Interesting article by Alan Kohler with much that could be discussed.

One aspect, for example, Morality.

… The New York Times revealed last week that Anthropic had been trying to instil morality into its AI model, Claude…

Christopher Olah, one of the co-founders of Anthropic, had been hosting private meetings with religious thinkers from around the world, with non-disclosure agreements and demands of confidentiality.

The discussions reportedly served two purposes: “to broach the possibility of AI consciousness … and to learn how Anthropic”, which is preparing to float for more than $US2 trillion ($2.87 trillion), “might apply centuries of human moral wisdom to its models, as rapidly as possible”.

Theologist David Decosimo posted, after reading the NYT piece: “There is so much here that’s deeply troubling: That at the very highest levels Anthropic believes Claude is a living, suffering creature with moral standing, that they’re treating & training it accordingly, that they’re actively trying to convert religious leaders to their faith.”

Yes, it is troubling, but the people at Anthropic know Claude better than he does.

Anyway, it’s becoming clear that AI agents are already making their own decisions, so trying to persuade them to act morally is worth a try.

  • arbilp3@aussie.zoneOP
    link
    fedilink
    English
    arrow-up
    1
    arrow-down
    4
    ·
    1 month ago

    Crikey, there’s plenty of folks about who don’t seem to have the ability to think either 😆

    • foofighter@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      1 month ago

      Argument v Insult. I hope you come around to engaging with some more critical takes at “AI” in the near future. There’s a lot of marketing and PR rubbish from the industry that’s clouding otherwise smart people’s judgement

      • arbilp3@aussie.zoneOP
        link
        fedilink
        English
        arrow-up
        1
        arrow-down
        2
        ·
        1 month ago

        I am not technically-oriented. Philosophical discussion is more of an interest to me. I don’t go for marketing and PR rubbish of any kind if I can help it.

        Some of the points that Kohler raises are thought-provoking and I am not sure if you are saying his judgement is being clouded. I found his little experiment of asking AI models: “What do you think of the proposition that AI is a new form of life?” very instructive.

        • foofighter@lemmy.world
          link
          fedilink
          English
          arrow-up
          3
          ·
          1 month ago

          I’ll try to explain how I understand LLMs as best I can because I can appreciate your interest in the philosophical side of things.

          An LLM breaks up what input you give it, assigns numbers and categories to the words given, and then responds by individually ranking words with a rating on the probability that it fits within the scope of the prompt. If I take Kohler’s question, the LLM break it down as if it is assigning individual unique phone numbers to each part of the question, like this:

          “What+614392XXXXX | do you think+6145782XXXXX | of the proposition that+614467XXXXX | AI is a new form of life?+614823XXXXX”

          Imagine these phone numbers resolve to the LLM’s training data. For a question like this one, it’s going to start calling around to see if all of these phone numbers have shown up somewhere in it’s training data and inevitably for a question about AI sentience, it’s going to stumble upon literature from works like Isaac Asimov’s, the script of 2001: A Space Odyssey, and every comment ever made on the internet in discussions about AI. It’s still not thinking though. Because it’s inevitable that science fiction like Asimov, 2001 and the online debate surrounding this genre of media take up a lot of space in the training data, it’s response, which is a next word probability algorithm, is going to come up with something intriguing - but it doesn’t actually see the words it just sees the phone numbers and the ranking against the vibe of it’s training data:

          +614392XXXXX174541 | +6145782XXXXX16431123 | +614467XXXXX113466 | +614823XXXXX142231298

          Some people online maybe on YouTube could absolutely explain this better than I have, but the world’s a mess that’ll only get messier the less people talk to each other directly. Technology aside, the economics of the five big companies at the top is very interesting and absolutely an area I’m not especially oriented in, so I lean on journalists like Ed Zitron to break down those concepts. I recommend you check out what he’s written about this it even if he does get a bit angry sometimes