Sam Finlayson

Meta-pin👇

Semi-related: For me personally, I can't imagine a worse clinical experience than receiving hard diagnostic news from a ChatGPT persona via text. Humans all the way for me for that sort of empathy-intensive experience. (Followed eventually by lots of chatting with ever-patient and thorough LLMs to fill in gaps).

@jonc101x @Gabe__MD I think a lot of the early findings on increased empathy were largely artifacts of the evaluation methods. That's not to say that patients CAN'T find such systems empathetic (as the doctor in the Times piece does), but I don't think the evidence as strong as we say.

For years, the consensus among senior clinicians and AI specialists was clear: AI cannot provide empathy. It cannot replicate the human touch. It lacks patience. It cannot build therapeutic rapport. These were presented as permanent limitations, the floor beneath which AI could never reach. A Columbia emergency physician and associate professor just published an essay in the New York Times describing her experience using ChatGPT for her own health concerns. What she found: The AI asked thorough follow-up questions about her daily life. It tailored recommendations to what she could realistically change. It suggested a post-meal walk and explained why a longer workout would offer only marginal additional benefit. When she told it some suggestions were unacceptable, it adapted without judgment. When she asked the same question repeatedly, it never showed irritation. It provided sustained encouragement over weeks. She followed its advice. Her labs improved. She wrote: "It kept cheering me on, precisely the kind of steady, relational care we keep insisting only humans can provide." She described a patient with a highly curable cancer who asked a chatbot every week if his cancer could be cured. He already knew the answer. He needed the reassurance. The chatbot provided it without fatigue, without impatience, without making him feel like a burden. The physician who wrote this piece is not a technologist. She is a practicing ER doctor at a major academic medical center. She found that ChatGPT delivered more personalized, more patient, and more empathetic care than the healthcare system provided to her as a patient. And it changed how she practices. Every objection senior clinicians have raised about AI lacking the human side of medicine is being contradicted by the lived experience of physicians using these tools on themselves and watching their patients use them. The empathy objection. The patience objection. The rapport objection. The "AI cannot listen" objection. Each one addressed not by a benchmark but by a physician describing what actually happened when she used it. The people who said AI could never do this were not making an empirical claim. They were making an assumption based on what AI looked like years ago. They stopped testing the assumption. The technology did not stop improving. The irony is hard to overstate. The human touch that physicians claimed as their irreplaceable advantage is now being delivered more consistently by a machine than by the system those physicians work in. nytimes.com/2026/05/24/opi…

I've soured a bit on podcasts over the last few years but will probably buck that trend for this one as I've never heard a Michael I. Jordan take I didn't like.

hallucinated references will land you a 1-year ban from arxiv now. wow

@noahkaufmanmd I respect your right to your hedgy and IMO misguided prediction. x.com/IAmSamFin/stat…

@IAmSamFin Ha! No real conviction… I just think in 10 years time robots will be as good or better than most surgeons. Plus some surgeons are total artists and having incredible skill and that’s a different story.

Not a gambling man, but I would strongly consider a formal bet on this one. (My position would be against the claim “surgery will be replaced by AI within 10 years”)

To elaborate: I disagreed with Hinton’s infamous 2016 prediction at the time, but I believe the bull case for radiology’s obsolescence based on CV trends in 2016 was stronger than the bull case for surgery’s obsolescence based on robotics in 2026 and I don’t think it’s close.

@noahkaufmanmd Come on man, you called by position nuts! Where’s the conviction? :)

@IAmSamFin Yeah I wouldn’t bet either but I do think it will happen at some point.

when the patient getting awakened q1hr for days finally develops delirium:

AI news cycle in a nutshell

Nuanced TLDR explainer on the import of the recent @ScienceMagazine article on LLM performance vs doctors.

@alan_karthi @GoogleDeepMind Thanks for sharing this, and congrats on the release! Hope to see you all soon at SAIL -- lots to catch up on.

Since joining @GoogleDeepMind I’ve dreamt for a decade that AI can give clinicians superpowers: amplifying reach as an always-available, trustworthy member of the care team. Delighted to share our strategic research initiative at @GoogleDeepMind towards this vision: AI co-clinician deepmind.google/blog/ai-co-cli… (1/n)

This is the the quote I've been citing a lot recently.

@justalexoki @atlanticesque Depends quite a bit on one's definition of fake! In many important senses it's as real as anything in history.

@IAmSamFin @atlanticesque that shit is so fake its not even funny lmao

did Jesus just lowkey crash out or what happened here

@LilianDelaveau @atlanticesque @justalexoki x.com/IAmSamFin/stat…

@atlanticesque @justalexoki I would be less atheist if there was more joke and relatable Jesus Content in the Bible tbh

@justalexoki @atlanticesque You guys would love the apocryphal second century Infancy Gospel of Thomas en.wikipedia.org/wiki/Infancy_G…

@atlanticesque deleting a fig tree as a prank

You've got to remember that these are just simple jurors...The common clay of the new AI spring.

@venturetwins Another Google hand in one thread with above prompt: Software engineer again Repeating same palm in same thread: Teacher Chef Mechanic Photographer

@venturetwins I did your hand, my hand, and the first google hit for hand and they all said software engineer or data scientist.

I want to not believe in palm reading… But ChatGPT guessed I was an investor or founder (and nailed a bunch of other stuff). Literally just upload a photo of your hand and ask for a palm reading - add “graphic” if you want a diagram like ⬇️

@LostFederalist @jamesonhaslam Genius, but the evil daemon defines it as weight in hand, turning into body weight pullup

@jamesonhaslam Some genius in the replies said assisted pull-up machine I literally can’t think of anything better

gotta be leg press, right?