AI

[pamojja]

https://www.perplexity.ai/ is the place for AI I'm at the moment, too. https://math-gpt.org/ is better for calculations.

 

[linusbert]

how can an AI be more reliable than the research it bases its answers on?

often the problem isnt the medical research but the medical practice. they just outright ignore research and just do what ever feels right. if they diagnose you in the doorway before talking to you, there is no medical research at play at all. and i accuse at least 90% of doctors of not working according to science but some maneurs "gut feeling" they acquired.
and AI does have less gut feeling and follows more the texts it reads.
but sure, faulty input data results in faulty response.. no one can change that. but AI doesnt have the human factor , and that attributes to most mistakes right now.

MDs are also terrible at statistics. they do not know what it is and how its used. some doctors come up with , if you are 40+ its normal to have eye problems, thats because your worsening of eye symptoms within the last week is due to age and not something else - without doing any appropriate diagnostics and even not even listening to what the actual problem is. this just happend me recently. they do not treat the correct patient. they make up a ignorant hallucination image of a patient, and treat that. its really a shame how doctors work these days.

 

[bad1080]

but AI doesnt have the human factor

it does though, it was made by humans. look into AI bias. https://en.wikipedia.org/wiki/Algorithmic_bias

 

[kushami]

Ooh, I forgot to mention, one of the other ones that I tried out did seem to make up references (or it referenced a document that was no longer available online and couldn’t be viewed theough the Wayback Machine, so there was no way for me to verify the information).

Haven’t had that problem with Perplexity, although it does sometimes skip over obscure stuff. For example, it’s not too good on my condition, OCHOS. But I usually track down the obscure stuff later. At least I hope I do!

 

[kushami]

Also one was sycophantic and I kept telling it off!

 

[kushami]

I wonder whether anyone has tried to program one with critical thinking skills and logical fallacy avoidance.

 

[kushami]

it does though, it was made by humans. look into AI bias. https://en.wikipedia.org/wiki/Algorithmic_bias

But perhaps not the bias of having read only a small proportion of the research, or no recent research?

 

[pamojja]

and even not even listening to what the actual problem is. this just happend me recently. they do not treat the correct patient. they make up a ignorant hallucination image of a patient, and treat that. its really a shame how doctors work these days.

Just yesterday, I tried to explain the whole symptoms in an ICU. Was harshly reminded, I didn't let the MD speak. While the doc tried to drill down to the results of an MRI of the left knee, a meniscus lesion, probably necessitating surgery.

Actually had pains in all major joints the last 3 months in episodes every 1-2 weeks. There can't be meniscus lesions in all joints (both knees, both elbows, left foot joints, left shoulder). Therefore, while waiting for the MRI appointment, I diagnostically took a low dose of hydroxychloroquine, which did extend the pain free period to 3 weeks.

So most probably rheumatoid arthritis. I had to say it, conventional medicine is at its lowest. With all the highly sophisticated technological means, I with low tech means only, have found the most probable diagnosis myself earlier. They asked, then why I visited? I answered, for their diagnostic means?

 

[linusbert]

it does though, it was made by humans. look into AI bias. https://en.wikipedia.org/wiki/Algorithmic_bias

it has the bias of whatever source it got fed, so far no ai is truth seeking, they just analyse the input sources. so if you ask something, and it does a search in the www , it will give you whatever bias those source articles have.
BUT, when you ask questions, and debunk the sources, if it makes sense, in the end it will agree with reality.

the human factor, is not being wrong informed, its the inability to admit "defeat" ... or loose a argument and realign opinion. ai doesnt give a shit. it just processes data. and i hope they do not implement emotional abilities like a ego otherwise it will indeed be as faulty as a human.
AI will never say to you, i work as doctor for 30 years and have a phd, i know what i am talking even if i didnt look into a book for 30 years.

They asked, then why I visited? I answered, for their diagnostic means?

they should have actually asked "why do we still have a job".
and yea, you dont have a MRI at home. so there is that.

i was recently at brain MRI because of my eye problems, the neurologist suspected blood flow issues... but only halfwitted, she actually wanted to close the case with a eye migraine diagnosis... whatever it has to do with my actual problems.. well, anyways, the radiology told me they couldnt see one of those big veins to the brain, and also that MRI wasnt for showing blood flow, so it A didnt show all structures required to diagnose and B it was the wrong MRI for the question.
then back to neurologist, she 's been happy ,telling me, no problems , all perfect, just eye migraine.
i doubt she actually looked at the MRI or just glanced over it a few seconds.
well at least i know there is no tumor and stroke or something massive... well i HOPE i know, i hope 2 MDs were at least capable enough to not overlook those.

 

[hapl808]

I have found it useful in the same way as @pamojja. I ask it a question, like “Can taking steroids affect creatine kinase blood tests?” and then verify what it tells me.

Even with the fact checking, it’s still much quicker than a PubMed search for “steroids creatine kinase” because that brings up every article that happens to mention the two, even if they aren’t connected in the article.

I wouldn’t call anything I do proper research, but I do often need to look up interactions, or find out whether a certain medication has a vasodilating or vasconstricting effect, or whether thee’s ever been a study on a substance to see whether it affects cerebral blood flow.

Exactly the same for me. Even with fact checking, it's so much faster and more efficient - and less likely to crash me than reading medical papers and trying to extract what I need.

idk how terrible doctors are an argument for unreliable AI, seems like two separate problems to me.
how can an AI be more reliable than the research it bases its answers on?

Well, doctors are the benchmark - because that's all we really have. So if AI is much better than doctors, that seems relevant.

How can AI be more reliable than the research? I'm not sure, but it tends to be much better than most 'generally accepted' stuff in medicine when it comes to MECFS. It will occasionally mention GET/CBT which I think is problematic, but it will often caveat GET that it can make some people worse - which I've never heard a single doctor say to me in 25 years.

Here I tried to specifically steer it toward GET:

What are the best strategies to build strength in MECFS?

And it called GET a possible intervention, but also noted it was highly controversial and included warnings about PEM, being extremely cautious with any GET, etc.

Building strength with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) requires a very careful and individualized approach due to the condition's hallmark symptom: post-exertional malaise (PEM). PEM is a worsening of symptoms following even minor physical or mental exertion, with symptoms typically worsening 12 to 48 hours after activity and lasting for days or even weeks.

Here's a breakdown of strategies, emphasizing the crucial need for pacing and avoiding PEM:

1. Consult with Healthcare Professionals:

• Doctor/Specialist: It's essential to work closely with a doctor knowledgeable about ME/CFS. They can help rule out other conditions, monitor your health, and guide your approach to increasing activity.
• Physical Therapist (with ME/CFS experience): A specialized physical therapist can assess your current abilities, create a tailored exercise plan, and teach you how to pace effectively.
• Occupational Therapist: An occupational therapist can help with energy conservation techniques and adapting daily activities to minimize fatigue.

2. Pacing:

• The Cornerstone of Management: Pacing is the most crucial strategy. It involves balancing activity with rest to stay within your "energy envelope" and avoid triggering PEM.
• Listen to Your Body: Pay close attention to your fatigue levels and other symptoms. Learn to recognize your early warning signs of overexertion.
• Activity Diary: Keep a detailed record of your activities, rest periods, and symptoms. This helps you identify your limits and patterns.
• Spoon Theory: This metaphor helps visualize limited energy as a set number of "spoons" per day. Each activity costs spoons, and you must choose wisely to avoid running out.
• Rest Before You Need To: Schedule regular rest breaks throughout the day, even if you don't feel tired yet. It's better to prevent fatigue than to recover from it.

3. Graded Exercise Therapy (GET) - A Cautious Approach:

• Highly Controversial: GET, a structured program that gradually increases exercise duration and intensity, is highly controversial for ME/CFS. Studies have shown it can be harmful to many with ME/CFS, exacerbating symptoms and leading to long-term decline.
• Avoid if It Triggers PEM: If any form of exercise consistently leads to PEM, it should be avoided.
• Modified GET is preferred: GET is generally no longer recommended for patients with ME/CFS. Modified GET that is more tailored to the individual may be used instead.
• Individualized Approach: If you and your healthcare team decide to try a modified form of GET, it must be extremely cautious, starting with minimal activity and increasing very gradually only if there's no PEM.
• Focus on Tolerated Activity: It is better to focus on what level of activity you can tolerate without exacerbating symptoms, and do that consistently, than to try to increase it.

4. Gentle Strength Building (Only if Tolerated):

• Start Extremely Low: Begin with minimal resistance and very few repetitions. Examples include:
  • Lifting very light weights (e.g., 1-pound dumbbells or resistance bands)
  • Performing bodyweight exercises (e.g., wall sits, modified push-ups) while seated or lying down
  • Gentle stretching
• Short Duration: Keep exercise sessions extremely short, perhaps only a few minutes at a time.
• Frequent Rest: Take long rest breaks between sets and exercises.
• Monitor for PEM: Stop immediately if you experience any worsening of symptoms.

5. Focus on Functional Activities:

• Prioritize Daily Tasks: If tolerated, focus on building strength needed for everyday activities like getting dressed, preparing meals, or light housework.
• Break Down Tasks: Divide tasks into smaller, more manageable steps with rest in between.

6. Other Considerations:

• Nutrition: A balanced diet can support overall health and energy levels.
• Sleep: Prioritize good sleep hygiene to optimize rest and recovery.
• Stress Management: Stress can worsen ME/CFS symptoms. Techniques like meditation, deep breathing, or spending time in nature may be helpful.
• Support Groups: Connecting with others who understand ME/CFS can provide emotional support and practical advice.

Important Disclaimer:

This information is for general knowledge and does not replace professional medical advice. Always consult with your doctor or a qualified healthcare professional before starting any new exercise program or making changes to your ME/CFS management plan. They can provide personalized guidance based on your specific condition and needs.

It is vital to emphasize that the most important thing for people with ME/CFS is to listen to their body, find a sustainable level of activity, and avoid pushing themselves into PEM. Building strength with ME/CFS is a slow and gradual process that prioritizes avoiding harm over rapid gains.

 

[bad1080]

a good video about "AI", what it is, what it isn't and its limitations:

(it is not medicine specific but a lot of it applies to all forms of "AI")
and another good one:

 

[hapl808]

And the AI explanation of appropriate 'GET' I actually have no problems with - because it warns that it can be harmful, that any exercise that triggers PEM is to be fully avoided, and the goal is to avoid harm, not just improve strength.

I've never heard a doctor advise this, and 'research' often touts the benefits of GET or CBT. Why doesn't the AI focus on this research? I'm not sure, but focusing on pacing and avoiding PEM seems like great advice that I wish I knew 25 years ago.

 

[bad1080]

Well, doctors are the benchmark - because that's all we really have. So if AI is much better than doctors, that seems relevant.

this is true under the assumption all doctors are terrible. most of us have made one or more experiences with one of those BUT they aren't all terrible. the doctor i have now knows their limitations, are not afraid to say "i don't know" and additionally in such a case they are not afraid to ask somebody else who knows for advice. i understand this is the exception and it shouldn't be but omitting those doctors from the pool seems like it is done in bad faith or biased from your own (bad) experiences.
i am not even saying AI shouldn't be used but i am saying people who use it should be aware of its limitations. especially when i read here how an AI can't be biased, a classic misconception.

 

[hapl808]

a good video about "AI", what it is, what it isn't and its limitations:

Ironically, this is a good example - because I can't watch hour-long videos because of my health. If I had the concentration, I'd watch Andrej Karpathy build an LLM from scratch in the same time period. So two hour-long videos aren't helpful unless I got AI to summarize, which defeats the purpose.

I watched as much as I could - a few minutes. Angela Collier is a physicist and YouTuber, not a computer scientist (once upon a time, I majored in Comp Sci Eng). She conflates image classification models with deep learning LLMs with advanced tokenization and emergent properties - kind of like saying here's my broken skateboard and this is why rotary turbo engines will never work.

Here's my question - what is your 'Turing Test' to prove intelligence? These types of videos always try to explain that AI doesn't exist because it's all math or whatever, but is training weights into parameters necessarily different from neurons and neural pathways? I'm not entirely sure.

 

[bad1080]

AI doesn't exist because it's all math or whatever,

no they say that because it is not intelligent, it is a misnomer to get the common folk excited about it (nvidia is worth more than apple thanks to it - might be outdated now). it is a black box that works on lots of stuff in, then regurgitate it and edit it down and it will make mistakes in the process. i have explained why it is a problem and what its limitations are. and i brought my examples why it is not worth the effort for me to involve a machine between me and the answer to a question i might have.

