Ensuring Accountability in AI Companies: Why Independent Oversight is Essential
Published on: March 10, 2024
Yoshua Bengio, a pioneering figure in artificial intelligence and co-winner of the 2018 Turing Award, is advocating for a significant shift in the governance of AI companies. He suggests that companies developing advanced AI systems must include independent board members who represent the interests of society.
Speaking against the backdrop of recent boardroom changes at OpenAI, including the temporary departure of CEO Sam Altman, Bengio stresses the importance of a 'democratic process' to oversee AI developments. He raises concerns about the rapid pace of AI advancement and its potential to endanger the public or be misused for power gains.
Bengio, a professor at the University of Montreal and head of Mila, the Quebec Artificial Intelligence Institute, believes that despite recent disruptions at companies like OpenAI, the tech industry is unlikely to slow down AI development. He warns of a focus on competition over public safety and calls for 'proper guardrails' in the field.
The AI expert's concerns extend to the safety risks posed by AI, ranging from disinformation to the development of artificial general intelligence (AGI) that could operate beyond human control. He emphasizes the urgency in establishing robust safety measures before the arrival of AGI.
Bengio criticizes the voluntary agreement between governments and AI companies, believing it favors the companies. He argues that the burden of proving AI systems' safety should lie with the companies, similar to the pharmaceutical industry's responsibility for clinical studies.
He also comments on the timeline for the emergence of uncontrollable AI systems, estimating it could happen within 5 to 20 years. Bengio suggests that governments must be prepared for even a small chance of adverse outcomes with AI technologies.
The AI scientist welcomes Joe Bidenβs executive order on AI, seeing it as a step towards better regulation. However, he notes the need for other countries to implement similar measures to ensure comprehensive global governance of AI technologies.