Meet SIMA, The 'AI Gamer' From Google Who Can Learn To Play Like A Human. Here's How It Works

SIMA does have more use than just gaming, especially as an AI tool that can understand instructions quickly and clearly.  

Gamers have been no strangers to AI, having been playing against it in almost every game from FIFA to Call of Duty. Now, Google has come up with an AI that can learn how to play games like a human. 

Google DeepMind, the tech giant's AI research firm, has introduced SIMA, an innovative artificial intelligence (AI) agent poised to revolutionise the gaming landscape. SIMA, short for 'Scalable, Instructable, Multiworld Agent', marks a significant advancement in AI capabilities, boasting the ability to learn and adapt gaming skills from diverse environments.

SIMA's prowess lies in its capacity to assimilate gaming skills through natural language instruction and image recognition, as detailed in a recent blog post by Google. Unlike traditional gaming bots, SIMA emulates human-like gameplay, responding to instructions and navigating game scenarios with agility and precision.

High Score Not Primary Goal

Presently, SIMA has honed over 600 fundamental gaming skills, spanning tasks such as navigation, object interaction, and vehicle piloting. Google emphasises that while achieving high game scores is not the primary goal, SIMA's ability to learn and adapt across various gaming environments holds promise for the development of highly capable AI agents.

Tried And Tested

SIMA has been tested on several games, including the likes of No Man's Sky and Goat Simulator

To bolster SIMA's capabilities, Google has partnered with eight renowned game studios, granting the AI agent access to a plethora of gaming experiences. SIMA has undergone trials across nine different games, including popular titles like No Man's Sky and Goat Simulator 3, enabling it to acquire essential skills such as resource management and navigation.

Why Does The World Need SIMA?

Well, if we look beyond the "learning how to game" aspect, SIMA does have more use, especially as an AI tool that can understand instructions quickly and clearly.  

Google's collaboration extends to four research environments, including a unique Unity engine environment where SIMA was challenged to sculpt, showcasing its comprehension and adaptability. Additionally, experiments involving human players have been conducted to refine SIMA's predictive capabilities, with one player engaging in gameplay while the other provided instructions to the AI agent.

Google envisions a future where advanced AI models like SIMA will execute complex gaming tasks with human-like proficiency, offering users invaluable support and enhancing gaming experiences.

Here's what Google said in its blog:

SIMA’s results show the potential to develop a new wave of generalist, language-driven AI agents. This is early-stage research and we look forward to further building on SIMA across more training environments and incorporating more capable models. As we expose SIMA to more training worlds, the more generalisable and versatile we expect it to become. And with more advanced models, we hope to improve SIMA’s understanding and ability to act on higher-level language instructions to achieve more complex goals. Ultimately, our research is building towards more general AI systems and agents that can understand and safely carry out a wide range of tasks in a way that is helpful to people online and in the real world.