In this episode of the Google DeepMind podcast, Professor Hannah Fry interviews Shlomi Fruchter and Jack Parker-Holder, the creators of Genie 3, a new interactive world model. They discuss Genie 3's capabilities, such as generating diverse and visually interesting worlds from text prompts or images in real-time, and its potential applications in agent simulation, planning, education, and entertainment. The discussion covers the differences between Genie 3 and video generation models like Veo, the emergent properties observed during its development, and the safety considerations for this technology. The podcast further explores the potential of Genie 3 to serve as a foundation model for simulated worlds, similar to how LLMs function for language, and its implications for achieving embodied AGI.
Sign in to continue reading, translating and more.
Continue