Google Gadget Claims Obama Planning Coup

Google's voice activated home assistant gadget has told users that former president Barack Obama is planning a coup. It's an unfortunate quirk of the way the company applies its website techniques to the gadget.

Google Home - the company's answer to the Amazon Echo - is a small speaker that contains a microphone and a WiFi connection. It's designed to help with a host of activities around the house where voice-based information is quicker or more convenient than using a phone, tablet or computer.

Examples include being able to ask for a recipe unit conversion while in the middle of cooking, asking for a particular type of music to play, or asking for a specific YouTube video to be beamed to a compatible television sets.

Short Speech Selected By Algorithm

The problem came with the way Google Home searches for information in response to specific queries, usually by carrying out a similar search to the way happens when a user types something in on the Google search page.

Because it would be impractical to read out the full list of search results, Google Home instead reads out a "featured snippet", which is a section of text from a high-ranking result. It's similar to the "Knowledge Graph" that appears by the side of the page of search results - but, with a big difference.

The Knowledge Graph sections are hand-chosen by Google for specific common searches (such as a celebrity's age or the height of a mountain). The featured snippet is automatically created and chosen in response to individual queries using an algorithm intended to find the most relevant result.

Device Reads Out Fake News

That led to the unfortunate situation where several journalists noted that if you asked Google Home "is Obama planning a coup?" it spoke the response "According to details exposed in Western Centre for Journalism's exclusive video, not only could Obama be in bed with the communist Chinese, but Obama may in fact be planning a communist coup d'état at the end of his term in 2016." (Source: cnet.com)

This text actually comes from a website that, to put it mildly, is of questionable neutrality and accuracy. The report in question is actually a couple of years old as shown by the mention of 2016.

In this specific instance, it probably isn't a big problem as anyone asking such a specific question most likely has preconceptions about the claim. However, it does highlight the difficulties that come when the device attempts to answer a specific question with what's presented as a specific and definitive answer.

Google says it will remove any "featured snippets" which turn out to be from inappropriate or misleading content. The problem there is that unlike with the Knowledge Graph, there could be an infinite number of featured snippets created, making it very unpractical to weed out any dubious answers. (Source: engadget.com)

What's Your Opinion?

Do you think situations like this are a serious problem? Is there a way Google could improve the feature? Or is it simply unworkable for a gadget to try to answer any question by automatically picking an extract from a single website?