Tens of millions of people use smart speakers every day to play music or search for all kinds of information. However, there are still those who resist inviting these devices and their powerful microphones to their homes because of concerns that someone may be listening. Something that may seem like something out of a science fiction movie but that really happens. A clear example is that of Amazon, which employs thousands of people around the world to help improve Alexa, the personal assistant of the purchase platform 'on-line', through its Echo speakers. The team listens to the voice recordings captured in the houses and offices of the owners of these types of speakers and then transcribes the conversations and re-enters this information in the 'software'. All this, in an effort to eliminate the understanding gaps between Alexa and the human being.
According to Amazon, Alexa "lives in the cloud and is getting smarter". But, like many software tools created to improve from experience, Humans have an important role in the learning process. In this sense, according to Bloomberg, the human team behind Alexa is made up of people who work in Boston, Costa Rica, India and Romania, according to some people who have worked on the project. According to the Bucharest office, each reviewer analyzes up to 1,000 audio clips per turn. The work, however, is worldly. A worker in Boston said he extracted accumulated voice data for specific expressions such as "Taylor Swift" and wrote them down to indicate that the search engine referred to the musical artist.
The team listens to the voice recordings captured by Alexa to later transcribe the conversations and re-enter this information in the 'software'
But also uncomfortable audios are collected and could be used to incriminate criminal acts. Two of the workers said they heard what they believe was a sexual assault. When something like this happens, Amazon says it has specific procedures, but the workers explained that, after requesting guidance for such cases, they were told that it was not the company's job to interfere. "We take the security and privacy of our clients' personal information seriously," an Amazon spokesperson said in a statement. "We just recorded an extremely small sample of Alexa voice recordings to improve the customer experience. For example, this information helps us to train our speech and natural language recognition systems, so that Alexa can better understand their requests and guarantee the service, "the text adds.
If the lack of privacy is a subject that can worry people who use Alexa, Amazon is clear: "Employees do not have direct access to the information that can identify the person. All information is treated with high confidentiality. " In addition, in the Alexa privacy settings, the company offers users the option to deactivate the use of their voice recordings for the development of new functions. But, sometimes it may seem that we are not aware that what we say is being recorded. "We do not think that another human being can be listening to what we say to our intelligent speaker in the privacy of our home," said Florian Schaub, a professor at the University of Michigan who has researched privacy issues related to this type of device. "I think we have been conditioned to the [suposición] that these machines are only doing a mechanical learning, but the fact is that there is still manual processing involved, "adds the professor.
"We just scored an extremely small sample of Alexa voice recordings to improve the customer experience," Amazon explains.
But Alexa is not the only device that has the help of humans to improve their learning. Apple's Siri also has human helpers, They work to assess whether the interpretation of the digital assistant's requests matches what the user says. In the case of the company founded by Steve Jobs, the recordings also lack personal identification information and are stored for six months linked to a random identifier, according to an Apple security document. At Google, some reviewers can access some audio fragments from their assistant to help train and improve the product, but they are also not associated with any personally identifiable information and the audio is distorted, according to the company.
When Echo debuted in 2014, Amazon's cylindrical smart speaker quickly popularized the use of home voice software. In a short time, Google launched its own version, called Google Home, followed by Apple's HomePod. According to researcher Canalys, Last year, 78 million smart speakers were sold. Millions more use voice software to interact with digital assistants on their smartphones.
When Echo debuted in 2014, Amazon's cylindrical smart speaker quickly popularized the use of home voice software