Re:AI beats human sleuth at finding problematic images in re...
by hirundo on Wednesday October 04, 2023 @07:41PM
(#37773195)
> She found that it contained text plagiarized from eighteen uncredited sources, which she categorized using color-coded highlighting.
That kind of plagiarism is a sign that AI wasn't used, since once tokenized the same concepts are seldom rephrased verbatim. The same problem applies to problematic images. Regenerated-with-alternations is harder to detect than Photoshop. Things will get interesting when you can cheaply generate middle-tier quality scientific papers with a short AI prompt.
That kind of plagiarism is a sign that AI wasn't used, since once tokenized the same concepts are seldom rephrased verbatim. The same problem applies to problematic images. Regenerated-with-alternations is harder to detect than Photoshop. Things will get interesting when you can cheaply generate middle-tier quality scientific papers with a short AI prompt.