Is AI an invaluable medical tool or a threat to the physicians who use it?
Can a computer algorithm replace a doctor?
A few years ago, a question like this might have seemed like something straight out of a high-concept sci-fi story. But as computing power has increased at an exponential rate, and as the amount of data we collect on everything from our eating habits to the number of steps we take grows at a similar pace, a better question might be: is it already happening?
It has already been proven that machine-learning algorithms – which are a subset of what is traditionally known as artificial intelligence – can diagnose illnesses like cancer and diabetes by interpreting digitized images of mammograms and retinal scans. But this, you may be surprised to learn, is just the tip of the iceberg when it comes to the power of AI. As things continue to evolve, imaging may prove to be the low-hanging fruit of machine-learning.
It can all be a bit overwhelming for those of us without computer engineering degrees. So let’s take a step back and try to better understand where the world of artificial intelligence and medicine overlap—and in which directions that overlap might be growing,
Machine-learning is exactly what is sounds like: inferences made from the study of patterns within vast numbers of data sets, including statistical models and other algorithms. Sounds familiar? It should, because this is essentially how one might describe the traditional diagnostic capabilities of a physician: they make inferences based on observable data and available statistical models.
But where human doctors are limited by – well, the human brain’s capacity to store and access information – and must rely on pre-existing (but proven) methods, an artificial intelligence program can instantly scan a bewildering amount of health data – from medical research both current and historical, to health records and insurance claims – and make correlations that no single human being could possibly make (except for maybe The Good Doctor, but he’s a special case…and also totally fictional).
Why are these patterns and connections important? Because, with access to them, a physician’s analytical speed becomes much faster, and they can get a much clearer perspective on the intervention and treatment options available to them.
And the benefits of artificial intelligence go beyond mere diagnostics. This capacity to identify new patterns in the function of the human body may someday change the way that we approach performing surgery and repairing damaged biological systems.
So: is artificial intelligence a replacement, or a tool?
However you want to define it, machine-learning algorithms will very likely redefine the way medicine is practiced over the next few decades. As AI becomes more prevalent in clinics and hospitals, certain pressures will be lifted from physicians—most likely, in the near term, the burden of memorization of empirical information.
With that, however, a doctor’s ability to clearly communicate this information to the patients they are treating will become vital. What is traditionally known as “bedside manner” is likely to become an even more essential trait in medical professionals, and an aptitude for creative problem-solving – specifically, how to apply that knowledge in the most effective way – will be one of the hallmarks of an effective physician.
AI’s predictive capabilities are beginning to flourish, as well. Some deep-learning models have been able to predict breast cancer up to five years in advance, and does so with a much higher level of accuracy than current clinical models. Scientists are also developing systems that can monitor people with multiple sclerosis and Parkinson’s disease, and can predict whether a person will develop Alzheimer’s disease.
However, as with so much else when it comes to the health care industry, there are ethical concerns. Particularly about the abuse and misuse of patient data. If an algorithm identifies you as high-risk to develop a chronic disease like diabetes, can insurance companies use that information to justify a higher premium? If the aforementioned deep-learning model predicts that you are likely to develop cancer in the next five years, might that make a potential employer think twice about hiring you?
Some worry that AI might be given an authority that takes control away from clinicians. Others believe that many AI models are just as flawed as the people who create them. Because machine-learning algorithms are dependent on the data sets they’re provided, they can still be subject to the same human biases that an individual physician might be. For example, certain algorithms that have been fed data from primarily Caucasian patient populations suffer a drastic drop in efficacy when diagnosing nonwhite patients.
If nothing else, this proves that the integration of AI technology isn’t a silver bullet that will solve all the problems faced by today’s health care industry. Ultimately, patient care is only as good as the doctors who administer it. Which means that the rise of AI doesn’t necessarily relegate the role of a human physician, it actually makes them – and their very human capacity for empathy and compassion – all the more important.
However this all plays out over the next few decades, even AI skeptics admit that the opportunities far outweigh the risks. As the technology becomes more efficient, reliable, and cost-effective, it may go a long way to democratizing health care across the world—patients everywhere, from rural areas to third-world countries, will have access to high-quality diagnoses.
Which might seem like the sort of far-fetched, utopian idea that you’d find in a high-concept sci-fi story—but, as we’ve learned, we’re kind of living in one right now. And for many patients – and the doctors who treat them – that’s a very good thing.