Google DeepMind’s new AI agent discovered the right way to play a bunch of video video games — together with No Man’s Sky, Valheim, and Goat Simulator 3 — to develop into a viable “interactive gaming companion.”
The brand new agent instrument, SIMA 2, builds on its earlier iteration, SIMA (Scalable Instructable Multiworld Agent), which DeepMind launched in March 2024. It additionally incorporates Google’s Gemini AI for the primary time, that means the agent can transcend merely following directions to “perceive a consumer’s high-level aim, carry out advanced reasoning in pursuit, and assuredly execute goal-oriented actions inside video games,” even ones it hasn’t seen earlier than, in response to a DeepMind weblog submit. It’s presently being launched to some lecturers and builders as a restricted analysis preview.
Regardless of SIMA 2’s gaming prowess, making a consumer-facing gaming helper isn’t the broader aim right here, members of the DeepMind crew informed The Verge throughout a Wednesday briefing. Jane Wang, a senior employees analysis scientist at DeepMind, known as it “a extremely nice coaching floor” for doubtlessly transferring the abilities to real-world environments someday.
And, as ordinary, all of it comes again to the ever-intensifying AGI race between Google, Meta, OpenAI, Anthropic, and others. “This can be a important step within the course of Synthetic Common Intelligence (AGI), with necessary implications for the way forward for robotics and AI-embodiment typically,” DeepMind’s weblog submit states.
Joe Marino, a analysis scientist at DeepMind, doubled down on that, saying that SIMA 2’s capability to take actions in a digital world and deal with environments it has by no means seen earlier than is a “elementary” step towards AGI — and doubtlessly towards constructing a general-purpose robotic down the road.
{content material}
Supply: {feed_title}