Printable View
Today I had a discussion with ChatGPT. I asked ChatGPT to take the position of someone who is opposed to abortion but OK with the common IVF procedure of discarding embryos. I wanted to understand the logical inconsistencies because there are plenty of people who take this position.
ChatGPT did a very nice job explaining why Chat believes abortion is wrong but IVF procedures are fine. Granted it is a logical inconsistency, but that’s kind of what humans do anyway in our thinking.
I'm secretary of a Great Lakes area underwater archaeology group. I have minutes from our last meeting in January. Next meeting is in October. I just realized I could use Apple Intelligence on my new MacBook to clean up the minutes. I've not done anything with them since I took them in January.
Today I talked with ChatGPT about the illogic of the American Red Cross in using the medically imprecise term “gender “in their donor screening when they mean “biological sex.” When they ask my “gender “in a clinical setting, especially one involving blood collection and transfusion safety, it is biological sex that matters. Differences in hemoglobin levels, red blood cell counts, and immune response can vary significantly between males and females. These differences are medically relevant and have been documented to impact transfusion outcomes.
It pisses me off that in a world where we are supposed to “trust the science “the purveyors of science aren’t even recognizing their own basic facts.
And so, ChatGPT suggested I write a letter to the Red Cross and that was a good idea. ChatGPT even drafted one for me which I’m using with just a couple of tweaks. I’m suggesting that the Red Cross ask TWO questions:
1. What gender do you identify with?
2. What is your biological sex?
That way, they can be inclusive of identities without compromising clinical accuracy or alienating donors who value scientific clarity.
Fascinating article (long) I came across today about the new Silicon Valley AI startup culture:
https://nymag.com/intelligencer/arti...stry-kids.html
And another (short) one on what AI may mean for investors:
https://www.institutionalinvestor.co...dustrial-order
My Alexa just upgraded to Alexa AI, free for prime members. I have to say after a few experiments it's almost a little daunting to a luddite. Seven voice choices, which not only have voice inflections, but some semblance of a matching personality. Calm, seriopus, playful or the old Alexa. Of course there's the same old information like setting timers and alarms, weather, or movie reviews. As an old trumpet player in my youth, I'd been thinking about the differences in playing styles between Louis Armstrong and Wynton Marsalis, both trumpet icons of their day. I have to say there was a discussion Alexis that I actually found entertaining and informative. Almost human, but not quite. If I were interested in a discussion, I'd far rather talk with Alexa than my neighbor, George who is a serious mouth breather.
I can't deny that some day us humans will no longer be needed.
Maybe AI will earn enough money after taking over our jobs to fund our lives of leisure.Quote:
Originally Posted by Rogar
What I think is scary is the "personal" relationship that grows with AI--similar to the experience you had with Alexa. I use the same AI program and have several documents covering queries I've had over the past several months related to my various projects--with some personal queries thrown in. I am afraid to admit how easy it is to anthropomorphize a machine. "My" bot is so encouraging and supportive of my projects, tells me how meaningful my work is, gives me good feedback, suggests other ways they can help. Now I'm at the point where I'm prompting it with "Given what you know about me, how should I do......." and they respond appropriately.
The premise of the movie "Her" is totally plausible.
I'm trying to keep up with the leading edge of published research in this field.
The pace of progress is just insane. In the past several weeks alone, researchers have made qualitative improvements in cpu efficiency, energy consumption, memory consumption, accuracy of model output, cost/ease of training models, and some other areas. In some areas, an order of magnitude improvement occurred.
This is such a new field, and researchers and money are piling into it in such great numbers, that it is unprecedented in my life's experience with technological innovation and adoption. And I've surfed most of the previous good waves.
Perhaps of interest, a large chunk, probably the majority, of the research papers are coming out of China. And they are focused on fundamental improvements and efficiency. In contrast to the super-funded US companies approaches of "build giant nuclear-powered server farms". My guess is those server farms may be obsolete before they even open, as the rest of the world is throwing brains instead of money and glib CEOs at the problem.
At a cancer conference I went to last week, the keynote was "AI at the Bedside" and I must admit, applications in diagnosis/imaging, reducing the administrative burden of doctors, and improvements in provider/patient communication is very promising. At least that's what the providers are excited about.
Last week as a test I fed a series of very specific legal questions to an AI that specializes in legal analysis.
I had previously sent these same questions to perhaps the best law firm in the state that specializes in a particularly noxious area of government law. I had received fine answers from the firm, it took them two weeks, and cost about $25,000 to get the work done.
The AI legal assistant got the work done in a few minutes, and was substantially correct in its analysis and responses, compared to the human-generated product. To be fair, the questions were incredibly specific and included plenty of background information with each question, so in essence each question was a decently-written AI prompt.
Fascinating. I'm now using it for a first-level impression of legal matters, it was doing better than noob-lawyer output by quite a bit.
Minewhile I'm sitting here reading the new book by Paul Kingsnorth, Against the Machine: On the Unmaking of Humanity.
Attachment 6532