@alexstandiford my take: in the race to train on more and more data, the bar has dropped. Scrape more and more data, quality be damned.
It’s now also being trained on some of the very same bullshit it previously regurgitated.
Garbage in, garbage out, or:
> We home-school 'em. I teach the big ones and the big ones teach the little ones. But no one ever taught me, which makes the whole thing just an exercise in futility.