Google DeepMind’s new AI agent has learned how to play a number of video games. no man’s sky, valheimand goat simulator 3 — to become a viable “interactive gaming companion.”
The new agent tool, SIMA 2, is built on the previous version, SIMA (Scalable Instructable Multiworld Agent), which DeepMind released in March 2024. It also incorporates Google’s Gemini AI for the first time, meaning the agent will be able to go beyond simply following instructions and “understand the user’s high-level goals, perform complex reasoning in pursuit of them, and deftly execute goal-directed actions in-game,” according to DeepMind’s blog post. Currently released to select academics and developers as a limited research preview.
Despite SIMA 2’s gaming capabilities, creating a consumer-grade gaming helper is not the broader goal here, DeepMind team members said. The Verge During Wednesday’s briefing. Jane Wang, a senior researcher at DeepMind, described it as a “really great training ground” for potentially applying the skills to real-world environments one day.
And as always, it all comes back to the growing AGI race between Google, Meta, OpenAI, Anthropic, and others. “This is an important step in the direction of artificial general intelligence (AGI) and has important implications for the future of robotics and AI incarnations in general,” a DeepMind blog post states.
Joe Marino, a research scientist at DeepMind, reinforced that point, saying SIMA 2’s ability to take action in virtual worlds and handle environments never seen before is a “fundamental” step toward AGI and potentially toward building general-purpose robots in the future.
