Thorny topics in AI and ML. Evaluation, bias (the statistical kind), bias (the sociological kind), trust and safety.