In Alan M. Turing’s 1950 paper, Computing Machinery and Intelligence, he makes and refutes arguments that opposed the idea that machines, like humans, could think. One of them, “The ‘Heads in the Sand’ Objection,” states that “The consequences of machines thinking would be too dreadful. Let us hope and believe that they cannot do so.”
Sixty-nine years later and, although many of us would now agree that “yes, computers think,” we are still troubled by the uneasy feeling that we could be replaced by them. I don’t believe we will be, but I do think we are at the point where symbiosis with artificial intelligence (AI) expands our ability to provide outstanding medical care.
Where once we consulted the four-inch-thick Physicians’ Desk Reference when looking for drug interactions, we are now far more likely to rely on some form of memory-freeing software on our smartphones or to consult the options given to us in Epic. We are fairly confident and comfortable with computers assisting with the storage, easy retrieval, and automatic updating of data. We understand that this can free our brain up for more complex thoughts.
With this improved ability to manage data, we have increased the amount and complexity of what we collect exponentially. Even the best of human brains cannot sort and evaluate the vast quantities of accessible data in the time available to us. Since the value of the information is only realized if we are capable of identifying, extracting, and analyzing the data relevant to our question, we are turning to AI (programming computers to perform human tasks) and its subset machine learning (where the computer is programmed to learn rules by recognizing patterns in data rather than being given the very rules).
AI is already in our lives in many ways: in voice recognition for dictation, machine vision to speed diagnosis, and predictive software that notifies us of potential errors in the medical record. It is not just a tool, like a microscope or stethoscope, to improve our senses. AI provides “thinking,” analysis of complex problems via powerful algorithms that result in rational assumptions. AI never gets tired or makes random mistakes typical for people (unless the data input is flawed or incomplete). However, AI needs to be complemented by human factors like collective folk wisdom, common sense, optimism, compassion, and understanding the personality and needs of the patient in front of you. It is only together that we can move forward more effectively. Some examples:
- Ideally, every medical decision would be reviewed by a team of relevant experts who could provide multiple levels of guidance on each decision and diagnosis. But there are not enough medical experts to staff it, it takes too long to read through a patient’s history, and privacy laws that prohibit this type of review would stop efforts before they even started. AI can evaluate a decision based on all the available data—both of the patient longitudinally and the other relevant cases recorded to date—in a fraction of the time it takes humans to do it.
- Patients should receive the most effective medications rather than the ones most familiar to the prescriber. But how can physicians stay on top of all the fluctuating medications, interactions, and side effects? AI can not only manage information, but also analyze genetic makeup, disease stage, and other factors to make the best recommendation.
AI gives us the ability in principle to generate a digital twin of the patient to predict their health trajectory. It supports clinicians in the tasks of knowing the latest medical science and individual patient’s health. By doing some of the heavy lifting of data management, AI can help reduce physician burnout. Ultimately, we will come to the tripartite coproduction of health care: patient, physician, and computer working together.
Artificial intelligence and machine learning are not going to solve all the present or future problems in medicine, but they are providing us areas ripe for research and rich in possibilities. In the words of A. M. Turing, “We can only see a short distance ahead, but we can see plenty there that needs to be done.”