Learning deeply and deeply learning: The nexus between human and machine learning
At first glance, the phrases learning deeply and deeply learning appear to be no more than a play-on-words ploy to catalyze clicks into this article. Well, not only. Each phrase has its own real independent meaning as well as complimentary connection to the other. I will (a) define each phrase, (b) highlight its connection to each other, (c) point out the emerging nexus between the two, and (d) identify practical applications.
In A nutshell learning deeply is related to the human experience of learning. Deep learning relates machine learning. A human can learn deeply through a multitude of experiences, and a machine with artificial intelligence can perform more like a human.
How I define learning deeply is full immersion into new experiences, while allowing the many parts of ourselves (intellectual, emotional, social, physiological, occupational, spiritual) to be influenced by how they differ from our past.
Deep learning (LeCun, Bengio, & Hinton, 2015) refers to advanced algorithms to help machines learn to mimic structures and functions of the human brain. It is also referred to as artificial neural networks.
The Nexus of Human and Machine Learning. In his seminal work, Experience and Education, John Dewey (1938) stated that we learn when our past experiences are carried forward into our new experiences, and we decipher the difference between the two. What is left is what has been learned. When I conceptualize Dewey's idea that experience leads to learning, I see a mathematical equation. As a data scientist and statistician, I know you are not surprised. Munyon's Calculation of Dewey's Learning Theory (2008) would read like this: Past Experiences + New Experiences = Learning.
The concept of deep learning has roots in Aristotle (300 BCE) and new developments as recent as 2012 (Wang and Raj, 2017). In recent years, with the crescendo of data science across numerous fields of learning, deep learning has become a prominent part of the conversation featuring the future of machine learning. Machine learning can now be simply categorized into two types: supervised and unsupervised. We supervise machines' learning -- supervised learning -- when we enter program commands for specific outputs (e.g., print file). However, with advanced algorithms, we can edge closer to unsupervised learning to allow machines to mimic the structure and function of and human brain.
When we supervise machine learning, we input and receive repetitive and predictable outcomes (with the exception of human error). When we implement unsupervised machine learning, we must prepare the machine to learn, to an extent, how the human brain learns, which includes 90% experience. Therefore, I posit that in order for machines to deeply learn (mimic the structure and function of the human brain), they must be allowed to learn deeply (become fully immersed in new experiences while comparing them to past experiences).
So now we face numerous questions and issues related to programming code, ethics, scope, function, purpose, and so much more. Stay tuned to your televisions, smart devices, and radios for updates on the nexus between human and machine learning. Many data scientists and stakeholders will be talking about the nexus. The day your device says "I" may be the day you put it down and take a step back. Until then, let's learn deeply about deep learning. Shall we begin?
