“Humans are never satisfied.” That’s what my high school English teacher would always tell my class. The angst-y fourteen-year-old that heard this the first time thought, “Well that’s kind of pessimistic” and viewed this as something negative. I refused to believe it and dismissed it as no more than a negative way of looking at life that would make anyone feel miserable. However, I started to notice the truth of my teacher’s words as I grew older. Humans are never satisfied, and we should all be glad that we are not. It is because of this that society advances. With each accomplished goal comes the hunger for a new one.
With society’s continuous augmentation of knowledge comes the increased intelligence of the technology and tools around us. Gone are the days of mindlessly purging someone to return balance to the four humors and in are the days of medications and vaccines that are overlooked by most. However, no matter how far we have come in our advancement of the medical field, the desire to develop tools to better help diseases, conditions, and our gain of knowledge is a fire that repeatedly ignites itself. While it is a great milestone, medicine is not satisfied with just the development of vaccines. The current thirst is to accelerate the process of when a doctor can start helping a patient to prevent detrimental and potentially life-threatening results, which can be accomplished with finding tools that pinpoint diseases and conditions.
The Director of the Institute for Next Generation Healthcare at the Icahn School of Medicine at Mount Sinai, Joel Dudley, best exemplifies this when he told AAMC News: “I’m not as interested in dragging 2017 medicine into the future as I am in building a new system that replaces it. There’s really fertile ground right now for a pretty big shift in health care.” Rooted in aiding artificial intelligence, machine and deep learning seem to be the answer to Dudley’s and the medicine field’s thirst. Machine learning is a way to essentially teach a program to code itself given certain information in order to achieve a particular goal, as well as progress itself. The entire coding process is automatically generated with little human interference. The impressive part of machine learning is that once the algorithm has been given a set of information, if a set of information is given to it in the future, based on how similar the two sets of information are, the algorithm will be able to detect a fact that binds these sets of information together. Say one feeds a program information on the symptoms of patients with mild Alzheimer’s from WebMD, which includes loss of recent memories, decreased interest in work and social activities, and communication issues. The program will compare one patient to another patient that has mild Alzheimer’s in one algorithm, and if the patient has most of the symptoms, then the program will determine that the patient has mild Alzheimer’s.
Deep learning is a way to achieve machine learning. According to Medium, deep learning was actually motivated to imitate the brain in its “structure, function” and “interconnection of many neurons.” In deep learning, a task is split into several others. Deep learning produces more accurate results than machine learning could alone because of its ability to pinpoint specifics. There is an algorithm to solve each of these specified tasks or to learn how to detect a certain property of a feature. For example, instead of having one algorithm to figure out if a patient has mild Alzheimer’s, there is a separate algorithm to determine if a patient has each of the symptoms.
Deep learning has the potential to solve problems and tasks that would have seemed impossible from a machine before. To put this in perspective, one should look at Google’s famous 2012 study on deep learning regarding cats. Computer scientists at Google had given millions of YouTube video thumbnails cats to a neural network without providing information on the properties of a cat to see if the neural network would be able to realize these properties. At the end of the study, the neural network was able to detect cats 75% of the time when provided images. Imagine how much more rapid and accurate the processes to identify something tricky like a skin condition such as rosacea could become using deep learning.
Many hospitals and universities have already started researching and taking advantage of these tools. For one, Georgia Tech, right next-door to us, in cooperation with Sutter Health conducted a study that showed that deep learning can be used to determine signs and symptoms of heart failure nine months before doctors could. In addition, the San-Francisco start-up Arterys developed a neural network that would measure how much blood flows through the heart using any MRI machine. Sudden and seemingly unexplainable events like heart attacks, which is part of the number one cause of death in America (heart disease), can be vastly aided and prevented with the use of the more accelerated process stemming from deep learning. According to NVIDIA, Massachusetts General Hospital’s Clinical Data Science Center “compare[s] a patient’s tests and history with data from a vast population of other patients to improve detection, diagnosis, treatment, and management of disease,” of which NVIDIA itself is a founding partner of. It would be interesting to see the possibilities made available if a nearby strong research institution that focuses on artificial intelligence, such as MIT’s planned future college for artificial intelligence, would aid in the design of the deep learning at the Clinical Data Science Center. Cancer may potentially be prevented with deep learning. Researchers at University of Toronto are advancing studies in cancer with deep learning by developing a neural network that pinpoints mutations that evoke cancer.
Machine and deep learning are certainly engineering masterpieces. While it may simplify processes, the complexity of the countless amounts of linear algebra via dimensional analysis (determining eigenvalues, matrix transformations, orthogonality in inner product spaces, etc.) and code encapsulation (in terms of organization) should not be taken lightly. In fact, the most harmful threat that machine learning and deep learning pose is that many scientists do not know about and what goes into these complex processes, and therefore, how something like deep learning functions. It is important to realize the interconnectedness of disciplines and the reliance of them on each other. Machine learning is capable of aiding diagnosis, targeting medicine, developing drugs, determining valuable candidates for clinical trials, and even making an educated guess of when and where an epidemic will occur. It is easy to see how technology can answer many of the medical field’s challenges. In fact, because of the value that machine and deep learning pose on modern medicine, many medical PhD programs involve computer science courses. If engineers and doctors work together, our capabilities would be endless.
At this point, one may start to realize the common fear of technology replacing us altogether. If technology reaches such a point of advancement that it can fulfill many of the capacities of doctors and others in the medical field, it is logical to think that people would become obsolete. Nevertheless, the “robot apocalypse” has always been an irrational fear because technology is only as smart as its human counterpart that creates and uses them. It is only meant to be used as a tool, similar to how a calculator can never replace an accountant. While deep learning may imitate the brain and may be able to use inductive and deductive reason, technology lacks common sense; while it may be able to piece certain pieces of information together, it will never be able to piece other sets like a human can. Perhaps the most valuable piece of evidence is exactly what my high school teacher told me. “Humans are never satisfied.” At the end of each goal, we always strive to solve another. Technology is no exception to this. Once we are able to develop and apply technology to certain aspects of our life, we will always strive to solve other issues, which can involve the development of current or new technologies. What this means is that people will always have a place in society because there are always issues in each era that people will want to solve. Technology is fortunate enough to bear such job fertility due to the fact that like an essay, it needs constant revision.