2 ms·

The Information Disorder Level (IDL) Index

• jruohonen 3y ago
  "We worked on this topic at the University of Bergen, which is part of the NORDIS hub. In a “fake news experiment” we asked ChatGPT to generate a corpus of 40 pieces based on fabricated and real events. During the generation process, we observed that despite clear instructions, ChatGPT generated additional fabricated things – regardless of whether the facts introduced in the prompts were true or false."