Demis Hassabis, CEO of Google DeepMind, discusses the progress and future of AI, addressing skepticism about LLMs and the path toward AGI. Hassabis argues that significant headroom remains in current AI techniques, though further breakthroughs in continual learning and memory are needed. He defines AGI as a system exhibiting all human cognitive capabilities, including creativity and physical intelligence, distinguishing it from superintelligence. The conversation explores Google's AI initiatives, including smart glasses, and Hassabis expresses caution regarding integrating advertising into AI assistants due to trust concerns. He also acknowledges Anthropic's advancements in coding models while emphasizing Google's broader focus on multimodal AI and world models, and addresses concerns about an AI bubble, highlighting AI's transformative potential across various sectors.
Sign in to continue reading, translating and more.
Continue
