Creative Red Teaming: Approaches to Addressing Bias and Misuse in Machine Learning