Humans don’t stand a chance.
We are expensive to train, subject to bias, and haven’t even reviewed all of your test results!
We are human doctors, and we will be replaced.
I studied at one of the top medical schools, and I’ve been in practice for over a decade. I can read the writing on the wall. The machines are coming for us.
As much as it pains me to say this as a carbon-based life form, we have our drawbacks.
The first and most obvious problem is that it takes a lot of money and time to train us. Medical education involves four years of college, four years of medical school, and three to eight years of residency specialization. That’s a lot of time and effort!
And once we finish all that training and have an MD after our names, we still suffer from the fact that our brains are human.
No matter how hard we study, we can’t know all of the available information. Hundreds of new journal articles and research papers are authored every month. We couldn’t possibly read them all, let alone remember everything. It just isn’t achievable. We are out of date from the moment we leave school.
Perhaps more problematically, we are subject to biases that can lead us astray. For example, the availability heuristic causes us to overestimate the likelihood of events that we can most easily recall. Imagine if a physician recently examined a patient with achy joints and correctly diagnosed her with Lyme disease. From now on, that physician will reflexively (and incorrectly) assume that every new patient with achy joints also has Lyme disease rather than a more likely problem like arthritis.
Another common mistake is confirmation bias. Once we reach a tentative conclusion, we subconsciously tend to look for information that supports our conclusion and discount evidence that disputes it. For example, if a physician believes that a migraine is causing a patient’s headache and vomiting, they may unintentionally miss or ignore evidence that the patient actually suffers from a brain tumor.
Computers are immune to these errors.
How Computers Could Do It
Let’s face it; artificial intelligence (AI) is awesome. AI can do so much more than suggest friends on Facebook or guess your search query on Google. While you were sleeping, computers have been composing symphonies, soundtracks, and trippy paintings.
On a more serious note, artificial intelligence can now diagnose complex medical conditions. A 60-year-old lady in Japan suffered from a rare form of leukemia which stumped doctors at the prestigious University of Tokyo. The doctors eventually turned to IBM’s Watson for answers. The supercomputer reviewed millions of papers and correctly diagnosed her.
It’s easy to imagine a patient filling out an electronic questionnaire, a specially trained nurse or physician assistant (PA) conducting a physical exam, and a computer taking care of the rest. Presto! You have a doctor visit without a doctor.
Sure, there are a few bugs to be worked out. AI isn’t yet ready to make all of the cognitive judgments one would expect from a physician. But it will be soon.
Technology and artificial intelligence are improving at an exponential rate. What computers can’t to today, they may very well accomplish tomorrow. Within a few years, technology may be nearly unrecognizable to today’s users.
Rapid advances in AI technology, coupled with the proliferation of electronic medical records (EMR), will produce machines with unimaginable potential.
It is just a matter of time before Watson, Google’s Brain, and other similar systems are trained as well as physicians. And these machines will know all past and current medical information, your complete medical records, and won’t forget anything. They’ll have access to all lab results and will be able to read imaging studies and slides without the aid of radiologists or pathologists. They will be an instantaneous one-stop shop for healthcare.
A Role for Humans?
Humans still have one advantage over our soon-to-be robotic overlords: empathy. After all, how do you think that our medical forefathers survived despite advocating dangerous “treatments” like leeches and mercury? They offered compassion.
While I have no doubt that computers will soon make complex diagnoses and treatment recommendations better than their physician counterparts, I suspect that machines will remain unable to determine the final treatment.
There is much more to selecting a treatment than probably of success.
Yes, if you have an infection and you’re choosing between different antibiotics, you’ll go with the computer’s recommendation. That’s a no-brainer. However, treatment decisions are rarely this black and white.
Often, patients and physicians must weigh competing values in a way that computers simply cannot.
It is worth injecting a local anesthetic to make childbirth less painful even if there is a minute risk of paralysis? Should people with cancer endure nausea and toxic effects of chemotherapy just to potentially prolong their lives weeks or months?
Artificial intelligence can’t help us here.
Ultimately, it seems likely that we will develop some hybrid model with computers carrying the cognitive load of making diagnoses and generating treatment options while physicians will return to their original role of offering care.
There’s one more reason to keep us humans around. Compassionate physicians will be our last line of defense against a dystopian world of machines placing a value on human life.
It’s easy to imagine an algorithm that calculates the cost vs. quality-adjusted life year (QALY) impact of each treatment option. It’s not hard to see what comes next. The computer won’t even offer potential therapies that are deemed too expensive. Do we trust a robot to be a patient advocate?
Consider the cancer example from above. Would we want to live in a world where our robotic masters won’t treat grandma’s cancer because they determined it was too expensive? To prevent this draconian future, human physicians must retain a fundamental role in healthcare.
As the technology evolves, physicians, researchers, patient advocates, and political leaders will need to work together with an open mind to tap this lifesaving resource in an ethical and humanist way.
AI is accurate, affordable, abundant, and may well be the solution to the healthcare crisis.
The computer will see you now.
Gregory Charlop MD
Dr. Charlop is a physician practicing in California, an artificial intelligence enthusiast, and the CEO of Visionary Remodels.