have some audioclips checked by people Instead, a person could listen to clips and see if the software interprets them correctly.
But Amazon could have described the role of human reviewers more explicitly. "We use your requests to Alexa to train our speech recognition and our natural language skills. Speech understanding systems," says Amazon on his frequently asked page with questions about Alexa – without mentioning the human element.
Amazon told Bloomberg that there are strict privacy safeguards to prevent system abuse. "Employees do not have direct access to information that can identify the person or account as part of this workflow," the spokesman said. "All information is treated with high confidentiality, and we use multi-factor authentication to restrict access, service encryption, and monitoring of our control environment to protect it."
Employees occasionally hear unexpected or extreme audio, reports Bloomberg. In one case, two workers heard what sounded like a possible sexual assault, but it was said that it was not appropriate to intervene. Employees can talk to other employees in an internal chat service about what they are hearing, and they can swap clips that have difficulty interpreting. The report also mentions that files are shared just because they are "amusing."
Bloomberg also says Apple and Google have human critics
Bloomberg says that Apple's Siri "also has human helper". The company refers to an Apple Privacy White Paper that describes how Apple uses the audio data recorded by consumer devices.
"User voice recordings are stored for a period of six months so that the recognition system can use them to better understand the user's voice." The White Paper says. "After half a year, another copy will be saved for Apple without an identifier to improve and develop Siri for up to two years."
What about the Google Virtual Assistant? "Some reviewers can use their assistant to access some audio snippets to train and improve the product," reports Bloomberg. "But there is no personally identifiable information, and the sound is distorted," says the company. "