Faced with such a lack of discernment of its algorithms, would we touch the finger of the limits of the AI?
While some are actively fighting fake news that bloom on the web and social networks, Google Home seems to have a hard time making a distinction.
“Former US presidents of the Klu Klux Klan”, or “Nazis within the Republican party” …: the “inbox” and rumors are going well and are in fact relayed By the voice assistant.
It all starts when the Search Engine Land site mentions an “algorithm concern” on Search. Indeed, according to the media, the feature “Featured Snippets in Search” is problematic in that it “takes information from the Web to provide the user with a definitive and out of context response to a query.”
It does not integrate the analysis or cross-tabulation of the thousands of results proposed. However, Google Home offers only one answer per question: the answer chosen by the home automation assistant is currently the most popular, but not necessarily the most reliable. This is where the phenomenon “One true answer” appears.
Given this, Rory Cellan-Jones, a BBC journalist, asked his home automation assistant last weekend if Obama was preparing a coup.
The answer was unexpected: Google Home said Barack Obama “may well be preparing a coup for the end of his term.” From there, other journalists tried to deceive Google Home, which stated, among other things, that five former US presidents had been members of the KKK.
In response to this little controversy, Google told AFP “work quickly” to remove these misleading allegations. The Mountain View firm would be refining the research to avoid “inappropriate and misleading” results.