Dont Try To Fool AI Detectors

This topic is: resolved

 

Thank you for contacting me. Please note that I live in the GMT+3 time zone - responses might be delayed by this.

This topic has 3 replies, 2 voices, and was last updated 9 months, 4 weeks ago by Szabi – CodeRevolution.

Viewing 3 reply threads
  • Author
    Posts

    • January 29, 2024 at 9:10 am #9779


      pagenet
      Participant
      Post count: 4

      I know I’m asking alot of questions today .. Sorry in advance.

      “Don’t Try To Fool AI Detectors (Disable Content Tricks)”
      Q1.  What exactly does this setting do?

      I noticed non-ascii characters (aka Diacritics) in the resulting output.
      Q2:  Am I correct this is how you are throwing off the detectors?

      Q3  Is that all that this setting does???  Or .. is there more to it?  .. that effects the actual order/pattern of the words being generated through creative AI prompting or word spinning?

      MY PERSONAL THOUGHTS
      Non-ascii characters might throw off today’s AI detectors, but Google will likely strip these out in the long-run and may even penalize their usage if their policy so chooses.

      I believe you are 100% right to try to throw off the AI sniffhounds, but we need to find a way that doesn’t leave a footprint that screams “we are cheating on purpose”.   There is no doubt in my mind that google will one day penalize websites that are made up of 90% or more AI content .. if not de-index them altogether.

      Today they say they don’t care, but of course they do.  They will tolerate some AI from websites, but I’m sure there they are going to measure what percentage of a website is auto-generated by AI.

      If there are other things you are doing to throw off the scent, how can we disable the non-ascii characters part of it?   Unfortunately, I do believe that one aspect will leave too much of a footprint for Google to eventually pick up on.

    • January 29, 2024 at 10:43 am #9782


      Szabi – CodeRevolution
      Keymaster
      Post count: 4577

      Hello,

      First of all, thank you for your purchase.

      Yes, currently the plugin is using the non-ascii characters to trick AI detectors. Currently, this is the only method for this. I experimented also with other methods, like adding intentional mistakes in the content, double spaces, wrongly written words or non-capitalized words. However, this method was removed, as customers did not like to see these kind of mistakes in their content.

      I am still researching this subject, looking into alternatives for this. I think the best option would be if I would build a separate API which would handle this task.

      So, if you don’t need the non-ascii characters, you can disable this feature of the plugin.

      Regards, Szabi – CodeRevolution.

    • January 29, 2024 at 12:53 pm #9783


      pagenet
      Participant
      Post count: 4

      I’ve only been trying for a couple days, but I found you can fool 1/2 the detectors with creative prompting alone .. when you add a spinner to it, you can fool most all of them .. WITHOUT leaving a footprint that makes it obvious you are trying to bypass detection.

      SpinRewriter’s humanize function is a bit weak .. but it does help.   You can add grammar mistakes, but I found that doesn’t help the detections scores as much as you think it should.

      I found Claude 2 to be the best at passing detection, especially when you add some creative prompting and a spinner .. but believe it or not, creative “write as a human” prompts alone appear to do most of the bypass.

      WORKFLOW
      Creative “Humanized” Prompt -> Claude 2 -> Spinrewriter Synonyms -> Spinrewriter Humanize function -> Then Fix Grammar

      So you can have a perfect grammar and still pass as human on just about every free detector (@ http://www.undetectable.ai).   Not sure about Originality or Wilson.

      I found it also works pretty well on OpenRouter LLMs .. even the free ones … This formula works right now, but who knows in the future.

      This WHITEPAPER seems to have a way to solve it for good though.
      https://arxiv.org/abs/2305.10847v5
      https://arxiv.org/html/2305.10847v5

      Python Github is here
      https://github.com/ColinLu50/Evade-GPT-Detector

      I don’t know much about Python, but if you do, you might want to check it out and see if you can convert it to PHP for your plugin, or better yet .. sell it as an addon rather than another API.   API costs are prohibitive to scaling   ;

      Keep at it. .. we’re getting close.

    • January 29, 2024 at 3:40 pm #9784


      Szabi – CodeRevolution
      Keymaster
      Post count: 4577

      Thank you for this info, I really appreciate it.

      If you have more info to share in the future, you can also contact me by email: kisded@yahoo.com

      Will work on this.

      Regards.

  • Author
    Posts
Viewing 3 reply threads

The topic ‘Dont Try To Fool AI Detectors’ is closed to new replies.