Community oncologist. I use AI to appraise clinical trials, then verify every finding against the primary data. The AI gets graded too. One trial, two readers, one bottom line. The Source Report on YouTube (in bio) and Substack. allenlimd.substack.com #onced #oncology #meded

We trained doctors to behave like machines, and patients went looking for a machine that behaves like a doctor. @DrArturoAI @EricTopol @Timothee_MD @FCademartiri @VPrasadMDMPH @oncodaily #AIinMedicine #EvidenceBasedMedicine

Everyone remembers the drama. @NatureMedicine paper said general LLM beats medical AI. @openevidence fired back with its own study showing the reverse. The whole field is arguing over who won, and they're all missing the same thing. Here's what nobody in this fight is measuring, and why the winner matters less than you think. @DrArturoAI @EricTopol @Timothee_MD @FCademartiri @VPrasadMDMPH @RoupenMD @HealthcareAIGuy @MattZirwas @Jean_J_Feng

AI can reproduce a treatment algorithm and cite the guidelines, but it cannot tell you when to deviate from them. In medicine, that distinction is the entire job. Today’s medical training raises a real paradox. Practicing clinicians know when to deviate because we built that judgment practicing without AI. If today's trainee works with AI from day one, where does that judgment come from? That is the risk of de-skilling and never-skilling. Could AI one day get good enough to truly be in the exam room and perceive the nuances we do? Maybe, even likely. Until then, judgment remains the most important clinical skill. The art of medicine. In AI terms, "taste". I don't have a neat answer. But this is the one thing medical education and practicing clinicians should actively hold on to.

"That was an older model" is how we let medical AI dodge every inconvenient finding. Good result: the tool works. Bad result: that version is obsolete. Nothing can count against it, which means it isn't being tested. Even when it's Microsoft's own team, with @EricTopol, finding the flaws. @VincentRK @DrArturoAI @Timothee_MD @FCademartiri @VPrasadMDMPH @RoupenMD @YLeyfman @hoifungpoon @YuGu_Aiden @HealthcareAIGuy #AIinMedicine #EvidenceBasedMedicine

AI Benchmarks are in vitro. That is where good medical AI starts, not where it is proven. Everywhere else in medicine, we already know the difference. @VincentRK @DrArturoAI @EricTopol @Timothee_MD @FCademartiri @VPrasadMDMPH @RoupenMD @YLeyfman @HealthcareAIGuy #AIinMedicine #EvidenceBasedMedicine

You would not use a drug because it worked in mice. So why is a benchmark from a lab treated as proof that a medical AI helps patients? It only shows a model improved. The clinic shows whether patients did. Article: A Benchmark for Medical AI Is Only Preclinical Data → open.substack.com/pub/allenl…