AI Safety Breakfasts with Rumman Chowdhury: Navigating the Complexities

In November 2024, Rumman Chowdhury took the stage during an AI Safety Breakfast in Paris, a part of a series leading up to the AI Action Summit in February 2025. The gathering, designed to energize discussions on AI safety, brought together experts and enthusiasts from English and French-speaking spheres.

Algorithmic Auditing: A Detective's Approach

Chowdhury articulated the need for a shift in how algorithmic auditing is perceived and conducted. Rather than traditional methods, she advocates for an approach akin to detective work. Why? Because identifying algorithmic issues requires understanding not just the technical output, but the broader social context in which these algorithms operate. In a rapidly evolving technological landscape, could this investigative method become the new norm?

Chowdhury's stance is clear: the existence of algorithmic bias or malfunction is merely the starting point. Tracing it back to its roots involves a mix of technical and social skills, qualities akin to those of a seasoned detective. This line of thinking challenges the status quo and questions if the traditional academic approach is adequate today.

The Question of Global Governance

As AI systems continue to gain autonomy, Chowdhury raises a critical point about global governance. She suggests that a truly independent global body is necessary to oversee AI. But how to ensure such an entity remains unbiased and free from corporate or governmental influence? The answer may lie in innovative governance models, like the IPCC for climate change, where scientists leave national affiliations at the door.

However, the challenge remains substantial. The complexities aren't just financial but geopolitical. How can we create a truly apolitical and effective governance structure in a world where economic interests often outweigh ethical considerations? Chowdhury suggests a focused approach, where the governance body is limited in scope but strong in its independence.

Impact on Society and Education

Chowdhury also touches on the societal impact of AI, emphasizing the need for critical thinking and discernment. The overreliance on technologies like navigation systems or AI-generated text could lead to a loss of fundamental capacities, such as spatial awareness or logical reasoning. Are we on the brink of creating a generation that may lack these essential skills?

There's a call to action for educational systems to adapt, ensuring that while students use AI tools, they also develop the critical skills necessary to engage with these technologies responsibly. It raises the question: as AI becomes ubiquitous, how can societies balance technological convenience with the preservation of human agency?

The discussions at the AI Safety Breakfasts underscore a pressing narrative: the governance and ethical deployment of AI aren't just technical challenges but societal ones that demand collaborative and innovative solutions.