There’s no question that it can take time to see a doctor in the US, whether you have insurance or not, and regardless of which insurance you have. Is the solution for speeding up access to promote AI to diagnose and treat patients, notwithstanding the risks? The Trump administration believes it is, reports Christina Jewett for The New York Times.
Unfortunately, it is not at all clear that AI diagnoses will be accurate, safe or effective. Many experts believe they could be dangerous; the harm outweighs the benefits. Others dsagree.
One billionaire, Vinod Khosla, sees a future in which we no longer need human physicians because AI can do their job. His son has an AI company aimed at meeting that objective. Even if it could and the evidence is not yet compelling, don’t for a minute think that AI would bring down health care costs, even though it could.
Everyone invested in AI expects a big payout. What you get for free now through AI could cost a bundle once there’s greater reliance on it.
To be clear, AI could have a valuable role to play in the health care sector. In many cases, you wouldn’t need to leave your house to get a diagnosis for a condition. And, who’s to say the AI diagnosing you wouldn’t be more compassionate and accurate about your condition than the primary care physician you would have had to wait to see?
But, right now, even leaders in the AI space are calling for an AI slowdown. It’s one thing to use AI to help with paperwork, emails, and billing, it’s another to use it to care for a patient. It goes without saying that the American Medical Association opposes the use of AI for the diagnosis and treatment of patients.
Before we rely on AI to diagnose and treat conditions, we should have a good sense that it is doing right by patients. Right now, we know that insurers in Medicare Advantage are using AI to deny care inappropriately. Where’s the evidence that AI will treat patients appropriately and not kill them?
Still, Medicare is preparing to pay for AI. Officials state that AI will reduce costs and improve care. Already, more than 1,500 medical devices rely on AI, for a wide range of uses, including prescribing drugs.
Right now, the Department of Health and Human Services is allowing more than 200 companies to run pilot programs with AI. One pilot provides AI talk therapy for people with depression or anxiety.
AI is particularly challenging to regulate because it is always evolving. Information it offers, advice it gives, treatments it provides, can and does change all the time. The government is hard-pressed to protect Americans, even if it wants to, from AIs that do the wrong thing. The risks are real. In one case, an AI is alleged to have provided medical advice to a man that endangered his life.
The administration’s appointee to a new post that would oversee AI is a tech expert who advocates for limiting oversight. Whether the views of independent expert scientists and consideration of people’s safety will be factored into the oversight is an open question.
The administration is considering paying tech companies for AI physicians as much as 60 to 80 percent of what physicians earn. It’s easy to imagine the AI physician bots going wild in their billing. There could be little if any physician oversight. Why rush?
