"Mayo’s LLM split the summaries it generated into individual facts, then matched those back to source documents. A second LLM then scored how well the facts aligned with those sources, specifically if there was a causal relationship between the two."
It doesn't sound novel from the article. I built something similar over a year ago. Here's a related example from Langchain "How to get a RAG application to add citations" https://python.langchain.com/docs/how_to/qa_citations/
I don't think you're getting it, it's not traditional RAG citations.
They are checking the _generated_ text by trying to find documents containing the facts, then rating how relevant (casually related) those facts are. This is different from looking up documents to generate an answer for a prompt. It's the reverse. Once the answer has been generated they essentially fact check it.
That would not have solved the problem in this fire since wind speed was so high. The videos showed embers traveling far and fast. Having a 10 foot fire break would not have prevented the spread. One thing to look into is how the fire started and if the electrical equipment can be made safer, like being underground in some places.
I made something like this to let people explore the lore of my game. It has an AI act like a dungeon master as you play through a story. You can try it for free at:
https://orbsccg.com/adventures
The main game is a strategic collectible card game, but the “adventure mode” is just for fun AI-powered stories.
I don’t know why this is downvoted. The lack of profit motive is a big reason that nutrition and supplements aren’t as well studied through rigorous trials as drug therapies. The ones that are run are funded by grants. Rather than just “more funding” I think there needs to be more systemic ways at reducing the cost of clinical trials or using alternate methods of getting high quality scientific data for answering these questions.
For example, there is a good trial running now on ketogenic diet in glioblastoma patients, NCT05708352, I think with a NIH grant and maybe the NCI as well. Here is a video about it: https://www.youtube.com/watch?v=W31kR0MzyRA
It still is pretty good. That and the ACM CACM are my only subscriptions. I independently purchase the Spectrum only because I don't want to be associated with IEEE anymore.
My reasons of cancellation are less fiduciary and more of a policy. In 2022, IEEE published an interview of two Russian researchers and effect of the intellectual & economic blockade in Russian academia after Ukraine invasion started. It sought pleas for donation & help. Many people pointed out it was tone deaf since the ones really suffering were the Ukrainian academics who were dislocated or persecuted in their own land. IEEE never commented or retracted those interviews & several people canceled their membership in disgust.
I subscribe to Spectrum for about $50 through Omeda rather than continue giving IEEE additional membership fee. I know some part will be shared with IEEE, but I am minimizing this by conscious choice.
> several people canceled their membership in disgust
Looks like nothing of value was lost.
Ukraine already gets 99.99% of airtime. Two Russian researchers -- not army generals or politicians, mind you -- were given a chance to describe the problems the conflict has created for their side, and we're supposed to pretend that those problems do not exist and their struggle doesn't matter because somebody else has it worse.
There's this thing called "journalism" where you're supposed to be given raw facts from all sides with as little interpretation (i.e. propaganda) as possible, and think for yourself. It used to be more common back in the day, though maybe not much. Now we only want to hear things that closely follow our own narrative, and ignore everything else. It's easier to live this way, I understand.
There are many cases where you're getting blood drawn already and if adding another test onto that is inexpensive, that's a pretty convenient add-on. That's useful for many reasons.
I was also wondering how it performed compared to a stool test, so I looked it up, and found "While the blood test caught 83% of the cancers found by colonoscopy, it missed 17%. That’s on par with stool-based tests."
And will combining the two tests catch a higher percentage? Could go as high as catching 97%, although I suspect lower due to some correlation between what cancers the two tests will pick up (eg. not advanced enough to leak blood into stool may equate to not advanced enough to leave markers in blood). But it will also increase chances of a false positive.
It doesn't sound novel from the article. I built something similar over a year ago. Here's a related example from Langchain "How to get a RAG application to add citations" https://python.langchain.com/docs/how_to/qa_citations/