Hacker News
new
|
past
|
comments
|
ask
|
show
|
jobs
|
submit
login
deevolution
on July 7, 2023
|
parent
|
context
|
favorite
| on:
Experiencing decreased performance with ChatGPT-4
Aren't they using RLHF? The feedback from humans might not always be the ~right~ feedback. Couldn't that possibly degrade the quality of its responses?
Guidelines
|
FAQ
|
Lists
|
API
|
Security
|
Legal
|
Apply to YC
|
Contact
Search:
