Text this: Attacking Deep Learning AI Hardware with Universal Adversarial Perturbation