Y
Hacker News
new
|
ask
|
show
|
jobs
by
Moru
588 days ago
They would not want to train their next model on the output of the previous one...