[ExI] chatgpt test 2: mostly successful was RE: chatgpt test 1: fail

Mon Apr 3 14:29:32 UTC 2023

>…Spike,
They added for sure many measures for ChatGPT to be conservative on certain topics. Medical advice is one of these. This why it tends to tell you to consult a doctor. I think with different training it could be more useful to explore innovation in the medical field. 
But notwithstanding these limitations you can still get some useful answer if you try to express more clearly your point of view to ChatGPT or continue in a given line of inquiry….

Hi Giovanni,

There is an interesting background story here.  A bit over 10 years ago, a family member developed a version of Alzheimers which was advancing more quickly than most cases do, and she was debilitated with it.  She would need to go into a nursing home, but the cost of those facilities was staggering, a strain on our family’s means.  I was on the Alzheimers network, where I was getting notices of any new developments.  All we had at the time was nicotine, but the patient was unwilling to take it because of religious considerations (reasoning: nicotine was derived from tobacco, and tobacco is evil, therefore nicotine is evil (however we had been thru this nightmare before with her father (and nicotine was the only medication that seemed to do anything at all (which isn’t much.))))

So I heard of the experiments on the Alzheimers mice and how bexarotene helped them.  Knowing just enough about chemistry to be dangerous, I figured out what dosage they were giving to the mice, then went into the solubility tables, figured out what it would take to deliver that to a patient, realized it wouldn’t work, unless… it was dissolved in ethanol.  If a prole knew how to measure out a few milligrams (I bought a milligram lab scale) and recognized that the solubility of bexarotene goes way up with temperature of alcohol, and people drink warm alcohol (the Japanese people drink sake) then with about 36 grams of ethanol a prole could dissolve the dose which was therapeutic to the mice.

I learned that the alcohol dissolved bex might be therapeutic and might cross the B^3, so… I wanted to try it, but… this particular patient would consider that a no-go because… alcohol is evil, all alcohol, including the solvent for vanilla, so… she had never actually tasted real vanilla.  OK then.

An adventure occurred when the Los Angeles times printed an article which had a critical mistake, which I spotted within a minute of its being posted online, ending up in my getting on the phone with the Times at about 9pm on a Saturday night, urging the, rather demanding, that they take that article down immediately before they kill somebody with an overdose of bexarotene based on information in that article.  For over three hours, that misadventure unfolded, and they took down the erroneous number.  The rest of the article was OK.

That adventure is what led me to say GPT gave the right advice: don’t do this crap if you don’t know what you are doing.  Doing this kind of crap is a possible reason we no longer have Robert Bradbury among us.  It blows my mind to realize he has been gone over 12 years now.

spike

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20230403/3b814f6c/attachment.htm>