- Share - [](https://www.facebook.com/sharer.php?u=https://about.ebsco.com/markdownify/node/163628) - []() - []() - []() # X + 3 = Y Worry? What would healthcare look like if we removed the human-in-the-loop from AI? Dr. Z proposes a series of thought experiments in his latest article. ![From the desk of Dr. Z. Image of clinician showing information on a tablet to a patient](https://assets.ebsco.com/asset/f5abcec3-cead-4f26-bbf0-4bec7af55128/from-the-desk-of-dr-z-october-2026-blog-780.png) [Roy Ziegelstein, MD, MACP](/health-notes/bio/roy-ziegelstein-md-macp) 28 October 2026 On October 15, 2026, the journal *Science* published a paper on [AI in mathematics](https://www.science.org/content/article/openai-just-changed-math-forever-exactly-how-no-one-yet-knows?utm_campaign=Science&utm_source=facebook&utm_medium=ownedSocial) that noted that following the release of AI-generated solutions to nearly 400 math problems, mathematicians are wondering, “Has their field been decimated or invigorated?” The article made me think of a speaker on AI in health care I recently heard say, “A clinician would have to be crazy today not to use an AI scribe.” The statement was provocative to be sure, although you wouldn’t have known it by the audience’s reaction. No one challenged the speaker. No one pointed out that the evidence that AI scribes are more accurate or save time is currently quite limited. And certainly no one admitted that they might be one of the crazy ones. My thoughts went in a different direction. I wondered to myself… if AI scribes are so great, do we even need a clinician in the room at all? Try that thought experiment for a moment. I did. I quickly recognized all the things clinicians do that AI scribes cannot. Most notably, clinicians can appreciate the subtle details in facial expression and body language, as well as the interactions between the patient and their partner or the child and their parent. To say nothing about the physical examination. In that same lecture, the speaker claimed that now that specialized clinical AI tools have largely eliminated hallucinations, as long as the AI relied on *credible source selection* (that is, selection of articles primarily from highly rated medical journals), “there really is no need for a human in the loop anymore.” The statement was provocative to be sure, although you wouldn’t have known it by the audience’s reaction. No one challenged the speaker. No one pointed out that the results of peer-reviewed studies – even randomized controlled trials published in leading medical journals – sometimes [can't be reproduced](https://jamanetwork.com/journals/jama/fullarticle/1902230) or that [subsequent trials report outcomes much less impressive than the earlier trial](https://jamanetwork.com/journals/jama/fullarticle/201218). No one noted that sometimes publications need to be retracted because of critical flaws. And certainly no one admitted that when served as a peer reviewer, they often reviewed the paper so quickly and superficially because of competing demands that they almost certainly missed important errors. My thoughts went in a different direction. I wondered to myself… if AI is so great, why does DynaMed need such incredibly talented, dedicated professionals? Why do we need terrific writers and editors? Why do we need clinical experts and experts in evidence-based methodology? Why do we need a human in the loop at all? Try that thought experiment for a moment. I did. I quickly recognized all the things we do that AI alone cannot: - Carefully review journal publications and guidelines and evaluate the strength of medical evidence. - Compare and contrast guidelines when recommendations from different specialty societies might not be in agreement. - Describe the benefits and harms of diagnostic procedures, medications, and interventions and indicate their potential relevance to patients. - Detail which trial outcomes are clinically important and which are less so. That paper in [*Science* on AI in mathematics](https://www.science.org/content/article/openai-just-changed-math-forever-exactly-how-no-one-yet-knows?utm_campaign=Science&utm_source=facebook&utm_medium=ownedSocial) also noted that when OpenAI released 722 preprints that included solutions to 372 advanced math problems, many mathematicians were despondent. Some wondered about the future of mathematics as a field and whether they now have a job… or even a purpose in life. You may be wondering that yourself. Before you worry too much, remember all the value you bring to your work, and all the things you do that AI cannot. And think about this sentence from the *Science* paper and read it to yourself a few times: “As *Science* went to press, [OpenAI said it had withdrawn three of the manuscripts and revised 14](https://retractionwatch.com/2026/10/08/openai-withdraws-preprints-722-manuscripts-unsolved-math-problems/).” Category: [From Our EIC ](/blogs/health-notes/category/from-our-eic) Tags: [AI](/blogs/health-notes/tag/ai) [From the Desk of Dr. Z](/blogs/health-notes/tag/from-the-desk-of-dr.-z) - Share - [](https://www.facebook.com/sharer.php?u=https://about.ebsco.com/markdownify/node/163628) - []() - []() - []() ## Related Posts [ ![From the desk of Dr. Z. Editor in Chief, Dynamed. Medical doctor pondering research in medicine at their laptop](https://assets.ebsco.com/asset/b969c128-772a-499a-be36-63ec91df4479/from-the-desk-of-dr-z-black-and-white-medicine-september-2026-blog-780.jpg) ](/blogs/health-notes/dont-squint) [Don't Squint](/blogs/health-notes/dont-squint) 17 September 2026 [ ![](https://assets.ebsco.com/asset/2214538d-ea74-4d58-afe9-4b507979b069/From-the-Desk-of-Dr-Z-doctors-discussing-August-2026.png) ](/blogs/health-notes/flight-2482-and-value-dynamed) [Flight 2482 and the Value of DynaMed](/blogs/health-notes/flight-2482-and-value-dynamed) 27 August 2026 [ ![From the desk of Dr Z](https://assets.ebsco.com/asset/2e72c72c-03af-40af-8cf4-3d729749d4af/From-the-Desk-of-Dr-Z-March-2026-780.png "From the desk of Dr Z") ](/blogs/health-notes/not-just-any-tom-dick-or-stanislav) [Not Just Any Tom, Dick, or Stanislav](/blogs/health-notes/not-just-any-tom-dick-or-stanislav) 11 March 2026