Y
Hacker News
new
|
ask
|
show
|
jobs
by
anu7df
137 days ago
I understand model output put back into training would be an issue, but if model output is guided by multiple prompts and edited by the author to his/her liking wouldn't that at least be marginally useful?