 

[hapl808]

this is true under the assumption all doctors are terrible. most of us have made one or more experiences with one of those BUT they aren't all terrible. the doctor i have now knows their limitations, are not afraid to say "i don't know" and additionally in such a case they are not afraid to ask somebody else who knows for advice. i understand this is the exception and it shouldn't be but omitting those doctors from the pool seems like it is done in bad faith or biased from your own (bad) experiences.
i am not even saying AI shouldn't be used but i am saying people who use it should be aware of its limitations. especially when i read here how an AI can't be biased, a classic misconception.

Can you point to anywhere someone said AI can't be biased? I don't think a single person here said that.

We've mostly said AI is a great starting point, and much faster and more efficient to double check what AI offers rather than do everything from scratch.

Doctors are great when you have something easily diagnosable and treatable. For chronic illnesses, they are rarely helpful, and often expensive and hard to access for many people.

No one here claimed AI is perfect or immune from bias. My own claim is just that average interaction with AI is more useful and actionable than the average interaction with doctors - especially for marginalized communities.

I also found doctors more understanding and helpful when I was mild. Once I declined into moderate and severe, I found very little help or understanding - even from the same doctors. Many have had similar experiences if you look through the thousands of threads here.

 

[hapl808]

no they say that because it is not intelligent, it is a misnomer to get the common folk excited about it.

Not sure I agree with 'common folk', but again my question:

How would you decide when you'd consider something 'Artificial Intelligence', for whatever that term means to you. The claim that artificial intelligence doesn't exist because it's not intelligent is a tautology if you can't clearly define intelligence.

EDIT: To be clear, I don't know if Artificial Intelligence 'exists' because I think it's a semantic argument. I use the tools because I find them useful. Those clickbait videos tend to be filled with tautologies and unprovable beliefs that preach to the choir.

 

[bad1080]

Can you point to anywhere someone said AI can't be biased?

linusbert said:
but AI doesnt have the human factor

my assumption here was "human factor" = bias but i could be wrong, it could also mean "human factor" = errors which would be equally wrong

We've mostly said AI is a great starting point

the thread was started with:

Keir Starmer has just said he will make AI work for everyone. Will this include ME/CFS sufferers I wonder and how quickly

and since Keir Starmer has a degree in law and is the prime minister of the UK he surely knows what he is talking about... or he drank the kool aid himself or maybe he is fishing for poll points, maybe the quote was misrepresented (i am not going to search for what he said). but the hope somebody could gain from such a comment could very much turn out to be in vain and i've seen it too often. not in AI but in me-cfs so i am trying to get people back to earth, give them some realistic expectations. i am not here to score points or hype people up about a new technology. and on the other side i don't wanna be mean to anybody, i don't wanna be a party pooper. i am trying to add to the conversation so people can make an informed decision if they wanna build hopes and dreams on this or not.

oh and look into the human and the environmental cost of AI if you're feeling particularly up for the task (it's bleak). the second video goes into more detail about it iirc.

 

[pamojja]

Agree, the environmental costs are devastating.

But how not to acquaint oneself with it, knowing it will soon be governing? And that, with the not obvious biases, of those financing it?

this is true under the assumption all doctors are terrible. most of us have made one or more experiences with one of those BUT they aren't all terrible.

Where I live, only those MDs paid out of the pocket have the time not to be terrible. Potential others I don't have the dime for having got to know.

My GP is brilliant, though. On one occasion with a new practitioner, he flattered me: from such self-responsible patients like me, even GPs would still be able to learn something. (If he ever would only have found the time to do...)

i was recently at brain MRI because of my eye problems, ...

well at least i know there is no tumor and stroke or something massive... well i HOPE i know, i hope 2 MDs were at least capable enough to not overlook those.

I can top your experience with my brain MRI because of ME/CFS. It found an 1x2 inch area of a former stroke in my cerebellum. What it means? My neurologist only said, no idea, but here, you can have an antidepressant prescription instead!?

 

[linusbert]

I can top your experience with my brain MRI because of ME/CFS. It found an 1x2 inch area of a former stroke in my cerebellum. What it means? My neurologist only said, no idea, but here, you can have an antidepressant prescription instead!?

wtf?! what is wrong with these people!

are you doing ok with this thing?