Machine learning can seem like an abstract concept that is too difficult to wrap our human brains around. Part of that feeling is based on misconceptions surrounding the concept of machine learning. What are some common misconceptions about machine learning?
The models computers learn are incomprehensible to humans
One of the most common misconceptions surrounding machine learning is that humans cannot comprehend what the computer is learning. In reality, while some models are indeed complex and difficult for humans to understand, most are not. Don’t immediately assume that you cannot understand the same exact way that the computer or machine can.
It’s all About the Right Algorithm
Many people believe that machine learning is simply coming up with the correct algorithm in order to solve a problem or identify a pattern. This could not be further from the truth. Machine learning is much more based in the data than an algorithm. As the CTO of Sift Science Fred Sadaghiani states, “data is orders of magnitude more important than the algorithm you use or any technique that you’re applying.” When we refer to data, that means both the amount and the quality of data. The more quality information that the system receives, the better the results will ultimately be.
Machine learning is absent of human bias
It is almost impossible to completely eliminate human bias from machine learning.
Quality data is crucial to machine learning; data filled with human bias can greatly impact machine learning applications. One of the best examples of machine learning being absent of human bias can be found in Microsoft’s bot named Tay, released in early 2016. The goal of creating Tay was to determine if the bot would be able to learn from interactions with social media users on certain platforms like Twitter. Within 24 hours, users had taught Tay to be both offensive and racist. Microsoft immediately pulled Tay from the market.
While machine learning can seem abstract and complicated, it is easier to understand once you debunk some of the common misconceptions surrounding it. For example, the models are not incomprehensible to humans. Machine learning it is not only about the algorithm and lastly, machine learning is can be biased by humans, as evidenced by Tay the bot.