But I believe the reason he is getting downvoted is because much of it is objectively incorrect, and people who live with ME/CFS don't have energy to go in depth to give a detailed reply.

I've replied with a more in-depth post where I cite studies that debunk many of the claims. (https://news.ycombinator.com/item?id=47914956)

ME/CFS is real. [1][2][3] Alzheimer's is terrifying, but I don't believe it is fair to say that it is obviously more terrifying than ME/CFS. People with ME/CFS have an abysmal quality of life. [4][5] And finally, ME/CFS is significantly underfunded by an orders of magnitude across all metrics, even when you account for the some of the concerns you raised such as dollars per patient, dollars per year lived with disability, etc. [6][7]

For those who haven't heard of ME/CFS, let me give you a high level summary.

We are all fatigued in this day and age, but fatigue isn't sufficient to describe what it is like living with ME/CFS.

At mild severity you are so fatigued that you have to decide between showering or doing laundry as your one physical activity for the week. But if misjudge and overexert yourself, it can take days or weeks before you get some of your energy back, assuming it ever does. Overexertion risks permanently decreasing your energy capacity.

As ME/CFS progresses in severity, the more comparable your life becomes (subjectively) comparable to someone with Locked-in syndrome. You become unable to talk and walk. Even looking at colors and hearing loud noises can be too much exertion, leading many people to wear earplugs and blindfolds for all hours of the day for months/years at a time. You have no choice to lay unmoving in bed all hours of the day with nothing but your limited thought capacity to keep you company. ME/CFS is currently incurable and virtually untreatable, but usually isn't fatal. It is possible that the disease eventually gets so severe that breathing and beating your heart is too much exertion, but that is almost unheard of. You have to live like that for rest of your life.

You are surrounded by people who don't even believe in the disease, and getting disability benefits is borderline impossible.

This disease isn't even rare - estimates are between 1-3 million people in USA alone. That's roughly 1 in 100 to 1 in 300.

Studies showing CFS is a real biological illness (there are plenty more, too many to list here) [1] https://www.mdpi.com/1648-9144/60/1/83 [2] https://pmc.ncbi.nlm.nih.gov/articles/PMC6787691/ [3] https://www.nature.com/articles/s41467-024-45107-3

Studies about MECFS quality of life: [4] https://pmc.ncbi.nlm.nih.gov/articles/PMC11356561/ [5] https://pubmed.ncbi.nlm.nih.gov/26147503/

Studies detailing how grossly underfunded CFS is [6] https://pubmed.ncbi.nlm.nih.gov/32568148/ [7] https://www.nature.com/articles/s43856-026-01516-7

I'll probably be downvoted for this, but I honestly think quality of life of CFS is lower than Alzheimer's.

I truly wish that disease funding was based on science and metrics rather than marketing and vibes.

That being said, Alzheimer's absolutely deserves it's funding and it is very sad to see setbacks related to fraud.

We have built a demo that showcases an application of this technology - natural language basketball statistic search. https://basketball.offbynone.ai

This enables asking data-driven questions in ways that have never been possible before.

Try asking it things like "If Shaq scored half as many points, which victories turn into defeats?" or "Of the players who have missed half their shots, who has played in the most games?" These questions are unanswerable by current LLM technology, even when enhanced by RAG.

The tooling we have built works from URLs or uploaded files. It auto-detects the structure and semantic meanings of the data into a blueprint. This blueprint is used to import the data into a mongo database. Additionally, it uses the blueprint to convert human-language questions into database queries.

You can have a high degree of confidence (but as with all llm technology, not 100% confidence) in the generated answers because you can verify that the LLM isn't making up the data. You can see the queries used to fetch the data as well as the raw data fetched from the database. This approach eliminates issues with the LLM making up data, but the generated answers are only as accurate as the provided source data. For example, some of the NBA records for 3 point attempts before 1982 can be a little spotty, so questions involving percentage accuracy before that date might not be accurate. Like all LLM technology, there is also a chance that it misunderstand what you are asking for.

I'm open to feedback and questions. We are open to the possibility of open-sourcing the tooling if there is enough interest.

Comments URL: https://news.ycombinator.com/item?id=43306546

Points: 2

# Comments: 0