We are in word games here about zero shot learning.
The fact is that with zero shot the models actually aren't learning in the sense of updating their knowledge structure, they are instead using background knowledge as inference. We have started calling it zero-shot learning but... well it isn't really.
The fact is that with zero shot the models actually aren't learning in the sense of updating their knowledge structure, they are instead using background knowledge as inference. We have started calling it zero-shot learning but... well it isn't really.