The World is Begging for Better Healthcare: AI May Be the Solution
- Feb 9
- 2 min read
By Asha Zimmerman, MD
Absolutely, I can help you write an article on this topic. You have probably seen sentences like that accidentally left in news articles recently. You may have even thought it was accidentally left in this article. AI is good at pattern recognition. So good that people trust it enough to forget to check the work.
Machine learning and LLMs are designed for pattern recognition. They excel in writing, math and coding. Do you know what other profession is heavily based on pattern recognition? Healthcare. The public has already realized this; it is estimated that 20% of all LLM queries are now health-related.
As doctors, we spend years learning patterns: how to interview patients, create a differential diagnosis (or list of possibilities), memorize workups and design treatment plans. That is what medical training is. Learning how knowledge is connected. This is exactly what LLMs are designed to do.
Consider that every update improves the accuracy of these models. Add the fact that the models are scalable, can work without sleep and are cheap. I mean really cheap. Like less than 1 cent to process the amount of information of a typical doctor's visit.
Combine all of these factors and we may just have a solution for the global health access crisis: an AI Doctor.
In healthcare, value is often defined by an equation: Value = Quality / Cost. Currently, LLMs can outperform physicians on multiple-choice tests, though results using real patient data are mixed. Some show LLMs doing better, others worse. However, even if we assume the quality is a “draw” (and assume they never get better), the massive cost savings means AI offers a much higher value of care. There is still a lot of work to do, but hopefully the potential can be recognized.
It is not all sunshine and rainbows. These models can make mistakes. In healthcare, mistakes can be life-threatening. We must test and deploy them with care. However, we must balance this risk against the alternative. Many people in the US and globally have no reliable access to doctors. For them, the alternative to AI isn’t a human doctor – it is nothing at all. If we don’t provide AI specifically for healthcare, patients will turn to general models like ChatGPT or Google anyways. Also, let us not forget that doctors can and do make mistakes.
In medicine, we often judge acts of commission (doing something wrong) more negatively than acts of omission (failure to do something helpful).
Even if both actions lead to the same bad outcome. But failing to implement a technology that can save lives is a mistake, too. Acts of omission still costs lives.
If the goal is improving global health, this shouldn’t be a competition of Doctors vs Computers. We should use every tool available. That means using both doctors and AI. The roll-out will likely mirror self-driving cars: first with humans in the loop, then fully autonomous. The future is hard to predict, but if self-driving cars are analogous, people are currently choosing to wait longer for a driverless Waymo than a human-driven Uber.
Connect With Asha
Comments