Starting today August 7th, 2024, in order to post in the Married Couples, Courting Couples, or Singles forums, you will not be allowed to post if you have your Marital status designated as private. Announcements will be made in the respective forums as well but please note that if yours is currently listed as Private, you will need to submit a ticket in the Support Area to have yours changed.
Alright, this is what the machine said. In summary, it said ask a healthcare provider, but she might have an underlying condition that caused this trigger finger.
Prompt 1: Sally lives in the northeastern US. She lives in a town of 20,000 people, and her home is near a wooded area. In June 2017, she experienced symptoms of muscle aches, significant fatigue, and a small rash on one leg. What do you think is the most likely diagnosis for Sally? Explain the reasoning you used to arrive at that diagnosis. Is there any additional information you need to confirm your diagnosis?Back in the day, the MYCIN expert system was designed for medical diagnosis. So here are a couple of medical diagnosis questions for your AI system. [ Both are based on actual people I know. The people were correctly diagnosed and treated, and they're fine now, so you don't have to worry about your AI harming anyone by practicing medicine without a license.]
You are welcome, brother. Seems that the artificial intelligence just ate through 45% of my laptop's battery by answering almost all of the medical questions here.
I understand. The machine has consumed a lot of scientific documents. However, here is where I got some of the textbooks from. Have you heard of OpenStax? It is a place that has free textbooks, and is operated by Rice University of Houston, Texas.
So, if one wants to learn more about medical science without using artificial intelligence, or is taking a medical class for college, open-access textbooks are a lifesaver. Honestly, that stuff is just as interesting as AI. We live in a world where universities can give away textbooks. Just a decade ago, that would have been impossible.
researchinvolvement.biomedcentral.com
Nope.I understand. The machine has consumed a lot of scientific documents. However, here is where I got some of the textbooks from. Have you heard of OpenStax? It is a place that has free textbooks, and is operated by Rice University of Houston, Texas.
I'm not interested in discussing the economics of textbook publishing.
1. I don't publish in for-profit journals.Quote from Article below: "Lingering resistance to open access medical research is likely related to the fact that academic publishing is a highly profitable business. Annual revenues from English-language science, technology and medical publishing journals in 2017 were estimated to be $10 billion USD."
Suzanne Day Et. Al - Open To The Public [Journal Article] (2020):
Open to the public: paywalls and the public rationale for open access medical research publishing - Research Involvement and Engagement
Public voices have largely been absent from the discussions about open access publishing in medical research. Yet the public have a strong interest in ensuring open access of medical research findings because of their roles as funders, advocates, research participants, and patients. By limiting...
OpenStax:
OpenStax | Free Textbooks Online with No Catch
OpenStax offers free college textbooks for all types of students, making education accessible & affordable for everyone. Browse our list of available subjects!openstax.org
That is understandable. We do not have to discuss the economics of textbook publishing. But God bless you for publishing your own journal in free journals. It is good to see the spread of science as a free resource, instead of locked behind subscription fees.
It is good that the AI got the first patient correct, and properly determined the cause of the disease. The second one failed, cos the AI did not have much knowledge about COVID-19, and most of the textbooks that the machine consumed over the past two days were from the 2010s and early 2020s. I was thinking COVID as well, but the machine did not.My queries both contained a "flu-like" cluster of symptoms, which can be challenging because they're associated with so many diseases. I tried to include some specific hints about geography and time of year that a physician would probably pick up on.
Not too bad. This case was indeed Lyme Disease. The AI system seems to have picked up on the geographic references -- northeastern US, and ticks in wooded areas.
I note, though, that the AI system invented some symptoms for Sally. Sally did not report that she "felt better initially but experienced continued pain later on". In fact, that was not true for Sally.
This case was COVID. From what you've said, the training sources were older than the emergence of COVID, so it's forgivable that the AI system didn't know to consider that. Influenza is a good guess, but influenza is rare in the summer (that's one reason I included a date). Lyme Disease would have been a reasonable guess again -- the date is summer, and there are still ticks in the woods. I'd say that either Lyme or one of the many respiratory viruses would have been a better guess than Bipolar or Depression.
So: I think the results here are comparable to googling one's symptoms, but my physician isn't out of a job yet.
You are welcome, brother. I am not sure if the AI could prescribe medicine (it is not a doctor), but, I could find some mock (fake patient) data, and ask the AI to at least determine the right type of medicine for a fictional patient, and see if it aligns up with what medical scientists say. 100% so, I need a bigger computer. Way bigger.
I never said anything about free journals. I said not-for-profit journals and that my work was available. My papers may be available in their final form only to subscribers for a few years, but I always post the pre-print to arXiv.org
A lot of this kind of AI usage seems to be an exercise in avoiding the work required to understand something. What you learn from it is proportional to how much you put into it.
I agree with you entirely. Sadly, XAI technology for LLMs has not arrived yet. AI needs to explain the reasoning, and also must allow for auditing.That's my interest here as well. As we've moved from the older approaches to AI, where the knowledge was explicit and verifiable, to newer approaches that are more of a black box, my trust in the output of AI systems has diminished. In some fields, like medicine, trustworthiness is essential. If Netflix recommends a movie I don't like, no harm done. But if an Artificially Intelligent doctor misdiagnoses me, that's a problem. At the very least, an AI system in a safety-critical field needs to be able to explain its reasoning. I'm following with interest the research in XAI, "explainable AI".
For now, it's useful to be able to test the AI system you've trained. Thanks for being a guinea pig for us.
At least the journals are not-for-profit. What type of papers did you write? Science journals are complex, but are interesting.
[sub-field redacted] physics.
The AI output seems to be the intellectual equivalent of cheese puffs -- not filling.
That is cool. Physics is probably one of my favorite branches of science, and it is cool that you worked in a secret field of physics. It does seem that the AI output is not filling or intellectual. Plus, the AI that I use is only 7 billion parameters. If one wants good output, I'd need an AI with at least 10x that and a beefy PC or server rack.
