Have you ever used Siri to check the weather for the week, for your trip to the beach, or to check if that cough you have this morning means you’re dying and don’t need to go to class? What about those pesky ads on Facebook or Instagram that seem directly related to your interests? What you’re seeing is machine learning at work and directly impacting our daily lives.
Machine learning is programs and algorithms that allow computers to imitate and “learn” human behavior without having to be told how to do every single minute thing. In other words, computers “learn” information directly from the data of their ‘experiences’ without having to be told explicitly how to. The algorithms adaptively improve their performance as the amount of information available for learning increases and are able to better predict the unknown.
This is not different from how our body and mind works. A machine gets an input like our body receives some sort of signal or stimuli. Both computers and our nervous system integrate and process information in order to come to a decision on how to act. The only difference is that machines are limited in which decisions are possible, whereas humans are not. Both humans and machines learn if this output was correct or incorrect in the scope of the situation and question. If our response was incorrect in some manner, we receive feedback and learn which aspect of our response we need to improve upon. We then improve our response the next time we are exposed to the stimuli or input. One of the main ways that we integrate machine learning into our daily lives is through their use in classification; the machine gets an input and learns how and where to categorize it. Your spam folder for your email is a classic example. Your email has a set filter for what it deems as spam or a genuine email. However, if you find an email in your inbox that’s actually spam, or conversely that all important college acceptance letter in your spam folder, you give your email service feedback on its categorization through a button. While it may just be clicking a button to you, this informs and refines the filter for your email. The email service learns by more narrowly defining what is spam and what is a genuine email for your specific account.
For things like the weather for your beach trip or stock prices, factors that are very unpredictable or influenced by many factors, machines use something called regression analysis. Because weather or stock prices are not something that can just be labeled as A or B, machines use all the previous knowledge it has access to make a prediction about what the future might hold based on trends in the past. In addition, machine learning uses your past website visits and browsing histories to determine which ads and products you may want to see.
However, some of the most powerful uses of machine learning come when we have no idea what the categories may be or have no efficient way of categorizing. Machines can go through large data and do something called cluster analysis to find trends or relationships in the data that would be easily missed if people were the ones digging through all the data. At the end what you will have is a set of different groups whose parts are more similar to each other than to another group. Machine learning is extremely useful for analysis like this. Machines don’t experience the boredom or fatigue that inevitably plague humans, and thus, won’t miss many things. Think of sorting skittles by color, except you’re sorting millions and millions of skittles, one at a time. Machines have limits, but they have the ability and storage capacity that just outclasses what the human brain is able to do.
This might sound terrifying; teaching computers to learn and think as humans do. Don’t fret, we aren’t about to enter the AI apocalypse… yet… Computers aren’t freely thinking on their own, making decisions based off a consciousness. Currently, we use these methods of machine learning for many great purposes that make our daily lives easier, facilitate research, and help people make money. Some may fantasize about the science-fiction stories that stem from machines learning so much about human life that they will eventually learn to think and learn on their own. But, it’s important to realize that machine learning is not going to rule the world; there are many limitations. So hold off on the science fiction pitchforks, and embrace and appreciate the ease that machine learning has brought to our lives.
About the author:
|Kush Bhatia is a PhD student in the Department of Genetics at the University of Georgia. In his spare time, he loves reading, drinking coffee, cooking, and gaming of all kinds. He also enjoys working with some high school STEM student organizations, such as the Technology Student Association. More from Kush Bhatia.|