In this case study, I will be sharing the vast amounts of false positives current AI detection software gives, specifically for this case study I will be demonstrating GPTZero. I personally want to…
  • Lettuce eat lettuceEnglish
    1·
    1 year ago

    Tried GPTZero with some GPT3.5 text and it was very hit or miss.

    Missed a ton of actual GPT texts.

  • BetaDoggo_@lemmy.worldEnglish
    21·
    1 year ago

    The issue with any attempt to identify generated text is that anything generated could have been written by a human. There’s also the issue with model diversity. Even if a specific model has common patterns that can be identified, there are hundreds of other models that will have different patterns.