Will Artificial Intelligence solve the riddle of ME/CFS and find an effective treatment?

[Wishful]

Yes, you can catch AIs having trouble with simple problems. Is it that much more difficult to catch researchers with fancy credentials also having trouble with simple problems? Furthermore, AIs are likely to advance in capabilities very quickly, while those human researchers are likely to get bogged down in dogma, or otherwise decline in problem-solving ability. On top of that, AIs are less likely to falsify studies to boost their careers.

At this point in AI development, I think AI will play a role in solving ME, but it needs human involvement. Maybe an AI will simply ask a question about parts of human biology that aren't yet understood, leading research into an unexplored direction.

 

[Wayne]

At this point in AI development, I think AI will play a role in solving ME

All good points! Right now, I think there's a LOT of potential for AI to assist in queries on health issues related to ME/CFS. But I think a lot of the onus is going to be on the person doing the queries, and steering AI in ways that are likely to give us the information we need.

And then it's also up to us to use (strategize?) those pieces of information to the best of our ability. In the end, AI is always going to be just a tool, although a powerful one at that. The human element is what can make the best of that tool for complex problems or situations.

 

[hapl808]

And then it's also up to us to use (strategize?) those pieces of information to the best of our ability. In the end, AI is always going to be just a tool, although a powerful one at that. The human element is what can make the best of that tool for complex problems or situations.

Agree that AI is just a tool. It is used by a human, which is why some of the moral panic is silly.

Usually the 'tests' aren't very good, because they focus on something we know AI is currently bad at - like fetching specific links. It doesn't 'think' like us in words or letters, it thinks in tokens. So a link is tough. It can spell a word perfectly, but it doesn't know how it did that.

If I asked Stephen Hawking to show me hopscotch - how can I trust someone on advanced physics if they can't do hopscotch.

Meanwhile I told GPT (in a convo about all my supplements) that my tongue / teeth felt weird - like my tongue was extra sensitive and my teeth felt weirdly smooth. It immediately suggested that my regular supplementation with magnesium might have thrown off the magnesium / zinc balance and to try maybe 15mg of zinc since it didn't look like I was taking a multi.

With all my years of research, I've never heard that too much magnesium without zinc balanced can make your teeth feel weird. Took the zinc and it quickly improved.

Never found a doctor where I could even ask those kind of questions, let alone access them multiple times a day.

 

[southwestforests]

If I was healthy enough to be a criminal mastermind it would be a hoot to write several different malware programs that infect AI with the AI equivalent of ME/CFS brain fog and rig them so that fixing one set of the malware enables another set of the malware.
And since people are already using AI to create and/or implant malware, it is only a matter of getting the programming done, assuming it isn't happening already, before the AI itself gets corrupted by malware.
Once AI gets corrupted, then what do you do?

 

[hapl808]

Once AI gets corrupted, then what do you do?

You fix it.

I wish it were that easy with the medical establishment. Once it gets corrupted, there's no fixing it.

 

[Rufous McKinney]

With all my years of research, I've never heard that too much magnesium without zinc balanced can make your teeth feel weird. Took the zinc and it quickly improved.

how fascinating

 

[Wayne]

I wish it were that easy with the medical establishment. Once it gets corrupted, there's no fixing it.

Well, I suppose we could "try" to take the huge monetary incentives out of it. .

 

[Wishful]

One thing I'm hopeful for with AI medical research is the ability to find connections from big data, which is a task I think humans are poor at. We get overwhelmed, distracted, have personal biases, etc. An AI could create hypotheses from subtle connections, and test them against existing data, quickly discarding ones that fail. It might find a hypothesis that doesn't fail from existing data, and come up with new experiments for more data from the list of known techniques, or point out a lack of data that should be investigated.

For ME, it's less important to train the AI on medical knowledge (which is lacking), and more important to train it to do proper science: propose hypotheses and test them, using available data.