no, there is a lot of research h that shows large foundation models display emergent behavior that go beyond their training data