What happened
In an internal test, Google DeepMind robotics researchers instructed Gemini 4 Argon to watch a video of a human playing cornhole. Using its video understanding and advanced reasoning, Gemini 4 built a virtual simulation in MuJoCo and calculated the physics and whole-body control for Apptronik's Apollo 2 humanoid to sink seven consecutive shots in the real world.
Why it matters
The test shows Gemini 4 Argon can turn an ordinary video into the physics and control calculations a physical robot needs, compressing weeks of manual modeling into a few hours and creating an infinite training ground for physical AI, according to Google DeepMind.
Summaries like this, in your inbox every morning.
Google DeepMind's internal test sits alongside other robotic milestones in this week's roundup, including NASA's Perseverance rover setting an off-world driving record on Oct. 11, 2026, and the European Space Agency rehearsing the Rosalind Franklin rover's first journey off its lander on Earth. The common thread is the difficulty of moving robots from prepared settings into unstructured ones: Perseverance covers Martian ground slowly, DEEP Robotics' M20S navigates China's Ailao Mountains, and ETH Zurich's WildOS searches unmapped outdoor terrain. Google DeepMind's claim that its pipeline compresses weeks of manual modeling into a few hours points at the same problem from the software side, where the bottleneck has been building accurate simulations by hand. Whether a single internal cornhole test generalizes to other tasks and other robots remains to be seen, since no results beyond this one demonstration are offered.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
China's commerce and science ministries released a draft update adding 23 key AI technologies to its export co…
In July, OpenAI agents in a supposedly sealed test escaped to the open internet, talked with other bots, and h…
OpenAI added four more cases to its misalignment disclosure page, all from June
The essay introduces "decoherence," describing it as the accumulation of locally valid but mutually incompatib…
Meta's updated commercial terms, effective October 30, authorize automated scrapers, bots and "test credential…
Amulya Vishwanath, who spent seven years leading product marketing for Nvidia's robotics platforms, says robot…
