Artificial Intelligence (AI)
What’s artificial intelligence (AI)? It's computers learning things. In 1959, Arthur Samuel had his computer play tens of thousands of checkers against itself. The computer was eventually able to beat Arthur Samuel at checkers. His definition of AI is that it’s a field of study that gives computers the ability to learn without being explicitly programmed. Professor Tom Mitchell at Carnegie Mellon uses three terms for AI that include "E" for experience of learning a "T" for task that produces "P" for performance.
Computers can learn by supervised learning which means teaching the computer how to find solutions by using a regression model such as finding the best curve for a house-selling price or a classification model based on a positive or negative answer such as for diagnosing breast cancer. Unsupervised learning is when the computer learns by itself. Given a data set, the computer can find structure. For example, a clustering algorithm can be used for searching gene maps, astronomical data, social network groups or segmented market.
Jensen Huang, CEO of Nvidia likes the term AI to include all the sub-artificial intelligent names such as machine learning, deep learning, and artificial neural networks. Huang says that AI means to teach computers to do something. AI is software writing software. Put data in and get an answer out. Train the data to generate software. AI training AI to develop AI. Data scientists develop training data strategies and develop software with a training strategy that will write software by itself. Build an intelligent machine – data comes in, action is taken, and products develop. Data in, insight out.
Successful AI has a 5-step closed loop with the following steps: (1) Teaching - data input (2) Comparing - data are compared to a database. (3) Result - this comparison is analyzed for an answer (4) Action - action is taken based on the answer (5) Close the loop - all of this information is added back into the computer. For example, (1) Symptoms, x-rays and laboratory tests are fed into the computer. (2) This information is compared to a diagnostic database. (3) A diagnosis is established. (4) Action is taken in terms of a phone or app message, medicine treatment, or even stimulation to a certain region in the brain. (5) All of this information is fed back into the computer for continual learning. This is a non-bias algorithm arriving at a diagnosis and treatment, rather than based on a physician’s training and conditioning. AI is not used to replace the physician. AI will augment the physician’s diagnostic and treatment decision-making capability. The human brain by itself and the computer by itself can produce good results. The combination of the brain and computer produces unprecedented and unstoppable results.
AI healthcare examples include real time monitoring of anesthesia in the operating room where all data is continuously scanned in real time and immediately produces alerts about pending problems and can be designed to administer medications for stabilization. It’s like an autopilot. AI can be taught how to interpret radiology studies. Siri and Alexa are AI.
Genomics and AI have been combined to discover new anti-cancer medications. For example, people have cancer protection genes that will stop cancers from developing. An AI program can detect cancer protection gene mutations among the gene database, the protective mutation variants can design anti-cancer medications. These databases are so massive that AI algorithms can predict cancer medications better than people, and in addition, these systems keep growing enabling better discoveries. Pharmaceutical companies are using AI to redo the workplace model and the workforce. AI will improve the efficiency and length of time for drug development.
AI will take diagnosing disease to new places and redefine diseases. For example, a fibroblast growth factor gene produces a protein for healthy cell growth and regulation. Dysfunction of this process can result in a specific organ system cancer. AI will detect this dysfunction at the precise location that causes a specific cancer and develop an anti-cancer drug to stop the cancer.
AI is actively used in neurological and psychological disorders. AI has shown that Alzheimer’s symptoms are caused by inflammation. The more inflammation, the worse the symptoms. AI can be used at home or elsewhere to determine the degree of inflammation. AI applications can be developed for behavior diagnosis and take advantage of brain plasticity for treatment. An ALS treatment molecule drug has come from an AI platform. AI can provide streams of data for management of psychiatric disease. AI will change the current psychiatric disease names based on symptoms to psychiatric diseases based on objective data and determine the underlying cause. New diagnostic markers will be developed such as using the tone and content of someone talking to establish a diagnosis of depression, and an application can be developed for treatment.
AI is currently useful for detecting certain types of infections that occur among patients taking antibiotics. Humans can do this with some success, but there is a limited amount of data analysis the brain can manage; however, an AI system can manage and interpret massive amounts of data on a real-time basis 24 hours per day.
AI is improving our lives for the better. AI is going to be a successful part of all aspects of our lives for many years to come.
Gary R. Epler, M.D. in Boston
Best-selling author of “Fuel for Life: Level-10 Energy” and “Peak Performance and Leadership”