Comment by lxglv

weird to learn that they do not generate inference requests to their network themselves to motivate early adopters at least to host their inference software