GPT-3 was prior to ChatGPT. Since it was a non-instruction tuned model, the outp... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		ipsum2 on Nov 17, 2023 \| parent \| context \| favorite \| on: What Meta learned from Galactica, the doomed model GPT-3 was prior to ChatGPT. Since it was a non-instruction tuned model, the output was quite bad unless prompted precisely.

danielbln on Nov 17, 2023 [–]

GPT-3 was instruct finetuned a while before ChatGPT was even released (see InstructGPT paper and announcement). What ChatGPT went through is RLHF to align its responses to a more conversational level, but you could give GPT-3 (davinci) instructions long before ChatGPT.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact