Google has been the most widely used online search engine for many, many years now. It’s common to say you want to “Google something” when in reality you just want to search for something on the internet, that’s how deep Google has embedded itself in our hearts and minds. One would expect that Google search has handled almost every question possible but the reality is that it does come across ambiguous questions from time to time that can throw a spanner in the works. To better answer those questions Google has created a new artificial intelligence called “RankBrain.”
Bloomberg reports that RankBrain has handled a very large fraction of search over the past few months. Apparently some 15 percent of searches made every single day haven’t been seen before by Google Search, so it can’t provide that level of information that the user might be expecting, and this is where the new artificial intelligence comes in handy.
RankBrain embeds large amounts of written language into mathematical values, known as vectors, using artificial intelligence. This allows the computer to understand what the user has written, if RankBrain picks up a word or phrase it hasn’t seen before it can guess what phrases or words might have a similar meaning and then filter results accordingly.
For example RankBrain can better handle questions like “What’s the title of the consumer at the highest level of a food chain?” The artificial intelligence works a bit differently from other technologies that power Google Search, a senior research scientist at the company explains “The other signals, they’re all based on discoveries and insights that people in information retrieval have had, but there’s no learning.”
He further explains that RankBrain is now one of the “hundreds” of signals that are taken into account by the algorithm which determines what to show in the results for a particular query, and how the results are ranked. Corrado says that it the few months since it has been deployed RankBrain has become the third-most important signal making sure that users get relevant results that they were expecting.
“I would describe this as having gone better than we would have expected,” he says.