ICU doc testing whether AI will take my job. Daily AI vs HI. Keeping patients alive by day, kids alive by night. Rockets & sci-fi. Hoping to see Kardashev 1.
@SyntheticBeef I would actually love AI to cure diseases. Medicine always lags behind other tech innovations. It often takes years before a new molecule can be used in clinical trials. I dream of the day when AI give me customized molecules for every type of infection and disease.
Oui. Nous sommes chanceux de vivre dans une ère d’abondance. Quelqu’un de la classe moyenne en Occident vit mieux aujourd’hui que les plus nantis d’il y a 100 ans, justement parce que nous avons eu de la croissance et du progrès technologique. La majorité de ce progrès est attribuable à notre système où le capital se concentre dans les mains de ceux qui sont les plus aptes à le déployer en nouvelles innovations.
@HariSel57511397 Yes. The issue is, most artist don’t realize that their first job is to learn sales. Probably thousands of great books are written every year that nobody ever reads because the artists don’t know how to sell themselves and their work.
I'm glad to hear that UpToDate is working hard on its newest version. I still happily pay for my UTD subscription and these days, when a question comes up, for fun, I compare what OE, UTD and now VerahHealth each answer.
I think one thing that will separate the winners from the platforms that struggle is how well they adapt to the question. An attending who needs a quick answer between patients doesn't want the same model or the same level of details as when they're reviewing guidelines.
The platforms that win will be the ones that switch easily between a quick answer and a thorough explanation, with complexity that builds step by step for someone who wants to review a subject in depth.
I think that a big part of the radiologist job will be accountability. Even if AI becomes better than the doctor, people don’t entirely trust the machine and need to know that a human has the final responsibility for their case. We see the same happening with FSD cars. Even if the machine is better than the human, there’s a lot of push back against relying to much on FSD and people expect way more from the AI than from human drivers. I think radiologists will become more efficient with AI and that their job will change, but they won’t disappear.
Yes. People forget that the natural state of humanity is poverty and injustice. The only thing that took us out of this misery is growth and technological development, both of which are the direct consequence of capitalism. I can understand arguments against oligarchs and monopolies that put barriers against innovation, but anybody with some real life experience know that some people have more talent in allocating capital and building than others.
Great case and always frightening. When these kids are rushed to the OR, there's always the dilemma about if you place the IV first or start with the sevoflurane inhalational induction. On one hand, you don't want to make them cry with the IV and worsen the obstruction. On the other, I feel safer when the IV is already in place. In my practice they usually come already having their IV line or I take the time to put one before starting the induction. What's your take ?
@Chiragrx07 I actually had to look this one up. That's far from my field of expertise. B. Acquired perforating dermatosis. Thx for teaching me something today !
Vera health versus OpenEvidence
I recently had a patient in shock from a massive iatrogenic air embolism that started a discussion about how much air it takes to cause cardiovascular collapse.
I asked both Vera Health and Open Evidence. Both gave very good answers. Vera health gave me the concise version: I wanted to know one thing, and it answered it directly. OpenEvidence gave me more text, more evidence and more nuance. I liked the details, but I had specifically asked for a concise answer. If I'd been in a hurry, it would have been too much to read.
Overall, both scored well.
Yes. Thats sound advice. First time I was left alone in charge of an ICU for a night shift as a junior resident I was sacred to death. Years later this is bread and butter. I always tell my fellow that it’s ok in a reanimation to be scared. The important job is to appear in control so the team can function well.
It depends on where you are in your curriculum. Nothing beats field experience. I learn best when I’ve got an interesting case and then expand on it. Now as an attending, I still use the same technique. I’ve had a huge pulmonary embolism recently. I used the case to review the new guidelines, indication of mechanical thrombecotomy versus systemic thrombolysis etc. I always have Claude open beside my papers so I can ask questions along the way while I read.
This has always been an issue with models from GPT though. You ask for X and then it goes tangential and hours later it drifted so far away from the task that you don’t remember what the goal was. On coding tasks and for specific research I often feel like Claude is better than GPT for this reason. It’s better at applying Pareto law and putting the emphasis on what is game changing.
@_Anesthesiology Ouff. When using ultrasound to put a pigtail though, it’s not always easy to differentiate between a huge bulla and a pneumothorax so it might happen more often than we think.
@TokenGremlin The issue with AI is that it's a classic prisoner's dilemma. If we don't push for growth, other nations will. If there's one thing I want even less than a dangerous AI, it's a dangerous AI owned by someone hostile to us.