A panel discussion on bias in AI is introduced by Michael DEA, with panelists Dennis, Mure, and Kim sharing their experiences and expertise related to AI governance. The discussion explores complexities of bias in AI, emphasizing the importance of clean data, challenges in addressing bias at various levels of AI systems, and the implications of bias in areas such as business, revenue, governance, criminal justice, healthcare, and linguistic diversity.
In the video transcript, a panel discussion on bias in AI is introduced by Michael DEA, the director of the Oslo for AI project. He asks the panelists to raise their hands in response to questions about intelligence, consensus on potential harms of AI systems, and existing models of governance capable of regulating emerging technology. The panelists, including Dennis, Mure, and Kim, share their backgrounds and experiences related to AI governance.
Dennis, a former CIA political analyst, talks about his interest in AI governance and the challenge of synthesizing vast amounts of writing on AI. Mure, from Latimer, discusses the company’s large language model trained on black history and culture to reduce bias. Kim, a UX and design researcher with expertise in computational rhetoric, highlights the need for new ontologies in discussing AI to move away from traditional biases and limitations in data discussions.
The discussion delves into the complexities of bias in AI, including data bias, human bias, and algorithmic bias. The panelists emphasize the importance of clean data and the challenges in addressing bias at various levels of AI systems. Mure explains how Latimer is working to include diverse data sources and develop bias detection tools to help companies mitigate bias in large language models.
The conversation also touches on the implications of bias in AI, such as its impact on business, revenue, governance, criminal justice, and healthcare. Kim and Mure discuss a project focused on preserving endangered languages using AI and the potential for AI to address linguistic diversity. The panel concludes by posing a critical question about AI governance: who designs intelligent systems and who has the authority to decide how they operate, emphasizing the need for thoughtful consideration of governance in AI development.