Machine Learning the Unexpected
Last week we learned about some of the natural, or expected use-cases of machine learning. This week I would like to show you some of the unexpected places machine learning is in use today. These are the things that you probably don’t really associate with the complexities of machine learning. Usually this is because relatively simple every day tasks are something that computers must learn in order to accomplish so we don’t consider them to be learned responses.
Have you ever been shopping for something on Amazon and bought something else just because Amazon recommended it? Where did that recommendation come from? A few years ago it was very hard for Amazon to make sales based on the recommendations, because they were not using machine learning techniques to model the recommendations, but using simple sales data and suggesting items that people frequently bought together. Amazon engineers have worked hard in developing machine learning techniques related to “Recommendation Systems” and have become the leaders in the technology. Amazon uses very large computer systems to train the algorithms based not only on past sales, but also on what customer shopping preferences and personalizes the recommendations to the individual. In my case I wind up seeing lots of recommendations for model railroading, 3D-printing and electronics. I even see electronics suggestions when shopping for food items, and most people don’t purchase food and electronics together, not even me.
The next example uses the same technology under the hood. YouTube, Netflix and other streaming services recommend shows based on your individual viewing habits. If you want to really mess with your friends, use their YouTube account for about a week. My kids have used mine to watch Minecraft videos and now every other video suggestion is Minecraft. A majority of my YouTube viewing are from two streamers, The King of Random (TKOR) for his crazy approach to science experiments and Robert Murray-Smith for his approach to clean energy.
One of the largest use-case for machine learning is actually a major field of study, Data Mining and Big Data. This is a topic all of it’s own, which basically means finding useful information in mountains of data. You might say searching for a needle in a haystack Computers have the ability to compare massive amounts of data rapidly and organize it in new and interesting ways. I grew up in a time when you went to the library to research a topic and every library had a card catalog, which was a bunch of drawers full of cards sorted by Author, Title and Subject. Those cards were all numbered with numbers from the Dewey Decimal System which helps to categorize and locate books in a library. Computers have found new and unique ways to organize data, much like the age old Dewey Decimal System and machine learning is a large part of this process.
The last and possibly largest use-case for machine learning is called “Regression Techniques” which are used to better access the market in stocks, finance and Real Estate. So as you probably realize after the last several weeks, machine learning is everywhere and it seems to be here to stay. Whether you are research and development, sales and marketing, or just watching videos online you rely heavily on machine learning every day of your life.
Until next week, stay safe and learn something new.
Scott Hamilton is an Expert in Emerging Technologies at ATOS and can be reached with questions and comments via email to sh*******@**********rd.org or through his website at https://www.techshepherd.org
