Google’s up-to-date DeepMind artificial intelligence agent has learned to play multiple video games, including No man’s sky, ValheimAND Goat simulator 3 — to become a real “interactive gaming companion.”
The up-to-date agent tool, SIMA 2, builds on its earlier version of SIMA (Scalable Instructable Multiworld Agent), which DeepMind released in March 2024. It also uses Google Gemini artificial intelligence for the first time, which means the agent can go beyond simply following instructions to “understand the user’s overarching goal, engage in complex reasoning, and skillfully perform goal-oriented actions in games,” even ones it hasn’t seen before, according to the DeepMind blog post. It is currently available to some researchers and developers as a circumscribed preview.
Despite SIMA 2’s gaming capabilities, creating a consumer-facing gaming assistant isn’t the broader goal here, DeepMind team members said Edge during Wednesday’s briefing. Jane Wang, senior research specialist at DeepMind, called it “a really great training ground” that could one day transfer skills to real-world environments.
And as usual, it all comes down to an increasingly 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 the application of artificial intelligence in general,” reads a post on the DeepMind blog.
Joe Marino, a research scientist at DeepMind, underlined this, saying that SIMA 2’s ability to take action in a virtual world and deal with environments it has never seen before is a “fundamental” step towards AGI – and potentially towards building a general-purpose robot.