We all learn in different ways. Some people are book smart, meaning they glean knowledge from reading books. Others learn better through classroom or one-on-one instruction. Still others learn by doing – maybe jumping into an assembly project without reading the instructions.
And then there are those that are more visual – they can better comprehend information when they see examples of it through pictures, videos and other types of images. This is the genesis of what is called deep learning. Deep learning is a subcategory in the study of artificial intelligence (AI), which is simply the practice of machines – typically computers – learning to mimic the thought processes of humans.
Deep learning is focused on learning through visuals, and it has a near-infinite capacity for both learning and applications. In fact, it is based on downloading vast stores of imaged data. The machine can then scan through this colossal amount of information and identify solutions. In this way, it actually mimics the human brain’s ability to identify collected knowledge and memories and evaluate what is relevant and useful for the current query. The difference is that the human brain has only so much capacity to upload and process information; a lone computer has near infinite capacity.
This concept of deep learning is best conveyed with examples, and there are plenty of potential applications. Let’s start with healthcare. A patient presents with multiple symptoms, which could point to any number of medical conditions. His physician could rely on a variety of screens to make a diagnosis, including lab tests, X-rays, MRIs, CT scans, ultrasound, physical exam, his formal education, his personal experience with previous patients, and consultation with any number of other physicians, radiologists and specialists. The sum total of this knowledge base then comes up with a diagnosis, but it might not be accurate. As we’ve observed on television shows such as House, it often becomes a process of trial and error to make an accurate diagnosis.
Deep learning, though, can exponentially improve both the speed and accuracy of this process. Imagine that every physician across the globe uploads his patient files, images, observations, etc. into a centralized database. When a doctor needs to make a diagnosis, he can enter specific personal information and text results about his patient. The machine then scans its vast universe of data to identify the most relevant cases, information and images that match this individual patient’s symptoms. In short, because a machine with deep learning capabilities can store, assess and identify a massive number of variables, it might be able to diagnose patient conditions quickly and more accurately – saving crucial treatment time, money and the discomfort of ineffective trial and error treatments.
Deep learning basically follows the human process of assimilating information to learn by example, only it has the capacity to sort through so many more real-world examples than any one human brain can compile, let alone assess.
The following instances are just the tip of the iceberg of the many ways that deep learning can be applied to help various professions become vastly more efficient.