Post #1632736

Except that OpenAI, Anthropic et al. have not been shy about the fact that they’ve been trying their best to reduce the amount of “hallucinations” that their models produce. It shouldn’t be surprising at all that ChatGPT confessed to a crime it did not commit.