In a recent interview on the "Machine Learning Experts" podcast, Dr. Margaret Mitchell — renowned AI ethics researcher, former co-lead of Google’s Ethical AI team, and now at Hugging Face — shared insights on her journey, the importance of ethical AI, and how teams can combat harmful bias.
Mitchell, whose background spans linguistics, computational linguistics, and computer science, noted that her realization of ethical AI's importance came during her work on Microsoft’s Seeing AI app. She observed that data sets are inherently skewed: “White people would be described as ‘people’ and black people as ‘black people,’ as if white was default,” she explained. This disparity drove her to focus on fairness and rigorous evaluation.
She emphasized that data ethics is critical in high-stakes applications like criminal justice, hiring, and healthcare, where biased models can have devastating real-world consequences. To increase awareness, Mitchell advises ML teams to invest in diverse data collection, perform disaggregated evaluations, and use tools like Model Cards — a project she helped create to document model performance across different groups.
On diversity, Mitchell argued that inclusion is not just a moral imperative but also improves model accuracy: "Having a more diverse set of people on an ML project results in better outcomes because you catch blind spots early." She also highlighted the need for transparency in decision thresholds and model limitations.
At Hugging Face, Mitchell is developing protocols for ethical AI research, inclusive hiring, and building a positive culture. For those entering AI, her best advice is to focus on understanding data and questioning assumptions.
When asked about the fear of AI taking over, Mitchell dismissed apocalyptic scenarios, stating that current AI lacks general intelligence and consciousness. Instead, she worries about biased systems amplifying societal inequalities. Her favorite ML papers include the original transformer paper and works on data documentation.
Mitchell’s work continues to push the field toward more accountable and inclusive AI, ensuring technology serves everyone equitably.