How to Build the Future: Demis Hassabis

Artificial General Intelligence (AGI) development hinges on solving fundamental challenges in continual learning, long-term reasoning, and memory, with agent-based systems serving as the primary architecture for progress. While current large-scale models provide a foundation, achieving AGI requires moving beyond brute-force approaches toward systems that can actively plan and adapt. Distillation remains a critical strategy, enabling smaller, efficient models to maintain near-frontier performance for edge and robotics applications. Beyond language, AI acts as a transformative tool for scientific discovery, exemplified by AlphaFold’s impact on biology. Future breakthroughs will likely emerge from interdisciplinary teams that integrate machine learning with deep tech domains like material science and drug discovery. As AGI timelines converge with long-term research cycles, developers must build systems capable of leveraging specialized tools while maintaining the creative impetus necessary for genuine scientific innovation.