Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


What do you use to differentiate AI sites from human sites?
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

What do you use to differentiate AI sites from human sites?

Recently I've looked at a multitude of websites regarding AI services. Lots, if not most, of them are AI generated. It's quite clear for me as I'm bilingual and the AI generated German translation is jarring at best. And in most cases even offputting.
So my question is, what indicators do you personally use to differentiate between AI generated content (websites) and human work?

How do you decide to trust a website with your money today?

Comments

  • I just get a feeling whenever the content is AI generated,

  • tsusutsusu Member

    LLMs are horrible at distilling knowledge because they don't truly understand anything. Most AI generated websites are the result of someone asking an AI to: "Generate a blog post about X."

    They say a whole lot of nothing with a whole lot of words. It's like listening to someone regurgitate what they heard from someone else, but unable to explain why or how that statement came to be.

    Also, most for some reason use GPT4o, and that model is in LLM Hell. It's been finetuned to shit to always follow a formula with its response, almost always in a numbered list, that is easy to recognize if you see it enough.

    Also, there is something known as GPTisms where certain phrases are highly probable in situations where humans would not use them. The most famous example is probably "multi-dimensional," "delve," and "tapestry" for GPT4o.

    Thanked by 1cainyxues
  • jarjar Patron Provider, Top Host, Veteran

    Brain

  • You can just see it, feel it & know it. BTW mostly it's about the images, text translation like you said, large amount of content where not even needed and the designs too sometimes.

  • AI models have a weird writing style. Once you see enough of it, it becomes obvious in most cases.

    Depending on how a person is using them, they might not know the date, so they'll act like it's still 2023 (or whatever their cutoff is). Or they do know the current date, but their info is still old, so they act like a Linux distro released in 2022-2023 is brand new.

    Obvious factual errors that anyone with any knowledge in the field would catch. Requires you to have a brain and knowledge of the topic at hand.

  • gksgks Member

    @tsusu said:
    LLMs are horrible at distilling knowledge because they don't truly understand anything. Most AI generated websites are the result of someone asking an AI to: "Generate a blog post about X."

    They say a whole lot of nothing with a whole lot of words. It's like listening to someone regurgitate what they heard from someone else, but unable to explain why or how that statement came to be.

    Also, most for some reason use GPT4o, and that model is in LLM Hell. It's been finetuned to shit to always follow a formula with its response, almost always in a numbered list, that is easy to recognize if you see it enough.

    Also, there is something known as GPTisms where certain phrases are highly probable in situations where humans would not use them. The most famous example is probably "multi-dimensional," "delve," and "tapestry" for GPT4o.

    AI regenerate content garbage, which is good to doom AI much earlier. In other way human greed destroy AI using AI itselfs. Google SEO will die soon

  • bdlbdl Member

    brain

  • It's not the perfection that makes us human. It's actually the imperfections that make us human.

    We are accustomed/habituated to see imperfections which doesn't follow a predefined pattern and can differentiate.

    AI can not compete that. Even if it tries, they aren't random enough as of yet.

  • EthernetServersEthernetServers Member, Patron Provider

    I can generally sense when something is AI written, but often use https://quillbot.com/ai-content-detector and https://www.zerogpt.com/ to double check that.

  • LeviLevi Member

    @EthernetServers said:
    I can generally sense when something is AI written, but often use https://quillbot.com/ai-content-detector and https://www.zerogpt.com/ to double check that.

    The most hilarious thing is that those tools use AI to detect AI.

  • I often look for a personal touch in the writing, like anecdotes or a genuine conversational tone. Also, I pay attention to how well the content addresses specific questions or concerns, AI can sometimes miss the nuances.

Sign In or Register to comment.