Google DeepMind’s new AI agent learned to play several video games – including no man’s sky, ValheimAnd goat simulator 3 – To become a viable “interactive gaming companion”.
The new agent tool, SIMA 2, builds on its earlier version SIMA (Scalable Instructional Multiworld Agent), which DeepMind released in March 2024. It also features Google’s Gemini AI for the first time, meaning the agent can go from following instructions to “understanding the user’s high-level goal, performing complex reasoning in pursuit, and efficiently executing goal-oriented tasks within the game” in ways that have never been seen before, according to the DeepMind blog. Post. It is currently being released as a limited research preview to some academics and developers.
DeepMind team members explained that despite SIMA 2’s gaming potential, creating a consumer-facing gaming helper is not the overarching goal here. The Verge During Wednesday’s briefing. Jane Wang, a senior staff research scientist at DeepMind, called it “really a great training ground” for potentially transferring the skills to a real-world environment one day.
And, as always, it all comes back to the ever-expanding AGI race between Google, Meta, OpenAI, Anthropic, and others. “This is an important step towards Artificial General Intelligence (AGI), which has important implications for the future of robotics and AI-avatars in general,” DeepMind’s blog post said.
Joe Marino, a research scientist at DeepMind, echoed this, saying that SIMA 2’s ability to take action in virtual worlds and handle environments it has never seen before is a “fundamental” step toward AGI — and potentially toward building a general-purpose robot.