Model collapse again
Ah ‘Model Collapse’ is rearing its head again. So the problem still stands. I haven’t seen anything recently on ‘Temporal Degradation’ - but they keep releasing new versions.
‘Model collapse’: Scientists warn against letting AI eat its own tail | TechCrunch
If the models continue eating each other’s data, perhaps without even knowing it, they’ll progressively get weirder and dumber until they collapse.

With Meta releasing an open source model that’s better than ChatGPT (although it will be expensive to run as it’s huge) I wonder if any of these risks were tagged by early adopters? Meanwhile Copilot is still not quite there yet.