Hacker News
new
past
comments
ask
show
jobs
points
by
pupppet
22 hours ago
|
comments
by
fastball
19 hours ago
|
[-]
But that is my point: if benchmaxxing was all the labs were doing, then surely the dumber model could/would have equivalent performance? Rather than noticeably worse perf on a (somewhat trivial to game) test.
reply