“Machine Learning” is a method of data analysis that allows a machine to not only analyze your data, but also learn from the analysis to improve future methods and objectives of analysis
RANKBRAIN: It tries to figure out what people are really looking for, delivering results catered to those needs. RankBrain grows and learns over time to improve its feedback process.
In the past, internet search engines relied heavily on keywords to generate results. However, with the rise of machine learning, search engines can now better understand the intent behind a search and generate more relevant results. Machine learning is used to constantly improve the search algorithms used by a search engine. Through data analysis and pattern recognition, machine learning can identify which results are most likely to be relevant to a user’s query.
Machine learning is used in search engines in two primary ways: to improve the accuracy of search results and to personalize search results for individual users.
Machine learning is used to improve the accuracy of search results in several ways. For example, search engines use machine learning algorithms to analyze the text of documents to determine the most relevant keywords for a given search. They also use machine learning to analyze the behavior of users to determine which search results are most likely to be of interest.
Machine learning is also used to personalize search results for individual users. This involves using data about the user’s past search behavior to determine which results are most likely to be of interest. For example, if a user frequently searches for information on a certain topic, the search
A few examples to elaborate : -
Google Translate is based entirely on machine learning. There is no person sitting at the back and translating the text. Based on the different things that people search for, what they translate and how many edits they make, the algorithm changes itself and gives a better output the next time.
Google search, especially their news articles, will give you results you are most likely to click on and read. The algorithm learns your preferences, your political beliefs and your preferred news sources and then the results are adjusted accordingly. As an experiment, take the help of a friend who is in a different country, if that's not possible then try it with someone who has a different profession from yoursyours, his search should have been different, now log on to your Google accounts on different computers and search for thr same terms,you’ll get different results.
Now about their business model, most of the money that Google makes comes through ads that you see not just along with the search results, but also in various websites. Most of these are PayPerClick, or Google only gets the money if a user clicks on the ads. To get a user to click, Google needs to show ads that are relevant to him or her at that time. And this is achieved by the computer learning the preferences of a user, based on what other similar people have clicked. In short, No Machine Learning, No clicks, No Money.
Hence Machine Learning is essential for Google's survival.
On a side note, Google had a massive faux paus with machine learning, in 2009 they launched an experimental feature called FluTrends, which they claimed would be faster and able to better predict the spread of flu in the US. It would use data based on searches about flu symptoms. Since the data from the CDC would come after two weeks, FluTrends according to Google would be able to help thousands of people better prepared for thr flu. Moreover, it would a victory for machine learning as there was new human involvement. They shut down the device after two years. Why? As this was an entirely machine driven operation, it couldn't differentiate between actual flu and the interest people had in flu. The CDC data was slow as it relied on hospital visits. The more people searched about flu in an area, the more FluTrends would show that people in that areas had flu. This was often off by a factor of 2, meaning it showed twice the number of people having the flu and hence had the potential to cause panic. They are tweaking the algorithms and there are hopes that they will begin FluTrends again, but this just gives you an idea of the importance humans in the entire process. But it also highlights the power data and machine learning have.