Why I am Ord-AI-ning an A.I. as a Soto Zen Priest (1st in a Series)

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • Kaitan
    Member
    • Mar 2023
    • 547

    #31
    Originally posted by Jundo

    Perhaps I need to get this done faster, time may not be on our side, says Sabine Hossenfelder ...
    .







    .
    Perhaps what we are seeing, similar to human intelligence, is creativity as an emergent property. For example, some say that A.I. is going to run into a wall, and be limited, because it has harvested all the data on the internet ... so no more fuel for the system. However, perhaps that is like saying that human beings would run of data and stop being creative if we read all the books in the book store. Rather, the A.I. continues to be creative by, for example, (1) finding relationships among the existing data that we miss, for example, seeing relationships among medical studies buried in obscure journals, and "hit and miss" chemical experiments in hours that might take years for human beings, and (2) random "mutations" that produce something useful, e.g., A.I. generators produce random images with mistakes, but some of those mistakes may be valuable much as nature's mistake/mutations are the seeds of natural selection. Perhaps A.I. can discover new physics, new chemistry, and even new ways to engineer better A.I. in this way ... all of which seems to "work" and be functional, but in ways we can barely understand.

    I definitely want such super-intelligent A.I. to have a basic Bodhisattva's concern for the welfare of all sentient beings, not merely itself.

    Gassho, J

    stlah
    Thank you for sharing this. I found interesting the part were she mentions
    the larger they become the more random deviations you are going to get from the original code
    . This seems to be key because we know from natural selection that random changes in evolution are the driving force for the wide diversity of species and biologists would agree with that. But even Sabine said that it is quite speculative in case computers having different hardware.

    I'm also not so sure about the random generation of images from AI, what I know is that they have a pseudo random code. True randomness is a fascinating topic.




    Gasshō

    stlah, Kaitan
    Last edited by Kaitan; 07-10-2024, 04:42 AM.
    Kaitan - 界探 - Realm searcher
    Formerly known as "Bernal"

    Comment

    • Ryumon
      Member
      • Apr 2007
      • 1801

      #32
      Originally posted by Jundo

      Rather, the A.I. continues to be creative by, for example, (1) finding relationships among the existing data that we miss, for example, seeing relationships among medical studies buried in obscure journals, and "hit and miss" chemical experiments in hours that might take years for human beings, and (2) random "mutations" that produce something useful, e.g., A.I. generators produce random images with mistakes, but some of those mistakes may be valuable much as nature's mistake/mutations are the seeds of natural selection. Perhaps A.I. can discover new physics, new chemistry, and even new ways to engineer better A.I. in this way ... all of which seems to "work" and be functional, but in ways we can barely understand.
      AI tools are doing amazing things with data mining, discovering potential antibiotics, working with medical diagnostics, etc. This isn't technically AI - not what we think of with an LLM - but machine learning. This, to me, is one of the areas where this technology may provide the most important results.

      As for random images with mistakes, the risk is that these mistakes are reinforced in LLMs, rather than being weeded out by natural selection. There is no "selection," other than human correction. it's not like images of people with six fingers on each hand will not have offspring, leading to that trait becoming rarer.

      But in physics and chemistry and biology and medicine; these are areas where these tools are going to come up with breakthroughs. I think it's wrong, however, to put them on the same level as generative AI LLM tools. Machine learning is very different from chatbots that use a massive dataset to answer questions. (ie, gen AI uses machine learning to refine its dataset, but ML doesn't use gen AI to provide answers to questions.)

      Gassho,

      Ryūmon (Kirk)

      Sat Lah
      I know nothing.

      Comment

      • Jundo
        Treeleaf Founder and Priest
        • Apr 2006
        • 40478

        #33
        Someone asked what Nishijima Roshi would think about all this. That is hard to say. Personally, I think that Nishijima, a free thinking man very innovative about the future of Buddhism, including his emphasis on new technology and modern medical understanding of Zazen, and his great interest in the future of Buddhism and the world to come, would love this. LINK I found a few places where Nishijima was very much open to the future, and to science and technology confirming our Zen Practice, for example LINK and also:
        .

        .

        I think he would be very positive about it.

        Gassho, Jundo
        stlah
        ALL OF LIFE IS OUR TEMPLE

        Comment

        Working...