Google Shows Off AI Agent That Helps You Play Games By Watching Player Actions

Google’s new AI model, Gemini 2.0, aims to revolutionize the way we think about AI with its multimodal text, sound, and image generation capabilities. It’s meant to be the next step toward agent-based AI — where bots can perform tasks on behalf of users. It can be used to create AI agents to provide advice and tips in video games.

As Google DeepMind CEO Demis Hassabis and CTO Koray Kavukcuoglu wrote in a company blog, the agents “make inferences about the game based solely on on-screen actions and offer solutions in real time.” They can also use Google’s knowledge base search to provide more complete information. The agents are already being tested in popular games like Clash of Clans and Supercell’s Hay Day.

In addition to helping with video games, Google is also developing other applications for Gemini 2.0. For example, work is underway on the Genie 2 project, which creates virtual game worlds based on just one image. So far, these worlds remain stable for only about a minute, but experts see great potential for the technology.