Saw this posted by “war and peas” on mastodon.

    • phonics@lemmy.world
      link
      fedilink
      arrow-up
      7
      ·
      18 小时前

      I had the same thought then saw it was war and peas and assumed otherwise. But i don’t feel like your wrong to assume that.

      • ComradeMiao@lemmy.world
        link
        fedilink
        English
        arrow-up
        9
        arrow-down
        1
        ·
        18 小时前

        Thanks.

        that’s not an accusation. That’s the business model .

        is exactly how Deepseek writes in my experience

        • AmyAye@nord.pub
          link
          fedilink
          English
          arrow-up
          12
          ·
          16 小时前

          It’s almost like these systems are trained on human writing and are literally just spicy autocomplete.

          • stateofaffairs@lemmy.today
            link
            fedilink
            arrow-up
            6
            arrow-down
            1
            ·
            15 小时前

            Spicy autocomplete. I like it.

            Avoiding AI accusations drove me nuts. I changed how I wrote. I tried to keep EM dashes and “if not this, then that” statements. But it forced me to change my writing to avoid sounding like AI.

            AI has improved and I am again deconstructing my writing. Not to get better, but to avoid klanker accusations.

            That is a new problem for writers.

        • prole@lemmy.blahaj.zone
          link
          fedilink
          arrow-up
          5
          arrow-down
          1
          ·
          14 小时前

          Yeah but that’s also how I write. Should I change the way I write, specifically dumb it down, so people don’t mistakenly assume I’m using an LLM? It’s fucking bullshit.

          • redjard@reddthat.com
            link
            fedilink
            arrow-up
            1
            ·
            14 小时前

            The actual tell is using it improperly. It’s meant to clear up plausible confusion, or at least meant o be used with words that are incompatible. But like in this example they are literally accusing Etsy of those “business practices”. So the “That’s not” is not fitting.
            When you use the device correctly it doesn’t carry the same AI feel to it.

          • ComradeMiao@lemmy.world
            link
            fedilink
            English
            arrow-up
            2
            arrow-down
            3
            ·
            14 小时前

            specifically dumb it down

            If you think that phrasing is not dumbed down that speaks volumes.

        • redjard@reddthat.com
          link
          fedilink
          arrow-up
          4
          ·
          18 小时前

          Also “Legally, that [changes nothing]. Emotionally, [it does].”. The “totally” and “may” aren’t something I’d expect ai to do though.
          I hope it’s just an unfortunate choice of phrases.
          We’re definitely at the point where people are abandoning devices they used to use to not sound like ai.

          • Simon_Shitewood@lemmy.ml
            link
            fedilink
            arrow-up
            4
            ·
            14 小时前

            It’s not even unfortunate phrasing, it’s normal phrasing that normal people use, which is the exact reason AI uses it. Your accusations are just Kafka traps where talking like a normal human being is held up as proof they’re an AI.

            • redjard@reddthat.com
              link
              fedilink
              arrow-up
              3
              ·
              14 小时前

              I wasn’t accusing anyone, and @ComradeMiao@lemmy.world seems to have been convinced it’s not ai too. If they were even doubting it to begin with.

              The thread is about why it feels like ai, it seems pretty clear that it is not actually ai.

              Stuff feeling like ai comes about naturally due to the way that it is. “Talking about it is good actually” cause then you learn when to trust it and when not to, and learn to understand it better, and how others perceive it.
              It’s a fact that using certain expressions in certain ways will now raise eyebrows.

              Btw. specifically the “it’s not X it’s Y” one is conditioned on using it poorly (at least for me). I’m sure there are people that legitimately used it poorly before ai, like presumably war and peas here, but it’s definitely not common.
              I mean if it was it wouldn’t be as strong a tell.
              My “<adjective>, not X. <adjective>, Y” in comparison is a weaker tell. Maybe because it is used properly, or more commonly. Who knows.

              Also ai doesn’t talk like normal people. It doesn’t even talk like professional writers, though it’s closer to modern pre-ai news article writing than most other styles.
              I’m sure it would have the ability to get somewhat close to other styles too, since in initial training the llm will have been exposed to it all and “have to have used it”. But it is trained not to later on, and people generally don’t bother jailbreaking their ai to get it to talk more like a normal person.
              Hence also “totally” just by itself being a strong no-ai tell.

              • ComradeMiao@lemmy.world
                link
                fedilink
                English
                arrow-up
                2
                ·
                14 小时前

                The thread is about why it feels like ai, it seems pretty clear that it is not actually ai.

                Stuff feeling like ai comes about naturally due to the way that it is. “Talking about it is good actually” cause then you learn when to trust it and when not to, and learn to understand it better, and how others perceive it. It’s a fact that using certain expressions in certain ways will now raise eyebrows.

                Thanks for a wonderful addition to this thread. Great point!