Recently, Google DeepMind unveiled a groundbreaking artificial intelligence model known as Scalable Instructable Multiworld Agent (SIMA). This AI model is designed to play 3D video games in a manner that mimics human behavior. The primary goal of SIMA is to learn how to interact within various gaming environments while performing a variety of tasks. Although the model is still in the research phase, it is continuously being trained to master more complex movements.
The Objective of SIMA
Contrary to popular belief, the purpose of SIMA is not to create a super-intelligent gamer capable of beating any game. Instead, Google intends to teach the AI model how to navigate through 3D games in open-world scenarios by analyzing human reactions and interactions based on natural-language instructions. Google emphasizes that this is a challenging task that warrants attention.
To provide a diverse learning environment for the AI model, Google DeepMind collaborated with eight game studios and exposed SIMA to nine different video games, including titles like No Man’s Sky, Teardown, Goat Simulator, and Valheim. Each game presented unique interactive worlds that required the AI model to familiarize itself with tasks such as navigating the environment, interacting with objects, and utilizing in-game menus.
In addition to real-life game environments, Google DeepMind developed four research environments, including a bespoke Unity environment called the Construction Lab. Within this setting, SIMA was challenged to construct sculptures using building blocks, testing its object manipulation skills and intuitive grasp of the physical world. The current version of SIMA underwent evaluation across 600 basic skills, ranging from simple navigation tasks to object interactions, all designed to be completed within 10 seconds.
The Impact of Training
Training an AI model on a wide array of 3D video games while adhering to human instructions has the potential to revolutionize the future of gaming. Google DeepMind is now concentrating on enhancing SIMA’s capabilities by introducing more complex tasks that necessitate high-level strategic planning and the completion of multiple sub-tasks. Tasks such as gathering resources and constructing a camp are currently being explored, with the belief that these skills can be applied to real-world scenarios.
The advent of artificial intelligence in the gaming industry, as demonstrated by Google DeepMind’s SIMA model, signifies a significant advancement in the field. By bridging the gap between AI systems and human instructions, the potential applications of such technology are vast and far-reaching. As research and development continue to evolve, the integration of AI in gaming may pave the way for innovative solutions that can benefit society as a whole.
Leave a Reply