The algorithms that power Google queries are incredibly complex and fluid, changing as the company sees fit in order to improve the speed and accuracy of returned results.
The Mountain View-based company relies on hundreds of “signals” to help determine what results to surface – and in what order. Remaining at the forefront of search is crucial to Google’s core and in order to do so, tweaking and developing new technologies to power search is mandatory.
Greg Corrado, one of Google’s senior research scientists, recently revealed to Bloomberg that for the past few months, a very large fraction of Google searches have been handled by an artificial intelligence system called RankBrain.
Corrado noted that 15 percent of the searches that Google fields each day are ones that it has never seen before. An example of an ambiguous query would be something like, “What’s the title of the consumer at the highest level of the food chain?”
Rather than try to rely on traditional signals to interpret such queries, RankBrain takes the written language and converts it into mathematical entities known as vectors. From there, it can make a guess at other words or phrases that might have a similar meaning and thus, return more accurate results.
Google has been working with artificial intelligence for half a decade now; RankBrain itself has taken five engineers a full year to develop. The effort has certainly been worth it, however, as RankBrain has become the third-most important signal when returning a query. That was surprising, Corrado said, noting that it has gone better than they expected.