It’s not true that it’s an inherent limitation of LLMs though. OpenAI just decid... | Hacker News

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

		circuit10 on May 9, 2023 \| parent \| context \| favorite \| on: LLMs are not greater than the sum of their parts: ... It’s not true that it’s an inherent limitation of LLMs though. OpenAI just decided that it was too risky to have ChatGPT give opinions or express preferences or feelings

SequoiaHope on May 9, 2023 [–]

I don’t think that’s the only reason they decided to use RLHF. I think the raw model without RLHF would just fail differently, rather than not failing.

circuit10 on May 9, 2023 | [–]

It’s possible to do RLHF without training that out

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