BLUF

New AI models show improved complexity in answers but also a greater tendency to provide inaccurate information—raising concerns about their reliability.

Summary

  1. The smarter the model, the more inaccuracies.
  2. Newer models falter on simple questions more often.
  3. Participants often failed to spot incorrect AI responses.
  4. Researchers suggest AI should admit uncertainty.
  5. AI companies may resist limiting models' responses.

References