• P03 Locke@lemmy.dbzer0.com
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 day ago

        LLMs have fully mastered the art of the Turing Test in spades. The layman’s experience with them is just whatever they see on the news or ChatGPT, with easy advice like “jUsT lOok At ThE EmOjiS aNd eMDasHeS”. But, rule files are incredibly easy to make (they are literally just Markdown files), and can tweak the default behavior of a model, including getting rid of behaviors that make it look like a bot.

        Or just give an LLM a large sample size of data, and say “create a rules file to emulate the chat behaviors contained in this file”. We’re talking a 5 minute operation, once you have enough data. And the amount of data can be as little as 25 posts.

      • Hugucinogens@lemmy.blahaj.zone
        link
        fedilink
        arrow-up
        9
        ·
        3 days ago

        For one, we’re way past the necessary level of complexity for bots to do that, many months now. If any even slightly large corporation wants to, they can easily and cheaply spin up chatbots to do that.

        On the other hand, I wouldn’t instantly chalk it up to bots, simply because there’s many many people, who comment backwards reactionary shit they don’t understand, just to feel better themselves. And few things are more satisfying to such people, than dunking on the new-and-alternative thing.

        It’s a very pleasant thing, to feel content that you’ve made the right choices, and don’t need to change anything.