The Dumb Side of Artificial Intelligence and its Impact in Health
Quick question: If you were really sick how would you feel about having a doctor that is brilliant at the “science” of medicine but unable to explain why they are recommending a treatment, clueless about how to listen and understand what you are feeling and not adept at managing the “softer side” of care delivery?
Such is the case with Artificial Intelligence (AI) today.
AI holds great promise to improve the quality and effectiveness of many aspects of health services delivered around the world today. A key factor in harnessing its power is to understand what AI is good at and frankly what it’s not so good at.
AI is increasingly good at things like being able to sense and predict things we care about like which patients are at high risk of readmissions, falls, or unexpected deterioration. It can help with predictions of which treatments may produce the best outcomes. It’s already making diagnostic images more “intelligent”. These are a few early examples that demonstrates the power and potential of AI to fundamentally change how health and medical conditions can be managed differently.
This point was brought home recently by Dr. Mobasher Butt from digital healthcare startup Babylon Health when he played to a packed house at London’s Royal College of Physicians. He was there to promote the capabilities of the company’s carefully trained “AI Doctor.”
A key measure of progress noted in Dr. Butt’s presentation was that the “AI Doctor” had scored higher than the average of human medical students on the MRCGP exam (an assessment system that determines whether a general practitioner is competent to enter independent practice in the UK.)
Clearly laudable, this work points to the potential for AI to improve the way we support the health needs of those we serve.
And yet, as we gush over the growing prowess of AI, it’s important to recognize and bridge around its limitation when applied to health and medicine.
Another quick question: While everyone’s talking and writing about Artificial Intelligence, have you ever heard anyone talk about Artificial Wisdom?
As “smart” as AI is at certain things, no one has figured out how to imbue or mimic those qualities that are essential to the care process. These include wisdom, reasoning, judgement, imagination, critical thinking, common sense and empathy. Such attributes remain as uniquely human characteristics.
Being able to sense or predict temperature variation is not the same as knowing what’s going on in a patient when they have a fever. Measuring spikes in blood pressure does not equate to understanding what anxiety feels like for a patient or family member and what to do about it.
As Artificial Intelligence gets better let’s recognize that there is no such thing as Artificial Wisdom.
In my hometown of Seattle, Washington, Paul Allen (the less-recognized co-founder of Microsoft) is ponying up an additional $125 million for the Allen Institute for Artificial Intelligence to attempt to shed light on whether computers can be “taught” how to have or replicate what we casually refer to as “common sense.”
We should embrace and adopt increasingly sophisticated ways in which we are able to create algorithmic-driven predictions. At the same time, any clinician or researcher who has spent time in real world situations is quite familiar with the "correlation versus causation" discussion.
One of my favorite books and websites on this topic is from Tyler Vigen called Spurious Correlations. Vigen lampoons the treatise that correlations on their own can be used to infer causation. For example, did you know that there is an almost perfect correlation (.995) between iPhone sales and the number of people who die each year by falling down stairs? Does this mean we can reduce the mortality rate from falls by reducing the number of iPhones sold?
As Artificial Intelligence gets better let’s recognize that there is no such thing as Artificial Wisdom.
Let's recognize the growing capabilities of AI to improve care while embracing the fact that wisdom and common sense remain uniquely human and essential to “systems of care” that incorporate the totality of physical, emotional and environmental factors in furthering the health and well-being of those we serve.
An additional read: Leading health and medical organizations are beginning to evaluate and frame approaches to how best to apply AI to improve the delivery of health and medical services. Last month the American Medical Association published it's first position paper on the opportunities and issues surrounding the incorporation of "Augmented Intelligence" in the provision of clinical care. A good read for any clinician interested in this topic.
Tom Lowry is Director, Worldwide Health, Microsoft.
Lead global business development to increase the use of cloud-based analytics and AI solutions by health clients and partners. Advise and support regional teams in developing and executing cultivation and sales activities that increase the innovative use of health analytic solutions in the provider, payer and pharma markets while increasing Microsoft’s global footprint in the Intelligent Cloud market. Advise and influence product strategy and development amongst Microsoft partners to accelerate the penetration of analytics and AI solutions in area such as Electronic Medical Records (EMRs), medical devices, wearables, medical imaging, and population health. Fluent in optimizing performance through cognitive computing, artificial intelligence, big data, machine learning and cloud computing.
