> Because EMX is a new model in machine learning, we do not yet know its usefulness for developing real-world algorithms. So these results might not turn out to have practical importance. But we do now know that we should be careful when introducing new models of learning. Moreover, we might need to look again at the many subtleties that can come up, even in established learning models.

In other words, "Machine Learning" here is a buzzword, without which the paper would probably get less attention.