See this. I use OpenEvidence. It has access to full text from some of the major medical journals. But generalist models seem to outperform it. Not sure what is going on there.
Physician reimbursement is only ~9% of national healthcare expenditures.
I tell you this with certainty as a 3rd year medical student: If physician wages go down and tuition stays as is, no one will do this. Intrinsic motivation to help people evaporates as soon as you see how enshittified healthcare in the US has become.
I do agree that medical school is far too restrictive to get into (For MD schools at least). However, if you want to make medical school easier to get into: Where will all those students rotate at for their clinical years? There aren't enough spots in hospitals to jam students in.
Stop taking aim at the people that sacrifice so much to help you. Take aim at the real drivers of healthcare expenditures: administrative bloat.
>Where will all those students rotate for their clinical years? There aren’t enough hospital slots.
This is a policy fiction. Residency slots are capped by federal law, not by hospital capacity. The Balanced Budget Act of 1997 froze Medicare-funded residency positions, and despite modest expansions decades later, the cap remains largely intact. Teaching hospitals routinely report excess clinical volume relative to trainee supply. The bottleneck is artificial and regulatory, not logistical.
>Stop taking aim at people who sacrifice so much to help you. The real cost driver is administrative bloat.
This framing collapses under scrutiny. Administrative bloat is real and well-documented, but pretending physician incentives are irrelevant requires willful blindness. Numerous studies show that U.S. physicians earn multiples of their OECD peers while delivering no commensurate advantage in outcomes. Many doctors are motivated by altruism, but many are also motivated by status, income, and professional gatekeeping—normal human incentives in a high-prestige, high-pay profession.
Further, high patient throughput is not an accident. Fee-for-service reimbursement structurally rewards volume over care quality. Seeing 20–30 patients a day is not a moral failure of individual doctors, but it does predictably lead to burnout, emotional detachment, and assembly-line medicine. Incentives shape behavior. Ignoring that is not compassion, it’s denial.
>Physician reimbursement is only ~9% of national healthcare spending.
That statistic is repeatedly used as a rhetorical shield, and it shouldn’t be. Cost systems do not fail because of a single oversized line item; they fail because multiple protected constituencies simultaneously extract rents while deflecting blame. Administrative overhead, defensive medicine, pharmaceutical pricing, hospital consolidation, reimbursement incentives, and physician compensation are jointly optimized for revenue, not outcomes.
Nine percent of a multi-trillion-dollar system is not trivial. More importantly, physician compensation is not isolated—it drives downstream costs through referral patterns, test ordering, procedure rates, and resistance to scope-of-practice reform. Treating physicians as a sacred class exempt from economic critique is precisely how you end up with a system that is unaffordable, unaccountable, and structurally resistant to reform.
If the argument is “9% is too small to question,” then by that logic no component is ever large enough to examine in isolation, which is how dysfunctional systems persist indefinitely. Real reform requires abandoning moralized narratives and admitting the obvious: healthcare costs are the product of aligned incentives across many actors, and physicians are not magically outside that system simply because the story is uncomfortable.
We are yet to see major, nationwide physician strikes. If that is what it will take for society to realize the value provided, so be it. Without physicians, there is very little healing going on. You can't say the same for so many other roles.
This is why I look up to physicians. The fact that you'll hold all of our lives hostage with the words, "Don't fuck with my salary, I'm valuable." That commands respect. More Darth Vader respect than Mother Theresa respect.
It's really just not there yet. I've been in medical school for >3 years now and have been using the latest models with good prompting. They have gotten much better, but I still see misses that my classmates would easily catch. This is not acceptable in healthcare. It's certainly not getting 100% on all my assignments, which are a step below the complexity of real-world clinical practice.
Before medical school, I was not so sure of the quality of your average doc. Now having spent a year in clinical practice across various settings, I am extremely reassured. I can say with certainty that a US trained doctor is miles ahead of AI right now. The system sucks really bad though and forces physicians to churn patients, giving the impression that physicians don't pay attention/don't care/etc.
Salaries for healthcare workers make up only a small portion of expenditures. You do not want to avoid a trip to your GP for an AI system.
It'll be great at first while in development. But when profits need to be generated, seeing a specialist will get harder. There will be less wiggle room. I predict we will see more GP utilization.
We would still have this problem. The heavier models make mistakes at too high a rate vs. a physician. Especially on imaging data. Real world data and patient presentations often deviate from the textbooks they are trained on.
That's a different class of problem. It will do just fine on text based queries spanning a few pages. Probably better than the average physician (average over all countries).
I do agree that LLM's are not there yet in the image part.
See this. I use OpenEvidence. It has access to full text from some of the major medical journals. But generalist models seem to outperform it. Not sure what is going on there.
reply