Q&A with Margaret Mitchell, an AI ethics researcher at Hugging Face, on a dataset called SHADES designed to test AI models for bias in languages beyond English (Reece Rogers/Wired)