Machine Learning Experts - Margaret Mitchell

Hugging Face · 4 years ago · 31

Margaret Mitchell, who founded Google's Ethical AI Group, discussed her work on ethical artificial intelligence and bias in machine learning systems during an interview at Hugging Face. While working on Microsoft's Seeing AI application, Mitchell realized that training data reflects skewed representations of the world—such as describing white people as "people" and Black people as "Black people"—and that models can learn dangerous associations, like interpreting disaster imagery positively based on color patterns. She now focuses on making machine learning teams aware of data bias through better education on concepts like marginalization and inclusion, and on creating standardized documentation tools like Model Cards to encourage ethical evaluation of AI systems.