Google has announced that its new artificial intelligence (AI) system, RankBrain, has been hard at work for months, using machine learning to help the search engine’s powerful algorithms interpret about 15 percent of all queries.
It focuses specifically on words and phrases that are ambiguous and have never been seen before.
In one test run, the engineers were asked to review a set of pages, and guess which would rank on top.
They were right about 70 percent of the time, not a bad score until the pages were scanned by RankBrain, which was right 80 percent of the time.
What is RankBrain?
Google’s breakthrough is using RankBrain to interpret new, unfamiliar multi-word, or “long-tail,” queries. The result delivers a more accurate, useful result for the searcher each time. The benefit of this improvement grows exponentially as the algorithm incorporates and uses learned information again for future complex searches. By the way, that 15 percent figure for the number of never-seen-before searches is 15 percent of 3 billion total Google queries daily, amounting to no fewer than 450 million per day! The remaining searches rely on information learned and translated by humans.
Machine learning and artificial intelligence are often used interchangeably to describe computers that make connections and learn, independent of human instruction. According to a Google senior research scientist quoted for the recent Bloomberg Business story that broke the news, RankBrain uses machine learning to translate unfamiliar language into mathematical phrases. These phrases, called vectors, are then fed into the computer. RankBrain’s ability to learn, guessing at unfamiliar words or phrases by comparing them with previously learned information, is the new function that should make Google’s search results more powerful than ever.
SEO: the Known and the Unknown
He goes on to explain that the overall search algorithm, called Hummingbird, contains hundreds of signals that help it determine page results, and RankBrain is now the third most important of those signals. Its contribution is the ability to recognize patterns at a faster rate than humans, so it can connect a new search term with a pre-existing record that is similar in meaning. Other important signals among those hundreds include, for example, the words on a page, the words in bold, and the links on a page.
At this stage, there’s plenty of room to wonder about the implications of this new system, which may be minimal to start, but should grow over time. Change is a constant where Google is concerned, so we’ll keep our ears to the ground for further insights into RankBrain. The more we understand about how and why Google’s algorithms make ranking decisions, the more effectively we can target our best SEO practices, like producing high-quality, Google-friendly content to enhance our clients’ user experience.
What’s the new marketing mindset?
It’s all about growth.
Learn the latest in our new eBook.