The field of Machine Learning isn’t a new one. In the recent decade or so it has really taken off with exciting new research and technologies. Today, we are going to take a brief look at why that is and try to understand the amazing potential of Machine Learning (ML). If you ever wanted to understand what ML can do, but didn’t know where to start, then this is the place.
What is Machine Learning?
Before we can answer this important question, let’s take a brief moment to understand the bigger picture of “What is Artificial Intelligence?” You have more than likely heard of AI, the idea that intelligent robots in the very near future will take over or maybe . Well, maybe not necessarily like that, but the topic has shown up in pop-culture and sci-fi novels. That idea, that computer systems created by humans are able to perform tasks that require higher intelligence, is the foundation of AI.
Going beyond the hard-coded machines of the past, ML is a subset of AI where computers are beginning to “learn” . You may have seen a similar graphic to the one below depicting AI encompassing Machine Learning and another amazing field, Deep Learning. We’ll cover that topic another time.
What is important to know is that the methods used in ML generally involve algorithms. Those algorithms are given data, possibly images, text, or numbers, and must find the patterns within the information. Then, if the algorithm was able to figure out the patterns while training, it can make an informed prediction on new data. With more training and the proper data, the ML algorithm can get better at the task it is designed to do.
More recent advances have also helped to accelerate the growth of ML. These include larger datasets (that often are made publicly available) and faster computing power thanks to GPUs.
Why is Machine Learning Important?
Despite Machine Learning seeming like it is a recent discovery, many of the mathematical algorithms that comprise it were discovered much earlier. Bayes Theorem and the Least Squares Method were both founded in the 19th century. Markov Chains were found in the early 20th century. In the 1950s, Alan Turing famously created the ‘imitation game’ to determine if a machine was capable of intelligence. A computer was also able to beat the world champion chess player in 1996.
With each passing year it seems ML is getting closer to us. What once was used by a few scientists for study, we now use every day. The web searches you perform on Google. When you take a photo and your camera recognizes your face. When you talk to Siri and ask her a question. On and on. ML is used in numerous fields, from education, to medicine, to business. And we are only still beginning to understand how to use this technology to help and improve our lives.
Getting a Little Technical
ML approaches to learning often involve decision trees, inductive logic programming, clustering, and Bayesian networks. Other types of ML algorithms are very useful for Computer Vision, a field dedicated to helping machines to visualize and understand their environment. Methods often use hand-coded classifiers to detect edges, shapes, or even colors. These CV algorithms use these filters to make sense of the images and “learn” what it is looking at.
However, many of these algorithms are prone to errors for new situations it has not previously seen. For example, have you ever noticed that when you are using your phone and directly looking at it that it can recognize your face? What if you slightly turn your head? Still no problem. But when you go full profile, it often can no longer tell, right? That is most often because the images that trained the ML algorithm probably never included this kind of data. They were simply trained heavily on frontal photos, not profile.
That is where Deep Learning methods can come in later and help to make better guesses about new data.
Summary
After this, hopefully you will have gained a better idea about what Machine Learning is and how important it is for the future. It is also important to note that there are many different methods to use when working with AI.
Next time, we will take a look at some of the important Machine Learning algorithms, such as decision trees, clustering, and Bayesian networks.
References
A Simple Comparison – Very short and nice read on the differences between AI, Machine Learning, and Deep Learning.
AI vs. ML vs. DL – A brief and good introduction to the differences between AI, Machine Learning, and Deep Learning.
The History of Machine Learning – BBC timeline of the history of Machine Learning with plenty of links to more documentaries if you are interested to learn more.
Pattern Recognition and Machine Learning – Written by Christopher M. Bishop. Definitely a good foundation for helping to get started learning about Machine Learning early on. Plus, it’s totally free to download. Just click on the link.
