A Deep Dive into Google’s RankBrain Launch
RankBrain, Google’s newly launched artificial intelligence component, which works in conjunction with the search engine’s Hummingbird algorithm, was announced late last month – setting SEO experts spiraling into a frenzy. While to most, Google’s RankBrain announcement felt like a huge surprise, we’ve taken the time to back track, and in fact have found two major events foreshadow the announcement of this artificial intelligence system now live on Google Search.
First, Eric Schmidt, the Executive Chairman of Alphabet Inc., is quoted on March 16.th
“I think the biggest trend is going to be the use of machine intelligence of large data sets to solve every problem,” Schmidt said. “I cannot think of a field of study, a field of research — whether it is English, soft sciences, hard sciences or any corporation — that can’t become far more efficient, far more powerful, far more clever.” (McFarland, 2015)
This quote appears prophetic, creating anticipation of vast changes to revolutionize Alphabet’s many companies. Eric Schmidt, the former president of Google, now Alphabet, has been spending many months trying to assure both national and international business interests, including public and private institutions, that the potential impacts of artificial intelligence, also called machine learning, on the world remain largely misunderstood. Then on September 28th of this year, Wired revealed another landmark Google announcement – that its vast quantum computer farms are undergoing a massive upgrade.
“GOOGLE IS UPGRADING its quantum computer. Known as the D-Wave, Google’s machine is making the leap from 512 qubits—the fundamental building block of a quantum computer—to more than a 1000 qubits. And according to the company that built the system, this leap does not require a significant increase in power, something that could augur well for the progress of quantum machines.” (Metz, 2015)
This change signifies the upgrade of integration between hardware and software breathing into Google’s live artificial intelligence, aka machine learning. This advancement means that Google Search’s speed has increased many times over, with little need to improve the power infrastructure.
Wide-ranging coverage on RankBrain in numerous articles, blog posts and tweets is now live all across the web. Moreover, like all multifaceted updates in Google Search, this change comes with numerous intricate details. I quickly realized that throughout this press storm, some critical facts have been misquoted or lost in translation.
RankBrain’s Launch Full of Press Surprises
Sundar Pichai, Google’s Chief Executive Officer made a startling announcement on October 22th at Alphabet’s earnings meeting, saying, “machine learning is a core transformative way by which we are rethinking everything we are doing,” (Clark, Google Turning Its Lucrative Web Search Over to AI Machines, 2015). Google has never been so bold as to announce that artificial intelligence, aka machine learning would take such an immediate and active role in search, let alone other parts of the company. Moreover, the surprises kept coming when a few days later, Bloomberg Business interviewed Greg Corrado, a senior research scientist who announced that a major update to Google.com went live a few months ago. The conversation revealed that a new component in their search algorithm now assists in delivering search results for a large portion of national and international search results. He christened this independent software “RankBrain.”
“For the past few months, a “very large fraction” of the millions of queries a second that people type into the company’s search engine have been interpreted by an artificial intelligence system, nicknamed RankBrain, said Greg Corrado, a senior research scientist with the company, outlining for the first time the emerging role of AI in search.” (Clark, Google Turning Its Lucrative Web Search Over to AI Machines, 2015)
This announcement is consistent with Google’s habit of announcing significant changes to its algorithm after the fact. However, with slight prodding Google uncharacteristically gave an interview offering answers to relatively direct questions. Greg went on to say, “…in the few months it has been deployed, RankBrain has become the third-most important signal contributing to the result of a search query.” (Clark, Google Turning Its Lucrative Web Search Over to AI Machines, 2015)
The final, and I believe most stunning revelation that came out of the Bloomberg Business article was a quote detailing how accurate RankBrain has become at matching search results with queries.
“Google search engineers, who spend their days crafting the algorithms that underpin the search software, were asked to eyeball some pages and guess which they thought Google’s search engine technology would rank on top. While the humans guessed correctly 70 percent of the time, RankBrain had an 80 percent success rate.” (Clark, Google Turning Its Lucrative Web Search Over to AI Machines, 2015)
Information experts and marketers have long known of Google’s, now Alphabet’s mission to develop and launch artificial intelligence on Google.com. However, this interview reveals how far Google has gotten ahead of the competition. We now know that their AI is alive, and it is learning within the machine at Google.com. I recommend following all the sources and that we each read every article carefully from this point on, so that we can each form our own insights and questions.
