• There is only a nature article about "low carbon emission" (Was: Holy S

    From Mild Shock@21:1/5 to So I on Mon Sep 2 19:02:47 2024
    I didn't find yet a paper that proofs "cheapness"
    related to humans, only a paper about "low carbon
    emission" related to humans:

    The carbon emissions of writing and illustrating
    are lower for AI than for humans https://www.nature.com/articles/s41598-024-54271-x

    And that costs go down is mentioned only
    in a vague tweet, even not related to humans:

    Perplexity CEO Aravind Srinivas says that the cost per
    query in AI models has decreased by 100x in the past
    2 years and quality will improve as hallucinations
    decrease 10x per year https://twitter.com/tsarnick/status/1830045611036721254

    So I wrote:

    Disclaimer: Can't verify the later claim... need to find a paper.

    "low carbon emission" can be an indicative of low
    price. But not necessarely. It could be also an
    indicative of "green" production of AI. And maybe

    the problem is that humans are not that "green",
    it could be easier to make a computing center "green",
    than a city full of humans. Also the nature

    article could be some fake news propaganda.

    So still waiting for more information...

    In some periods in history,
    the price of oil is such,
    that people pay
    to have it hauled away.

    So, it's not so much that AI is cheaper than people,
    though it is a great source of advantange,
    as that there's a great embarrassment of riches
    of availability of computing resources,
    what's set the price point so low.

    Then, the idea is that people can afford their own agents,
    on their own computing resources, not so much as that
    it's cheap for the industry to offer that as a service,
    to anybody, because the agent belongs to them,
    and it's inscrutable and so on and worthless,
    in terms of what value it takes.

    So, "the cost", is plenty high, "unregulated AI".



    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)