January 26^th 2014, the day Google acquired DeepMind Technology, marked the birth of widespread use of Artificial Intelligence in Google. DeepMind Technologies’ goal is to solve intelligence by combining techniques from machine learning and systems neuroscience to build powerful self learning algorithms. What it does is it takes in a large set of raw information to its algorithms so that the system itself can learn the very best representations in order to use those for action or classification or predictions. After implementing the system in different wings of Google, in 2015, Google engineer, Amit Singhal (who is in-charge of the company’s search engine) built a deep learning system called RankBrain. The system helps generate responses to search queries, and is proven to play a role in “a very large fraction” of the millions of queries that go through the search engine with each passing second.

RankBrain 101

RankBrain is a machine learning (AI) algorithm that Google uses to sophisticate the search results. Moreover, it allows Google to learn about search queries that are unfamiliar to it. The system’s job can be narrowed down into two steps:

1. Interpreting search queries (keywords):

Before RankBrain, when a person would search something, Google would scan pages to see if they contained the exact keyword someone searched for. For example, if a person searched for “the bezel-less phone by Samsung”. Google would look for pages that contained the terms “bezel-less”, “phone” and “Samsung”. As of now, RankBrain actually understands what you’re asking. Today, if you search for the same thing, the SERP (Search Engine Results Page) would directly narrow the results down to, for example, “Samsung S8”.

2. User interaction with the results (user satisfaction):

Initially if a user searches for something and the results don’t seem to be up to their mark, the RankBrain is going to take a few backlashes but surely provide better results the next time they search for something similar. They do this by showing a set of search results that they think the user will like. If a number of people like one particular page respective to their search, they will give that page rankings boost. Incase the user didn’t like it, they will drop the page and replace it with a different page. The next time a user searches for that keyword, they will record how it performs. The user interaction is recorded according to the following criteria:

• Organic click-through-rate
• Dwell Time
• Bounce rate
• Pogo-sticking

All in all, Google has been developing RankBrain similar to what is widely known in the marketing field as “behavioral targeting”. Their system is developing in every way possible to give us the best result. Hence the day is not too far that we step into the future where the internet would know what we are looking for before we finish our typing, or even type.