Behind Amazon’s digital assistant Alexa are thousands of humans reviewing some of the millions of utterances people make to their smart speakers, according to a new report from Bloomberg.
These teams, Amazon said, primarily exist to make Alexa smarter. The company confirmed that the teams annotate a small percentage of clips, and the reason for doing so is to strengthen Alexa’s ability to recognize speech and understand commands.
As Bloomberg notes, there is already a segment of the population that avoid smart speakers because of privacy concerns. News of teams on the other end of Echo devices listening to what some users are saying to Alexa could inflame privacy concerns even more.
BBC notes that the practice of employees listening to requests of digital assistants is fairly common among big tech companies. Google and Apple both have humans reviewing conversations with the Google Assistant and Siri respectively, to improve the digital assistants and make sure they are interpreting requests correctly.
Amazon reiterated that privacy and security are a top priority and that “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.”
Here is the full statement from Amazon:
We take the security and privacy of our customers’ personal information seriously. We only annotate an extremely small number of interactions from a random set of customers 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. While 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, customers can delete their voice recordings associated with their account at any time.
Amazon emphasized that Echo devices are designed to only activate when someone utters the wake word. When the device is activated, the blue ring on top flashes, and that is when the speaker is sending the conversation to the cloud. The devices are always hearing, looking out for the wake word, but they’re not always listening and recording.
Bloomberg reports that the teams are a mix of contractors and full-time employees who work in offices in Boston, Costa Rica, India and Romania. Workers review as many as 1,000 clips a day, most of which are mundane. However, occasionally the clips pick up disturbing or criminal behavior.
The digital assistant has been at the center of criminal cases in the past, most famously a 2015 murder case where after some resistance, the tech giant handed over audio recordings and other records for an Echo belonging to the suspect
Alexa, and competitors like the Google Assistant and Siri, are still young and still learning. Humans have an important role to play in making these digital assistants smarter by understanding more languages, keeping up with slang and answering questions.
Another example of Amazon’s use of human knowledge to improve Alexa is a program unveiled last year called Alexa Answers. A select group of customers invited to the program can go to a website to see a list of questions Alexa can’t answer today and provide information to fill in the blanks.