How AI Will Change the Job of a Doctor
Discover the moral imperative of AI in providing improved access to healthcare and envisions a future where AI-driven systems enable personalized and cutting-edge care, revolutionizing medicine.
My dad sometimes took me to the bank when I was a child. He ran a business, needed credit, and built close relationships with a few bankers. Each banker had known my father and his business for years, was familiar with the company's ups and downs (and my father’s), and made decisions on how much credit to offer at what terms to keep it going.
Nowadays, I run my own business -- and drive myself to the bank. It’s on a different continent, but the building looks strikingly similar -- and so does the banker. He sits in the same office, wears the same suit, has the same VP title, and always is pleasant and helpful. But there’s a big difference: My banker no longer makes any of the actual banking decisions. He doesn’t decide whether I’ll get credit, how much, or at what price (interest). He doesn’t even decide what products to offer me -- the computer does. He’s basically a friendly and familiar interface to a fully automated back end.
As a result, being a mainstream business banker (or personal banker) is now an entry-level job. It requires training on software tools and customer service instead of profoundly understanding business plans or the community in which the bank operates. The same thing is also happening in retail, where expert buyers and merchandisers with keen eyes for fashion and trend-setting are gradually being replaced by algorithms. It’s hard to see from the outside, but the manager of a clothing, hardware, or food retailer no longer does the core retail tasks of deciding what to buy, where to place it in the store, how to price it, and what to promote. This also happens in advertising, video games, cybersecurity, and legal work.
Where Healthcare AI is Different
Health care seems different. Health care is a clear laggard in adopting digital technology, rated well below finance, oil and gas, real estate, education, and even government. Between the very human-centric service model, the educational background of the people involved, deep-rooted culture, perverse revenue incentives, and the sheer size and complexity of the industry, it’s hard to change.
There’s a strong consensus that AI won’t replace doctors. Arguing that it can go not just against the complexity of what doctors actually do, but such a stance fails to realize the need for a human touch. You will want a human to hold your hand when discussing your cancer diagnosis. Empathy is critical in such life-changing moments. The level of human connection you’ll have with your doctor will directly influence how well you feel, how likely you are to stick with a treatment plan, and how you and your family will remember the trauma for decades to come.
But agreeing that the human doctor will always be there doesn’t reflect the massive changes and risks to their jobs. Doctors essentially do three things: diagnosis (what’s wrong with me?), treatment (what’s the plan?), and prognosis (how long before it gets better?). All three core tasks are being gradually performed by AI systems that employ machine learning, deep learning, natural language processing, and time series forecasting.
There’s not going to be an AI that solves health care -- just like there isn’t one solution to all retail (groceries, clothing, and music all took very different paths to digitization) or all marketing (as the MarTech 5000 technology landscape shows). Rather, based on one process within one specialty at a time, automated systems are showing up and ready for prime time: identifying abnormal chest x-rays, diagnosing common pediatric diseases, analyzing lymph node slides in pathology, detecting eye disease early on and assessing mental health.
The Moral Imperative
This process has been underway for a few years now, and while there’s a lot of good debate on specific studies, a few things are clear just from a current literature review:
• AI technology gets more accurate every day. Human experts do not.
• AI will be able to explain its results clearly.
• AI may be the only way to provide access to best-in-class health care to more than 6 billion people who make less than $32 per day.
The third point is a moral imperative. While U.S. radiologists will make the claim that AI systems cannot possibly replace them (and their $371,000 average base salary), the majority of people in the world don’t have, and will never have, access to a radiologist. There are 25 MRI machines per 1 million people in the northern hemisphere, compared to 1 MRI machine per 25 million people in sub-Saharan Africa. With a population that’s growing faster than the growth in training doctors, automated solutions are a pragmatic necessity.
So the next century of progress in medicine may be driven by AI, deployed to give your local primary care doctor, as well as the world’s top specialists, superpowers to diagnose and treat any disease. But first-world specialists are also not exempt from change. Consider the financial pressures on healthcare providers and how in the past decade, they’ve taken the exact same path that banks took in the 1980s and that retail and manufacturing (by deploying ERP systems) took in the 1990s:
• Move everything from paper to digital systems.
• Design the first user interfaces to replicate the paper workflows they replace.
• Realize that this works poorly and reduces productivity.
Knowing that the healthcare industry is 20 years behind in adopting technology, here is what happens next. With some luck, five decades from now, I’m still around, but I'll be an old man with chronic conditions. One day, I roll into a hospital needing to see an oncologist, cardiologist, endocrinologist, and psychiatrist (for good measure). I meet just one person who wears a white coat, gives a reassuring smile, and has just the right amount of gray hair to instill confidence. Various gizmos and screens are applied to adapt my treatment plan in a holistic, evidence-driven way that applies to cutting-edge medicine and matches my personal lifestyle goals. But I would not be surprised if that person’s total medical education was a 12-week entry-level training program.
Thanks for reading AI in Healthcare! Subscribe for free to receive new posts and support my work.