3 ms·
And I'm saying that any task that wasn't known in advance immediately becomes a task known in advance once it's done once. I'm not arguing that there is no pla
by lolinder 3y ago
And I'm saying that any task that wasn't known in advance immediately becomes a task known in advance once it's done once.
I'm not arguing that there is no place for large, general models—they're great for exploration—just that a smaller foundation model shouldn't be dismissed offhand based solely on parameter count.