Ordinary gradient descent will get stuck at a local minimum as opposed to a world minimum, resulting in a subpar network. In usual