Self-driving cars, humanless production lines, robot doctors? Artificial intelligence is already disrupting global economics systems at an unprecedented rate, replacing millions of jobs in manufacturing, finance, and even law. Such sweeping systemic changes to the future labour markets beg the question: is healthcare next? To answer this, we first have to grasp what artificial intelligence is and how it works.
AI: Apples and Intelligence
AI is short for artificial intelligence. In actuality, it is a broad umbrella term used to describe the capacity for machines and computers to mimic the human ability to think and reason as a means of problem-solving.
Within AI is machine learning, which refers to the use of algorithms that “learn” to improve at a specific task without being explicitly programmed. For instance, if you were to design a program that could identify images of apples out of an assortment of images featuring 10 different types of fruits, you could train the program to learn which fruits were apples by manually telling it whether each image it identified was the correct output. Eventually, through “practice” the program would be able to identify which fruits were apples.
Within ML is “Deep Learning”. Deep Learning refers to the ability of a program to use a prebuilt neural structure to classify inputs automatically. To use the same example as before, instead of a human training a computer to identify apples, a computer with a preprogrammed architecture could extract out the defining features present on the various types of fruit and then classify all the fruits into different categories without a human providing continuous feedback. It is this very technology that is now putting millions of jobs worldwide at risk of automation; the replacement of human labour with robots, causing unemployment.
Who Gets Replaced and Why?
Since the basis of AI is pattern recognition, jobs that are repetitive and require little formal education are at the highest risk. For example, if you’ve ever been to McDonald’s, you may have used a self-serve kiosk, where you place your order simply via the touchscreen with minimal human to human contact. Been to a supermarket in the last 5 years? Self serve checkouts split the work of entry-level cashiers. Pick up your smartphone! That junk call you received last Friday selling you car insurance probably wasn’t from a real person.
However, it isn’t just part-time summer jobs for high school students that are being replaced. Due to major breakthroughs in machine learning and specifically deep learning, “safe” and cognitively taxing jobs are now also at risk.
It is estimated that 25% of legal tasks such as proofreading contracts and legal research can be done with greater efficacy by algorithms.
In finance, insurance underwriting is losing its human bias as AI learns to provide more objective recommendations. Since analysts of all stripes use previous data to predict future trends, it too is becoming home territory for artificial intelligence.
As a matter of fact, in Canada alone, The Brookfield Institute for Innovation + Entrepreneurship at Toronto’s Ryerson University predicts that 42% of Canadian jobs are at high risk of automation within the next two decades. This begs the question…
If intellectually intensive work can also be automated, is healthcare next on the chopping block?
Well, yes… and no. In the same way that the field of business consists of accountants, analysts, administrators, bankers, and marketers who all possess different skills and abilities, healthcare cannot be likened to simply pediatricians and registered nurses. The field of medicine is highly diversified in the sense there are a significant magnitude of jobs that don’t require directly prescribing controlled substances. Thus, it is important to draw such distinctions when discussing which jobs are at medium, low and high risk.
A General Overview: Replacement or Displacement?
Generally speaking, the field of medicine is at a lower threat of automation in comparison to sectors of the economy that require less formal education. While there is a plethora of expert opinions regarding the future of AI and medicine, the general consensus is that AI will displace rather than replace work in healthcare. Furthermore, the field of healthcare is a uniquely human enterprise. There is a unique personal connection that comes with helping people at vulnerable times in their lives. Empathy, reassurance, humor, and general rapport simply cannot be replicated by abiotic entities such as machines.
With diagnoses concerning mental health increasing at an alarming rate, there will still be demand for human counselling no matter how accurately a machine can calculate one’s required dosage for antidepressants because doctors treat patients, not symptoms.
There are also distinct patient care scenarios where human involvement is a necessity. Consider death counselling, terminal illnesses, drug-seeking behaviours, patients who are convinced that they need antibiotics for viral infections or unnecessary screenings, and those who refuse to take their medications or do not do so properly.
In addition, with an ageing population, the WHO predicts that there will still be significant shortages of healthcare professionals even if 10% of tasks in the field can be automated.
That being said, not all jobs in healthcare are safe. Here are healthcare jobs that are at high, medium or low risk of being automated.
High Risk – If it’s Repetitive it Ain’t Competitive!
Not all healthcare jobs will be safe. The first to go will be those with little to no direct patient contact. These include…
- Medical transcriptionists
- Medical records and health information technicians
- Medical secretaries and billing
- Medical equipment preparers
- Low-level medical assistants
- Hospital janitorial services
- Medical triage
As is evident, these jobs empathize repetition with minimal variation, making it prime for near-full automation by AI algorithms. The amount of high-level patient-to-provider contact is also minimal, as any contact between patients and those working in such professions revolves around non-specific knowledge that lacks individuality.
Medium Risk – More Displacement Less Replacement
These jobs will likely see shifts in the nature of work rather than full-scale automation of core tasks. What may be surprising is that many high salaries, specifics, and “safe” specialties are included in this category.
Remember the example of an AI algorithm classifying different images of fruits? To some extent, that’s what radiologists, dermatologists, and neurologists do. Radiologists are trained to read medical imaging technologies such as MRIs, CT Scans, and X-rays to detect the presence of disease symptoms. Likewise, dermatologists determine whether skin abnormalities are benign or cancerous. Such work is highly pattern intensive which is one of AI’s greatest strengths. So much so that in some cases AI is currently better at diagnosing breast cancer than radiologists by the sheer ability to reference a greater database of knowledge. However, radiologists still prescribe appropriate treatment plans and counsel patients; a task AI simply cannot do effectively.
Such can also be said about epidemiology and pathology, where AI predictive models are already being used to predict and monitor disease outbreaks and disease progression respectively.
Furthermore, though some menial tasks done by pharmacists can be automated, humans will still be required to ensure patients are compliant and consistent with their medications.
Displacement over replacement is the name of the game.
Low Risk – But… not risk-free
To be fully transparent, “low risk” doesn’t mean “no risk”. AI will still impact these professions to some extent, however, to a lesser magnitude. Jobs in this category are highly patient-centric. Human to human interaction is maximized and repetitive administrative tasks are minimized.
- Therapists of various sorts
- Physical therapy
- Genetic counsellors
- Physicians in general
- Emergency Medical Technicians
Generally speaking, people who work in such professions treat patients more than they fill in paperwork, and must exercise compassion in order to build patient rapport. Only under those pretenses will patient outcomes be optimized.
It should also be noted that nearly all of the examples are in highly specialized fields that deal with objectively nuanced patient profiles. AI algorithms would have a difficult time convincing an elderly patient with Alzheimer’s that they need to take Simvastatin twice a day to lower their cholesterol, or educating concerned parents on the importance of vaccinating their 12-month-old, let alone revealing to a physically active 40-year old that he has a 50% risk of dying from terminal lung cancer simply because of his genetics.
All the aforementioned occupations, especially psychologists, deal not just with symptoms, but integrated and complex mental conditions while providing holistic medical care. It is exactly the high degree of variability that distinguishes “safe” and “unsafe” jobs.
Facts Over Fear
At the end of the day, foreseeing the potential impacts of AI on future jobs in the field of healthcare is by nature, speculative. Healthcare is a uniquely human enterprise, in the sense that accurate diagnosis and effective treatment is only part of the equation. Every technological revolution thus far has also created novel jobs that simply hadn’t existed before. So keep your head on a swivel, ensure that your education provides you with irreplaceable skills, and be alert, not anxious.