That being said, here are a few more insights I’ve gathered from relevant sources thus far.
The Bloomberg Business Video Interview & Jack Clark Article Insights
Bloomberg Business published an article, which I feel is the trumpet announcing that Google’s AI is conscious and that Google’s yearlong effort to launch the update has been successful, or in the words of the interviewed Google engineer, “I was surprised,” Corrado said. “I would describe this as having gone better than we would have expected,” (Clark, Google Turning Its Lucrative Web Search Over to AI Machines, 2015). Here are a few key takeaways:
RankBrain is an artificial intelligence system that is live on Google.com for several months
RankBrain process a “very large fraction” of the millions of queries a second
RankBrain is the third-most important signal contributing to the result of a search query
RankBrain is more accurate in guessing new search queries than software engineer
Bill Slawski’s RankBrain Coverage
Bill Slawski’s RankBrain coverage stands apart based on his thorough and extensive experience exploring technology patents. In his latest article, Investigating Google RankBrain and Query Term Substitutions, Slawski examines Kedar Dhamdhere’s patent titled Using Concepts as Contexts for Query Term Substitutions. It is relevant to note that one of the inventors of this patent, Thomas Strohmann has written multiple papers on Machine Learning – Sparse Greedy Minimax Probability Machine Classification and A Formulation for Minimax Probability Machine Regression to name a few.
The patent reveals a treasure trove of insights into how RankBrain is an independent, autonomous entity that stands outside the traditional Google algorithm factors.
Bill summarizes the patent in the following processes:
Acquisition of “query term replacement” information for one or more terms in a query search.
Acquisition of “query term replacement” information for one or more query terms that precede or succeed “subsequent queries” that include the “”
The “idea” is adjacent to the one or more query terms in the “subsequent ”
Acquisition of “query term replacement” information of one or more query terms in a “context” of the “”
Assign a substitution process in a “context” of the “idea” based on the acquired “query term replacement” information.
Bill’s article brings the most clarity to the RankBrain coverage – though it still leaves many questions unanswered due to the shallow content provided in the initial Bloomberg interview and article.
The concepts discussed in Bill’s article sparked a lively conversation in our office. After concluding our banter, the fact dawned on me that I might have made an error in my understanding of RankBrain. I went back and re-read all the articles and was shocked that I had misquoted a central concept. In the original interview with Greg Corrado and Bill’s treatment of the Google patent discussion, I understood that RankBrain processes “new queries” that Google has never seen before representing 15% of “completely new” Google Search queries. That seems like a huge figure and appears to support the earlier Bloomberg quote, “(a) very large fraction” of the millions of queries a second that people type into the company’s search engine.” (Clark, Google Turning Its Lucrative Web Search Over to AI Machines, 2015) However, I believe that Greg Corrado, the senior research scientist at Google was pointing to a much larger figure that I initially inferred wrongly, and thus used the quote out of context. Greg never says that RankBrain currently processes 15% exclusively. Moreover, I suspect the percentage is much larger. How large is impossible to know at this point in time, but what I took away from the discussion was that RankBrain is a lot bigger update that I first realized.
Further insights from Bill Slawski’s article:
The search results will reflect much greater diversity around core ideas and websites that include “replacement phrases” around the core ideas on the page.
In the comments of the article, Bill addresses the issue of 15% being misquoted. Here is Jack Clark’s tweet: “@glenngabe It is invoked on a very large fraction of queries and is particularly good at dealing with the 15% per day that haven’t been seen” (Clark, Jack Clark on Twitter, 2015)
What We Can Expect of Future RankBrain Conversations
The more people understand RankBrain, the more faceted the coverage and conversations will become. Review the next section for further coverage available on the rising influence of artificial intelligence systems. I highly recommend reading the span of quality reporting on the web when you have a moment to take a deeper dive.
There are two additional articles published in The SEM Post that ask and answer a lot of important questions. The first article, RankBrain: Everything We Know About Google’s AI Algorithm, is a suitable primer on RankBrain with post-article/interview conversations from Bloomberg journalist Jack Clark and expanded coverage of the original interview with Greg Corrado. The second article, Google’s RankBrain: 9 Industry Experts Weigh In, includes conversations between nine thoughtful search engine experts and Google engineers. These two articles close the gap in understanding about RankBrain including addressing some of my discoveries.
