Google Deepmind's new AI agent 'SIMA' will be your in-game teammate: Check details
Google DeepMind has introduced SIMA, an AI agent designed to develop gaming skills that align more closely with human gameplay instead of being too strong and doing its own thing. SIMA, which stands for Scalable, Instructable, Multiworld Agent, is still in the research stage.
SIMA aims to eventually master playing any video game, including those with no linear path to end the game and open-world games.
Also read: Google Gemini vs Microsoft Copilot: Clash of AI chatbots in 5 different scenarios
While not meant to replace existing game AI, SIMA will be more like an additional player who blends seamlessly with your group.
It combines natural language commands with the ability to understand 3D environments and image recognition.
Also read: Google announces these 4 updates at Games Developer Summit 2024 for better gaming experience
“SIMA isn’t trained to win a game; it’s trained to run it and do what it’s told,” Google DeepMind researcher and SIMA co-lead Tim Harley said during a briefing, reports The Verge.
Google collaborated with eight game developers, including Hello Games, Embracer, Tuxedo Labs, Coffee Stain, and others, to train and evaluate SIMA. Researchers immersed SIMA into games such as No Man’s Sky, Teardown, Valheim, and Goat Simulator 3 to familiarise the AI agent with the fundamental aspects of gameplay.
According to Google, SIMA doesn’t require a custom API to play games or access source codes.
Harley mentioned that the team opted for games with a focus on open-ended gameplay rather than a structured narrative. If you’ve experienced Goat Simulator, you understand that its essence lies in random, spontaneous actions. Harley stated it was this kind of spontaneity they hoped the AI agent would learn.
Currently, SIMA has approximately 600 basic skills, such as turning left, climbing ladders, and accessing the menu to use a map. According to Harley, SIMA could eventually be trained to execute more complex functions within a game. However, tasks like “find resources and build a camp” remain challenging because AI agents cannot carry out actions on behalf of humans.
SIMA is not designed to function as an AI-powered NPC similar to those developed by Nvidia and Convai. Instead, it will operate as another player within a game that impacts the result.
