A blind test for human-generated (Wikipedia) vs ChatGPT-generated text on their competence, trustworthiness, clarity, and engagement.
Result: No significant difference.
Not answered: Is the result the same for texts containing hallucinated facts?
Follow
@redfrog How about the extremely high probability that the ChatGPT text was actually strongly influenced by the corresponding Wikipedia article? If this might be true, a related hypothesis would be that ChatGPT would not be able to stand up to the quality for texts that have been generated after its scraping cut-off point.
