Jump to content

Talk:GPT-4

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia


JAMA pediatric study[edit]

This study used GPT 3.5, not 4. The study (technical it was a research letter, not a peer-reviewed article) was widely discredited because the prompt was simply, "List a differential diagnosis and a final diagnosis" and the case information. Given that no attempt at prompt engineering was made and the old GPT model was used (6 months after the release of GPT4), it appears more like a smear attempt. 170.85.70.173 (talk) 16:06, 13 June 2024 (UTC)[reply]