Machine Learning Use-cases
By Scott Hamilton
The last two weeks, we covered the basics of machine learning algorithms, how they are classified and a little about how they work. Now we are going to learn about how they are used in everyday products. There are some natural use-cases that even the person who doesn’t believe machine learning exists must admit that there is some extremely hard, mind-crunching computer wizardry driving the technology.
The first of these natural use-cases for Machine Learning is Natural Language Processing. This is where a machine learning algorithm has taught a computer how to interpret human speech or text. When you call your bank, auto insurance company, computer tech support or any other phone based service and a computer voice answers, asking you how it can help, this is natural language processing. Most of these systems have a fairly limited language set that they understand, but more and more of them are becoming complete enough that they can simulate human interaction. Natural language processing is not limited to speech, but also written text. The most advanced of these systems is the auto-correct, grammar checking, spell checking, and auto-complete technologies you see in use on sites like google and Facebook. If you have held a chat session with Amazon customer service, you were talking to an Artificial Intelligence (AI) customer service algorithm. I find the whole idea rather amazing.
The second is similar in nature, Computer Vision, where the AI is taught to recognize things in the real world. Most of this is accomplished through pattern recognition techniques. Computers are really great at detecting the edges of an image and, as such, utilize the edges to detect patterns and recognize objects. These are primarily used in factories to inspect, orient and validate parts in an assembly line. However, there are some very clever use-cases of machine vision in medicine to detect cancer.
The third combines the first two, along with computer control systems, to make self-driving vehicles a reality. I really struggle to trust self-driving cars, but not so much because of the car, but mostly because of the random interaction with human drivers. We are far too unpredictable.
These are just a few of the applications of machine learning. Next week you will be surprised to learn that quite a few seemingly simple tasks you do online everyday are tightly intertwined with Machine Learning and Artificial Intelligence.
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*******@te**********.org or through his website at https://www.techshepherd.org.