Until we talk again, here are some major mysteries and rabbit holes to explore that the coverage leaves unanswered.
What is the actual percentage of search queries RankBrain processes each second?
Does RankBrain replace the requirement of “core ideas/concepts” to be present within the keyword density of the page?
Does RankBrain replace the “authority” of inbound link popularity? We know there is a dampening of exact match anchor text, but will “anchor text” remain relevant?
Google’s patent was granted in August 2015, but it was filed on Oct 12, 2012, what other patents apply to RankBrain?
How long before RankBrain becomes the 2nd most important factor within the Google Algorithm relevancy?
When RankBrain becomes the 1st measurement of matching all query intent and context, won’t Google’s Algorithm become a black box?
Did Matt Cutts “extended leave” predict the introduction of RankBrain?
More Key Takeaways
There are several key takeaways that the process of researching RankBrain revealed. Moreover, the most palatable and I hope not too surprising is that AI still has a long way to go before the Robot Uprising begins. However, I am starting to get excited that AI might be taking its first breaths. Here are my takeaways from this article.
Eric Schmidt announces that Machines will take over Alphabet, Inc. very soon.
Google Quantum D-Wave Computers are upgraded from 512 qubits to 1000 qubits
Sundar Pichai: “Machine learning is a core transformative way by which we are rethinking everything we are doing.”
Millions of queries a second are interpreted by an artificial intelligence system nicknamed RankBrain
RankBrain can correctly identify search queries 80% of the time while Google engineers have a 70% success rate.
RankBrain is the third-most important signal contributing to the result of a search query.
RankBrain cannot learn on its own yet.
RankBrain requires periodic updates.
Gary Illyes, Webmaster Trends Analyst at Google claims, “SEO Magic” still works
Acknowledgments: Thanks to Brandon Schakola for inspiration and support. Thanks to the significant research and articles by all the thinkers and writers in my industry.
TSA Upcoming Series: How to Thrive after the Machines have already Risen
Important questions we will try to answer in the new TSA series:
Why is it a terrible idea to bad-mouth Ray Kurzweil’s dad?
What do Stephen Hawking, Elon Musk, Steve Wozniak, Daniel C. Dennett, Noam Chomsky and Bill Gates all have in common?
Why does Alphabet keep buying up robotic companies?
Why is Warren Buffet so terrified of self-driving cars?
Why is Alphabet’s Executive Chairman, Eric Schmidt, formerly and Google President, assurance not to worry about AI seem so terrifying?
Clark, J. (2015, October 26). Google Turning Its Lucrative Web Search Over to AI Machines. Retrieved from Bloomberg: http://www.bloomberg.com/news/articles/2015-10-26/google-turning-its-lucrative-web-search-over-to-ai-machines
Clark, J. (2015, October 26). Jack Clark on Twitter. Retrieved from Twitter: https://twitter.com/mappingbabel/status/658780659889143812
Kedar Dhamdhere, T. S. (2015, August 11). Using concepts as contexts for query term substitutions. Retrieved from google.com: https://www.google.com/patents/US9104750
McFarland, M. (2015, March 16). Google’s Eric Schmidt downplays fears over artificial intelligence. Retrieved from The Washington Post: https://www.washingtonpost.com/news/innovations/wp/2015/03/16/googles-eric-schmidt-downplays-fears-over-artificial-intelligence/
Metz, C. (2015, September 28). Google’s Quantum Computer Just Got a Big Upgrade. Retrieved from Wired.com: http://www.wired.com/2015/09/googles-quantum-computer-just-got-a-big-upgrade-1000-qubits/
Slawski, B. (2015, October 26). Investigating Google RankBrain and Query Term Substitutions. Retrieved from http://gofishdigital.com/: http://gofishdigital.com/investigating-google-rankbrain-and-query-term-substitutions/
Slegg, J. (2015, October 27). Google’s RankBrain: 9 Industry Experts Weigh In. Retrieved from thesempost.com: http://www.thesempost.com/googles-rankbrain-9-industry-experts-weigh-in/
Slegg, J. (2015, October 27). RankBrain: Everything We Know About Google’s AI Algorithm. Retrieved from The SEM Post: http://www.thesempost.com/rankbrain-everything-we-know-about-googles-ai-algorithm/
Source: The Search Agents
Link: RankBrain Rising