Google DeepMind has released a new version of its AI “world” model called Genie 3. This advancement enables the creation of interactive 3D environments where users and AI agents can collaborate in real-time. Genie 3 allows for longer interactions, with users able to explore worlds for up to a minute without losing visual memory of objects.
The new model improves upon previous limitations by offering better interaction capabilities and improved memory retention. Google’s Genie 3 creates immersive video game-like environments using AI-generated assets. With this technology, users can input prompts that allow them to manipulate world events, such as changing weather conditions or introducing new characters.
While the technology is promising, it remains a research preview designed for academics and creators. However, Google plans to expand access to additional testers in the future.
Source: https://www.theverge.com/news/718723/google-ai-genie-3-model-video-game-worlds-real-time