Geekier search marketers often make jokes about Google becoming a real-life Skynet (the artificial intelligence that becomes self-aware and destroys humanity in The Terminator movie franchise), but they’re closer to the truth than they might think. According to recent reports, Google has deployed an artificial intelligence nicknamed RankBrain to handle approximately 15% of the volume of search queries Google receives every day.
For science and tech geeks, this is nothing short of incredible. In this post, we’ll cover everything you need to know about RankBrain and what it means for not just the future of Google, but the future of information itself.
RankBrain is the nickname given to the proprietary artificial intelligence (or AI) used by Google to handle search queries. The AI may never be given a formal name, or maybe RankBrain will stick, but for now, it’s the name being used for the artificial intelligence that’s handling 15% of Google’s search volume.
An important distinction to be aware of is that RankBrain is not (currently) a standalone technology at Google – it’s incorporated directly into the existing algorithm that Google uses to rank web pages. However, while there are literally hundreds of ranking signals (and nobody outside Google knows for sure how many there are), RankBrain has quickly become one of the most important. Greg Corrado, a senior research scientist at Google, told the Washington Post that RankBrain is now the third most-important factor in determining how a search query will be answered on the SERPs.
Image via Brafton
Aside from its growing importance at Google, RankBrain is already exceeding expectations. In one test, Google engineers were asked to look at some web pages and estimate how many Google’s algorithm would rank highly. The engineers were right roughly 70% of the time. Under the same test conditions, RankBrain was correct 80% of the time.
Artificial intelligences are computer systems that mimic human cognitive functioning – the way our brains work and the ways we think – to solve problems. Artificial intelligences can perform many tasks that, until recently, required human intervention to complete, such as image and speech recognition, translation, and basic decision making.
Today’s computers are incredibly powerful, but most are also incredibly stupid – computers have to be told what to do, what things are, and virtually everything else before they can accomplish the tasks they’re built for. This is not necessarily true of AI systems. Think of an AI as a “learning machine.” The more data an AI handles, the more it learns and the more effectively it can approach new problems in the future.
A great way to think about artificial intelligence in the context of search is by looking at the semantic search functionality of Google’s Hummingbird update.
Using Google Now, users can take advantage of semantic search by relying on Google’s technology intuitively understanding the relationships between two search terms without being explicitly told that the two terms are related. For example, you could perform a search for “Dame Helen Mirren” to find out more information about the actress, before asking “Where was she born?” In this example, Google would understand “she” in the second query to mean Dame Helen Mirren, based on the semantics (or meaning) and context of your previous search.
Using artificial intelligence, Google can not only handle the increasing volume of search traffic that comes as a byproduct of increasing Internet access, but also offer a more intuitive, responsive experience that allows users to find the information they need faster and in a more fluid way.
By all accounts, RankBrain performs strongly at determining the intent of ambiguous searches – some of the most challenging queries for Google to interpret – meaning that search results are only going to become more accurate as the technology develops and handles greater volumes of queries.
There are limitations to this technology, as demonstrated by the image above, but these challenges may soon be overcome as AI becomes more powerful and “learns” to interpret and solve problems once thought impossible.
The easy answer is, “We don’t know.” The more likely answer, however, is “It already has.”
Google developed RankBrain with the intention of providing users with more refined search results based on their queries. It follows that, as search results become more accurate, then ads served based on those queries become more targeted and, as such, more effective.
It’s impossible to say specifically how RankBrain is being used to handle commercial search queries, but considering that approximately 15% of the 100 billion searches Google handles every day, it’s practically inevitable that RankBrain is handling queries involved in the AdWords auction.
Again, this is pure speculation, but it seems likely that as RankBrain becomes increasingly sophisticated, it will be applied to other parts of Google’s (sorry, Alphabet’s) business.
Google has long had an eye toward the future, having diversified into areas of research including longevity (through Calico Labs, an Alphabet company). The hiring of renowned futurist Ray Kurzweil in 2012 – a man who wants to literally live forever – was perceived as a strong indication of where Google sees itself in the near future, as was the company’s acquisition of machine learning startup DeepMind for $500 million last year. Machine learning is the process of enabling computers to make “decisions” without being explicitly programmed to do so. Google’s driverless car project is a prime example of a practical application of machine learning, as the algorithms in the software that operates these vehicles must learn how to continually adapt to changing road conditions and variables such as drivers’ choices, in real time, without being expressly programmed to do so by a human being.
Images created by a neural network, a “learning” machine, via the MIT Computer Science and AI Laboratory
In addition to its internal research and development, Google is competing with other tech giants at work on their own AI systems. Facebook is already using AI to filter results in users’ News Feeds and “learn” their preferences for the type of content that should take precedence,
Jokes aside, we don’t really know the dangers of AI right now. Sure, movies like The Terminator showed us that artificial intelligences are a very bad idea, and although many scientists remain confident we can control AI to a sufficiently safe degree, some of the world’s leading technologists and scientists – including physicist Stephen Hawking, Tesla and Space X’s Elon Musk, and Microsoft founder Bill Gates – have all warned of the grave dangers that unchecked development of artificial intelligence could pose to the future of mankind. Only time will tell.
We may end up waging a bitter war against the machines amid the ashes of the nuclear fire, but for now, at least our search results will be more accurate.
UPDATE: Well, if this were a Terminator movie, we’d be at the part just before Skynet becomes self-aware and destroys us all. According to Google, RankBrain is now processing every search query Google receives, and is now the third most-important ranking signal in the Google algorithm. Interestingly, although every search query is being processed by RankBrain, that doesn’t necessarily mean that it’s influencing every query. According to Jeff Dean, a computer scientist at Google, RankBrain exerts influence over “a lot” of queries, but not all of them.
RankBrain has become a core part of Google’s approach to search. However, Google’s interest in machine learning goes far deeper than that. Google has invested millions of dollars in its burgeoning machine learning division, but more importantly, Dean says Google is becoming a “machine-learning first” company.” This means that artificial intelligence and machine learning aren’t just a fad for Google, or a way to make their processes smarter – it’s becoming the foundation upon which the company’s technology is built.
The increasing reliance upon RankBrain at Google has the potential to radically upset the world of SEO. It’s too soon to say with any certainty how Google’s heightened focus on machine learning will have on SEO, but it’s safe to say it’s probably going to be a really big deal.
Originally from the U.K., Dan Shewan is a journalist and web content specialist who now lives and writes in New England. Dan’s work has appeared in a wide range of publications in print and online, including The Guardian, The Daily Beast, Pacific Standard magazine, The Independent, McSweeney’s Internet Tendency, and many other outlets.
See other posts by Dan Shewan
Please read our Comment Policy before commenting.