Pioneers of AI: OpenAI’s Tibo Sottiaux doesn’t want humans to be an ‘afterthought’

OpenAI’s product lead, Thibaut Sottiaux, details the company’s approach to building and deploying frontier AI models like Astra and the new "Dots" agent suite. The discussion centers on integrating safety and alignment into the product development lifecycle, emphasizing that rigorous testing and human-in-the-loop oversight are essential for scaling AI responsibly. Sottiaux explains how AI agents now assist in internal operations, such as monitoring performance and automating back-office tasks, while highlighting the shift toward models that prioritize human utility and well-being. The conversation also addresses the complexities of recursive self-improvement, where AI models contribute to their own architectural and infrastructural advancements under strict human supervision. Ultimately, the focus remains on creating technology that empowers users, reduces stress, and avoids the pitfalls of power concentration, ensuring that AI serves as a collective benefit to humanity.