Google's new AI agent DeepMind has learned to play many video games, including No Man's Sky, ValheimAnd Goat Simulator 3 – become a viable “interactive gaming companion.”
The new SIMA 2 agent tool is based on the earlier version of SIMA (Scalable Instructable Multiworld Agent), which DeepMind released in March 2024. It also includes Google's Gemini artificial intelligence for the first time, meaning the agent can go beyond simply following user instructions to “understand the user's high-level goals, perform complex reasoning to achieve, and skillfully execute goal-directed actions in games,” even ones it hasn't seen before, according to the DeepMind blog. fast. It is currently available to some scientists and developers as a limited preview research version.
Despite SIMA 2's gaming prowess, creating a consumer-focused gaming assistant is not the broader goal here, DeepMind team members said. Edge during a briefing Wednesday. Jane Wang, a senior scientist at DeepMind, called it “a really great training ground” for potentially transferring skills to a real-world environment one day.
And, as usual, it all comes back to the ever-intensifying AGI race between Google, Meta, OpenAI, Anthropic and others. “This is a significant step towards artificial general intelligence (AGI), with important implications for the future of robotics and AI implementations in general,” DeepMind said in a blog post.
Joe Marino, a research scientist at DeepMind, confirmed this, saying that SIMA 2's ability to perform actions in the virtual world and manipulate environments it has never seen before is a “fundamental” step towards AGI – and perhaps a general-purpose robot in the future.
