views
Will Artificial Intelligence replace physicians?
Here is our CEO, Deepak Mittal’s opinion about AI replacing physicians:
The same opinion others leaders have expressed in a poll conducted by NGI. Results were quite clear, 63% of the respondents believe that AI won’t replace physicians:
Artificial Intelligence (AI) acts as an enabler to medical care. AI/ML shines the most when it is assisting physicians in making better medical decisions. More than accuracy, we as humans need human empathy from a physician along with effective treatment.
Al/ML is bringing the power of object identification, classification along with question/answering, but a physician’s power lies in linking various pieces of information to make decision. Diagnosing a condition is an np-complete problem (specifically set cover: http://en.wikipedia.org/wiki/Set_cover_problem) and even with quantum computers, np-complete problems cannot be solved in polynomial time. The point can be well proven by the experiment of Waldo.
Waldo experiment: Waldo wears a stocking cap, even in the summer, is skinny, usually wears a striped shirt, needs a haircut, and always hangs out with lots of other characters. Can you find him in the image? And if you did, how long did it take?
The viral video where a robot built with Google AI finds Waldo from a cluster of images within seconds which a human eye would normally take minutes. But if we change the question to who all needs a haircut in the picture then will the computer do that. That’s where in our opinion, a physician’s power lies. Or better question will be which haircut will look best on which person based on his liking/ethnicity etc. Please note that we are only highlighting two of the thousand parameters that might be going in a physician’s mind. Unfortunately, not all decisions are black and white in our human world.
Even if we consider fully automated surgeries, we have history books to offer wisdom. The advent and progress in AI has been remarkable, and we have had our fair share of lessons from shortcomings and mistakes like Therac-25 in the past, where admittedly so we’ve realized there is a long, long way to go before AI can even dream of replacing surgeon.
I would also like to point to the problem with malpractice insurance and FDA approval. Think of a hospital having 1000 physicians and malpractice points to one physician vs malpractice pointing to AI that does the work of 1000 physicians. Imagine a situation where If we may run out of physicians overnight because of one blunder or mistake.
In nutshell, we believe that it is unlikely we’ll get humanoid “robot physicians” for a long time to come, though technology will replace some of the more routine aspects of medical care, or improve it but for now, the physicians are here to stay.