Keynote: Gemma 4: Compacting Intelligence for the Edge - Léonard Hussenot

This talk explores the philosophy and engineering behind Gemma 4, emphasizing that the future of AI depends on "intelligence per byte" rather than just size. The speaker discusses the importance of compacting intelligence, which maximizes the reasoning and instruction-following ability of each token, as a crucial factor for developing genuinely useful AI. By focusing on token efficiency and minimizing memory footprints, this approach enables the creation of faster, more private, and accessible applications in the field of artificial intelligence.