Google search engine also includes deep learning based signal
We knew for a while now that Google had embedded some deep learning algorithms into Youtube in order to qualify videos, in Google Now in order to improve the quality of the speech recognition and in Google translate in order to output sentences that actually makes sense.
We discovered 2 days ago that deep learning was also implemented into their core product, the Google search engine. The existence of this system, nicknamed “RankBrain”, was revealed by Greg Corrado, a senior research scientist with the company, to Bloomberg. According to him a first deployed was made at the beginning of the year and a full release was made “few months ago”. It is important to understand that RankBrain is not a new algorithm but one signal among many others (“Top Heavy” for blocking ad overpopulated pages, “Pirate” for copyrights violations, “Pigeon” for pushing local results or “Panda” against spamming) that impact Hummingbird, the main algorithm.
The project was initiated about a year ago by a group of five Google engineers, including search specialist Yonghui Wu, and deep-learning expert Thomas Strohmann. Getting there was not easy but they seem quite pleased with their work: “I was surprised,” Corrado said. “I would describe this as having gone better than we would have expected.” It now became such a success that turning off this feature “would be as damaging to users as forgetting to serve half the pages on Wikipedia.”
It seems that RankBrain is particularly useful for never-seen-before queries, which represent up to 15% of Google queries, that is to say around 450 million queries per day. This AI embeds vast amounts of written language into mathematical entities, called vectors, which the computer can understand. Hence, if RankBrain encounters a sentence it has never seen before, it can make a guess as to what phrases might have a similar meaning and filter the result accordingly.
I personally see two major step forward out of this roll out:
- It brings us one step closer to a human like conversation where Google search is able to pins down the exact question to any basic question it is asked
- It goes along the mobile trend where people ask full human language sentences to their audio assistants
Even more surprising, still according to Corrado, RankBrain has already become the third-most important signal contributing to the result of a search query. He refused to state the two first ones but we can reasonably assume that it is “links” and “keywords”.
Observing one of the largest companies in the world enrolling Artificial Intelligence into the very heart of its core product, re-inforce the idea that machine learning and deep learning in particular is not simply a hyped technology. “Machine learning is a core transformative way by which we are rethinking everything we are doing” said Google’s Chief Executive Officer Sundar Pichai, ealier this week.