I prefer "confabulation," which describes the analogous human behavior where you have no idea what the objective truth actually is, so you just make up something that sounds right
That's exactly what happened to me when I tried to get the open-source models to extract a CSV from textual data with a lot of yes/no fields; i.e., the model forgot the column order and started confusing the cell values. I found I had to use more powerful models like Mistral Large or ChatGPT. So I think that is a valid thing to worry about with smaller models, but maybe less of a concern with larger ones.
True, this is how RAG works, but this is why I prefer to use open-source LLMs for RAG: because the token costs are less opaque and I can control how many chunks I pull fromthe database to manage my costs
It's based on understanding the (theorized) disease process and trying to prevent it. If the problem is that you're looking at nearby stuff too much (e.g. phone screens held less than a foot away from the eyes), then that means your eye muscles are contracting too much. Atropine paralyzed the eye muscles and forces them to look further away because nearby stuff would be blurry.
Technically, copyright promotes creativity and patents promote technological progress. I think patents block progress more than they help it, especially with patent trolls and all, but copyright is probably a net positive for society.
> Until recently, the algorithm that was protecting all organ donor patient information in the country, so STI status, mental health, every physical history, was from 1996.
I take issue with the way this point is presented. I'm willing to believe that healthcare data wasn't encrypted with the SOTA methods and one that was regarded as "good enough" instead, but the age of the algorithm has nothing to do with it's security. After all, RSA encryption is from 1977.
This is getting down voted, but I hope people realize that what he's saying is that the Venn diagram of crypto grifters and AI grifters is, in fact, a circle
Practical AI did a podcast episode about the dangers of using AI models as a shield to hide behind in justifying your decisions. The episode was titled "Suspicion Machines" and based on the libked article [1], and I think it's worth a read/listen.
The problem is that it could work that way, or it might not. It depends on what tax rate the government sets, which could change depending on the time, the economy, and the current administration. If the government guesses the magic rate that discourages speculation without discouraging other relatively low-value businesses that society still needs to function, then yes, that's a great system. However, I'd the government overshoots or undershoots, then the land value tax doesn actually fix anything.
That makes more sense for IBS (irritable bowel syndrome) than it does for IBD (inflammatory bowel disease). IBS is caused by the gut-brain nervous system acting up due to stress, and vagal maneuvers forcibly turn off the fight-or-flight response from the nervous system. On the other hand, IBD is caused by your immune system attacking itself in the gut, and the T cells aren't really connected to the nervous system in any physiologically relevant way. So your friends were probably talking about IBS, not IBD.
I don't see why not though? One of the benefits of a decentralized currency is that it is not subject to censorship by any central authority who disagrees with the content you are purchasing. That is exactly the problem these guys are having. Granted, they'll have to pay a lot of transaction fees/gas, and the volatility still makes it hard to fix prices, but for one-off transactions, you can ignore day-to-day volatility as long as the minute-to-minite volatility is bearable.
As a medical student learning thousands of facts about the human body, I think I have a relevant perspective on how to learn effectively.
The meta in medical school is currently to focus on active learning via flashcards and practice questions, and to minimize the amount of time spent watching lectures passively. I think this is, generally speaking, a good idea. The things I learn from practice questions stick in my head a lot better than the things I heard once in a lecture. However, where practice questions and flashcards come short is in making cohesive mental frameworks that organize several related topics in your head. I still think that good old-fashioned lectures are the best way to present those frameworks to students, because they don't lend themselves well to the rapid-fire questions I use for active learning. However, for learning more discrete details, I think sitting through a lecture is a waste of time. So it's about knowing the tools in your toolbox and when to use which one, if you ask me.
TL;DR active learning generally good, but need some passive lectures to show the big picture
Is it wrong for desperate patients to spend all their savings on ayahuasca-carrot juice? I don't think it's very ethical of the juice seller to exploit patients by selling them hope, but it doesn't feel like it should be illegal, since the patient does get some value out of the purchase, even if it's just (false) hope
Why does it make me happy to see that as soon as Oracle bought Cerner, all of their users ditched them? So much so that they're pivoting to "cloud services" instead?
I think it's absurd to ask OpenAI to just "recall" their trillion-dollar cash cow, but there should absolutely be legislation limiting the use of LLM's (or really any black box AI) in the criminal justice system
The problem is that it's an awful lot of work for everyone to decide for themselves, and I don't think the average user is going to care enough to build their own moderation algorithm. They'll just complain and move on
Cool, does your product fall into the category that they call "patient passports"?
Also, I looked into your website, and it looks like you have the https://www.fastenhealth.com, but you forgot to register the subdomain without the www. I remember having to configure that the last time I deployed on Netlify. Also, I was gonna try the Careers link on the website but it looks like it's a dummy link, which is fine if it's intentional.
> The risks involving hallucinations are too damn high still, and may always be.
Yes, but I think in the limited realm of people who otherwise wouldn't get any advice at all, I think LLMs could play a useful role. American healthcare is so prohibitively expensive that many people with potential medical issues will avoid seeing a doctor until it is too late to do anything. Checking in with an LLM could help people at least identify red flags that really can't be ignored, and it would be more helpful than WebMD telling you that everything is cancer.
For medical topics, I recommend Anthropic Claude. Don't want to jinx it, but so far, I've been able to get actually helpful medical information from Claude where ChatGPT just says "I'm sorry Dave, I'm afraid I can't do that"
The problem is that in the US, many EMR's don't automatically connect to each other.
If you're going from a hospital owned by one chain (say, Hospital Corporation of America) to another hospital owned by the same corporation, then yes, it works like you said and doctors have access to your historical data. But God forbid that you should ever end up at a hospital owned by a different corporation, because then you have to call the Medical Records department to release the data, which often takes a couple business days to go through. The system doesn't work if it isn't open, and unfortunately, our patient privacy law (HIPAA) is often used by corporations as an excuse to silo their data away and prevent others from accessing it.