> Every machine learning application ever involves both training and inference. Not really, no. There are a lot of "classical" models that you don't want to up

> Every machine learning application ever involves both training and inference. Not really, no. There are a lot of "classical" models that you don't want to update on the fly. I'd even argue that it's a minority.