Common Myths of Machine Learning

AUTHOR:  Dan Dang

Whenever the phrase “machine learning” is mentioned, one might be inclined to think of artificial intelligence, which then leads to thinking of the movie “Terminator”, and then needless panicking ensues. There is no need to worry because machine learning doesn’t mean the end of the world. However, there are many unsure ideas around the topic of machine learning. Here are some of the most common myths of machine learning debunked!


Myth #1: Machine Learning is Only For Predictive Analytics


This is a common misconception and with good reason since machine learning is often used for predictive analytics. However, machine learning can be applied beyond that. Anomaly detection, data classification, and data recognition are some of the many applications derived from machine learning algorithms, outside of predictive analytics. An anomaly detection algorithm can be embedded into security cameras to alert you of any unusual activity happening at building entrances. Data classification is used in Google’s search engine to bring up the most relevant search results. And finally facial recognition is used in Facebook to tag friends in photos. Bet you didn’t think of that one!


Myth #2: Machine Learning Eliminates Human Bias


We as humans can become emotional which affects our decision-making, thus leading to human bias. With machine learning, computers are built to process data and make decisions with minimal human supervision. However, human bias is not completely eliminated. To use machine learning someone must decide the use scenario to follow. Afterwards, someone needs to choose the data sets to analyze. Next, someone has to select the machine learning processes. Finally, outcomes are interpreted by, you guessed it, a human being. Each step of the procedure has a certain amount of human bias, regardless of how advance machine learning is.


Myth #3: Machine Learning is Only Used with Big Data


This myth developed from the current rise in importance of big data. Today’s big data is too large for basic analytical methods. This is where machine learning truly shines; machine learning can extract valuable insights from copious amounts of data to create accurate models. However, machine learning may not even involve data at all. For instance, one use case is image processing for facial detection on phones. Machine learning algorithms are run against one hundred images or so to train a model to detect faces. One hundred images is rather a small training set size. Note the more images fed into the algorithms, the more accurate the model becomes. Can you say, “selfies”?


Myth #4: Machine Learning Doesn’t Apply to my Company


On the contrary, machine learning can be more beneficial to your company than you may think. Anti-virus software use machine learning to better detect malicious software on computers. Autonomous self-driving cars rely on machine learning to interpret sensor and image data in order to identify objects around the car. Pandora Radio utilizes machine learning to predict a person’s taste in music for better recommendations. These are just a few examples showcasing how diverse machine learning applications can be. Next time you have a problem with your business, machine learning could be the solution.


Myth #5: Judgment Day is Coming


Again the world isn’t coming to an end. Machine learning will not give birth to artificially intelligent androids that will destroy mankind. Many may believe that machine learning and artificial intelligence are one and the same, however, that could not be further from the truth. Machine learning is derived from the study of pattern recognition and learning theory, which originated from artificial intelligence. Machine learning is essentially a tool; you are the one choosing to use it or not, much like using a cell phone. Human decision-making is a vital part of machine learning application.




There you have it, five common myths of machine learning, busted! Now that you’re more knowledgeable about this field, you can make better well-informed decisions of using machine learning in your enterprise. Medium One’s data intelligence platform has built-in machine learning libraries, making it easy for businesses to utilize on a daily basis. For more information about machine learning or the IoT world, you’re already in the right place,'

Author: Joyce Li