3 ms·

I hate to break it to all the doomsayers here, but... "According to a 2014 press release issued by the Agency for Healthcare Research and Quality (AHRQ), a div

I hate to break it to all the doomsayers here, but... "According to a 2014 press release issued by the Agency for Healthcare Research and Quality (AHRQ), a division of the U.S. Department of Health and Human Services, diagnostic errors affect about one in 20 adults, or about 5% of adult patients in the United States. That means about 12 million U.S. adults will become victims of a misdiagnosis during a single year. These figures come from 2014 study published in BMJ Quality and Safety." Source: https://www.wieandlaw.com/how-often-do-doctors-make-mistakes/ https://www.wieandlaw.com/how-often-do-doctors-make-mistakes... 5% error rate. If this tool is even 1% lower on average, it's already better than a human. The next thing to control for will be severity of the mistakes made. Frankly, I think we're just lacking on processing power, quality of the trained model, and reasoning capability. Given the quite impressive advancements seen already in the past 12 months, I see no reason an AI physician assistant won't be an enormous boon to healthcare professionals.

ejstronge 3y ago
> If this tool is even 1% lower on average, it's already better than a human. Without diving into the source data behind the 'wiendlaw.com' article (are they authoritative sources? Can we believe their interpretation of a biomedical journal article?), it is facile and likely wrong to conclude, a priori, that one distribution of errors (that of humans) is indistinguishable from another distribution of errors (your hypothetical LLM with '1% lower on average' errors). It is often said that 50% of biomedical facts are wrong. The trouble, of course, is determining which half of the textbook to disregard. While this trope is cute and also incorrect, it is difficult to imagine how an LLM trained on past knowledge will be able to create the new knowledge required to establish new facts
- famouswaffles 3y ago
  >It is often said that 50% of biomedical facts are wrong. The trouble, of course, is determining which half of the textbook to disregard. So abduction ? Good thing that's what LLM do best I guess. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10425828/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10425828/
- whatscooking 3y ago
  Most people can’t create new knowledge from established facts in the first place. Have you met some of the providers? They have a hard time discerning a rock from a pillow
johntiger1 3y ago
Yep, this space is definitely heating up, see: [1]: https://www.reuters.com/business/healthcare-pharmaceuticals/hippocratic-health-raises-50-mln-seed-funding-build-ai-model-2023-05-16/ https://www.reuters.com/business/healthcare-pharmaceuticals/... [2]: https://cloud.google.com/blog/topics/healthcare-life-sciences/sharing-google-med-palm-2-medical-large-language-model https://cloud.google.com/blog/topics/healthcare-life-science...
gentleman11 3y ago
About half of what chatgpt tells me is a hallucination. So humans are 10x better by your metric

ejstronge 3y ago

> If this tool is even 1% lower on average, it's already better than a human. Without diving into the source data behind the 'wiendlaw.com' article (are they authoritative sources? Can we believe their interpretation of a biomedical journal article?), it is facile and likely wrong to conclude, a priori, that one distribution of errors (that of humans) is indistinguishable from another distribution of errors (your hypothetical LLM with '1% lower on average' errors). It is often said that 50% of biomedical facts are wrong. The trouble, of course, is determining which half of the textbook to disregard. While this trope is cute and also incorrect, it is difficult to imagine how an LLM trained on past knowledge will be able to create the new knowledge required to establish new facts

famouswaffles 3y ago
>It is often said that 50% of biomedical facts are wrong. The trouble, of course, is determining which half of the textbook to disregard. So abduction ? Good thing that's what LLM do best I guess. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10425828/ https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10425828/
whatscooking 3y ago
Most people can’t create new knowledge from established facts in the first place. Have you met some of the providers? They have a hard time discerning a rock from a pillow

johntiger1 3y ago

Yep, this space is definitely heating up, see: [1]: https://www.reuters.com/business/healthcare-pharmaceuticals/hippocratic-health-raises-50-mln-seed-funding-build-ai-model-2023-05-16/ https://www.reuters.com/business/healthcare-pharmaceuticals/... [2]: https://cloud.google.com/blog/topics/healthcare-life-sciences/sharing-google-med-palm-2-medical-large-language-model https://cloud.google.com/blog/topics/healthcare-life-science...

gentleman11 3y ago

About half of what chatgpt tells me is a hallucination. So humans are 10x better by your metric