In 2013 the world first heard about the concept of Rank Brain, which is an artificial intelligence (AI) program that Google was developing to make search engine technology smarter. It was officially put into use in 2015 and since then has been getting smarter with each and every search.
That’s the thing about AI, the more you use it, the more it learns.
Right now, Google uses hundreds of ranking singles, to determine what web pages come up when a person searches for a specific thing.
Some of these things are simple like does the search term appear in the HTML title tag? Does the search term appear in the body of the page? Does the page in question use header tags that feature the search term?
However, the three most important are PageRank, RankBrain and the content itself.
What exactly how does RankBrain work? Well from what I can gather after listening to Google talk about it, is that it is mainly used as a way to interpret the searches that people submit to find pages that might not have the exact words that were searched for.
When you search for “Apple” are you talking about Apple the technology company or are you referring to the fruit?
While Google previously had technology that could help with this problem, RankBrain does it better and it continues to get better at it each and every day, with each new search a person makes. That’s the beauty of technology that learns. The more people that use it, the more it learns to give people what they want.
A few years ago, people laughed when it was revealed that Google engineers were using romance novels to teach its artificial intelligence systems to better understand how people communicate. In all Google fed it 2,865 romance novels.
Yet when that same AI brain learned how to write its own romance novel, people were stunned. The world didn’t think it was possible that a computer could actually write fiction.
Of course, that wasn’t the goal. What they were trying to do is make a computer known for its stuff, factual style, to be more conversational.
The better we teach Google’s AI to communicate like we human’s do, the more it will know what we mean, truly mean and that means it will return better search results.
In other words, it will know instinctively if we mean by “apple”.
Something interesting to note was that RankBrain does all of its learning offline. It’s learning is done based on batches of historical searches and then learns to make predictions from those.
Once those predictions are tested and proven to return positive results, then the Google engineers push the latest version of RankBrain to the live servers and the learning process begins again.
Google processes over 40,000 searches every second of every day. That means that every day, RankBrain has an opportunity to learn over 3.5 billion things from us and be all that much smarter the next time we search.