Google admitted on Thursday that “language experts” hired by the firm listen to about 0.2% of the conversations that users have with your virtual assistant, which means that a part of these interactions are not completely private.
The assumption is usual, and often reiterate the companies that manage virtual assistants as Amazon, Samsung and Apple, in addition to Google, is that the conversations between a user and his assistant are completely private and that the interaction occurs exclusively through artificial intelligence, that is to say, that the only ones that “listen” to the user are robots.
However, the admission on Thursday by Google that 0.2% of these conversations are listened to by human beings to ensure, improve the quality of service sheds light on a practice that the companies usually avoid to advertise, though it is known within the industry that to a lesser or greater extent, is something usual.
The revelation came from the hand of the responsible product of searches of the californian company David Monsees, who posted an entry on the official Google blog as a response to an information appeared yesterday on the belgian television VRT NWS, which could access to about a thousand recordings of anonymous individuals.
The recordings were supplied to the belgian television in Dutch language by one of the “experts” that Google had hired in that country to listen to segments of the conversations and “understand as well the particularities and accents of each language in specific”.
The firm, which has already announced that it “will take action” by filtration considering it a “violation” of their data security policies, he admitted to having “experts around the world” whose role is to listen and transcribe “a small part of the dialogues to help us better understand those languages.”
In particular, the firm of Mountain View (California, USA) was the percentage of interactions analyzed by humans at 0.2% and ensured that these fragments are not associated to the accounts of the users and the experts will tell them that is not transcribed sounds or conversations in the background that are not directed at Google.
However, the belgian television was able to identify “postal addresses and other sensitive information“, in the recordings, which allowed them to get in contact with people whose voice had been recorded and confirm that indeed it was them.
“A couple of Waasmunster (Belgium) immediately recognized the voice of his son, and his grandson,” put as an example from VRT NWS.
Google indicated that the virtual assistant only sends the audio recordings once this has been detected that the user is interacting with him after that, for example, “Hey, Google” and that several tools are available to avoid the “false activations”, that is to say, that the software will interpret a sound erroneously as the key word to be activated.
Despite this, VRT NWS reported that around a thousand fragments of a voice to which he had access (all of them in Dutch language), 153 were conversations in which no one gave the order of activating the virtual assistant, but this misinterpreted a sound.
Among others, the average belgian said he had heard talks in the bed, between parents and their children, business calls, discussions, and scenes of both sex and violence.Translated from ABC News https://www.abc.es/tecnologia/informatica/hardware/abci-google-admite-escuchar-02-por-ciento-conversaciones-asistente-virtual-201907120231_noticia.html