Episode 374

Google DeepMind is simulating entire worlds using AI - that can be interacted with in real time.

“World models” simulate the environment and physics of the real world. And DeepMind’s Genie 3 model allows people to create these worlds with basic image and text prompts.

The idea is not just to allow people to explore these worlds, but to serve as a testbed for AI agents to learn how to interact with the world before they are deployed in humanoid robotic bodies. 

Could this be the next big step towards artificial general intelligence (AGI)?

Joshua Howgego speaks to Jack Parker Holder, Research Director at Google DeepMind, about the latest developments.

To read more about these stories, visit https://www.newscientist.com/

Learn more about your ad choices. Visit megaphone.fm/adchoices