Anna Makanju, OpenAI's VP of global affairs, discussed how emerging "reasoning" models like o1 can reduce AI bias by self-identifying biases in their responses. Despite some improvements, o1 still faces challenges like performance issues and high costs.