Google DeepMind has pushed the boundaries of artificial intelligence (AI) with the introduction of its latest model, Scalable Instructable Multiworld Agent (SIMA). This groundbreaking AI model is designed to play 3D video games just like a human player would, marking a significant step forward in the realm of AI gaming.
To create a learning environment for SIMA, Google DeepMind collaborated with eight game studios and exposed the AI model to nine different video games. These games, including popular titles like No Man’s Sky, Teardown, Goat Simulator 3, and Valheim, provided a diverse set of interactive worlds for the AI model to navigate through and interact with various objects.
Complex Tasks and Evaluations
The current version of SIMA has been evaluated on over 600 basic skills, ranging from simple navigation tasks like turning left and driving a car to more complex object interactions such as climbing a ladder and crafting a helmet. Google highlighted the importance of training the AI model on a diverse set of 3D video games, emphasizing the impact it could have on future AI capabilities.
While the focus is currently on training SIMA with more complex instructions that require high-level strategic planning and multiple sub-tasks to complete, Google hints at the potential real-world applications of the AI model. By teaching SIMA to follow human instructions and act accordingly in virtual environments, Google aims to explore ways in which this technology can be beneficial to humans in various tasks and scenarios.
Overall, Google’s DeepMind has taken a significant stride in the field of artificial intelligence with the development of SIMA. By bridging the gap between AI and 3D gaming, the possibilities for this technology are endless. From learning how to navigate through virtual worlds to understanding human interactions, SIMA represents a crucial advancement in the quest to create more intelligent and adaptable AI systems.
Leave a Reply