Google’s robot AI just got a lot bolder — now it has to prove it’s more than a demo

Google’s robot AI just got a lot bolder — now it has to prove it’s more than a demo
Google is turning up the volume in the race to build smarter robots. Its new Gemini Robotics 2 promises machines that can move with more dexterity, plan across multiple steps, and even work together — but the bigger question is whether that software leap can survive the messiness of real-world hardware.

The company’s pitch is sweeping. DeepMind says Gemini Robotics 2 is the “intelligence layer” for adaptable robots, extending control from “feet to fingertips” and pushing beyond narrow, pre-programmed routines. In Google’s telling, the upgrade means humanoids that can “walk, crouch, stretch, and manipulate objects to clean up a cluttered room,” while also teaming up with other robots to finish jobs faster.

That aligns with broader industry momentum. Axios frames the update as part of a wider push to use AI models to make humanoid and other robots more versatile, with separate systems handling both low-level motion and higher-level planning. The Verge, meanwhile, underscores the headline technical claim: Google DeepMind’s new model can control a robot’s entire body.

Still, the skepticism is baked into the moment. The promise of “whole-body intelligence” sounds impressive on paper, but robotics history is littered with flashy demos that stumble when confronted with battery limits, unreliable hardware, and unstructured environments. Even Axios points to a practical bottleneck beyond the model itself: U.S. companies may face a tougher time sourcing robots after the FCC moved to ban future sales of Chinese-made robots over security concerns.

On X, Google executives leaned hard into the breakthrough narrative. Demis Hassabis said robots can now “reason through every movement” and handle tasks that “weren’t possible before, like tying delicate knots,” while also coordinating on complex workflows. Sundar Pichai amplified the same message by reposting Hassabis’s announcement.

That leaves the story in a familiar place: the software looks more ambitious, the corporate messaging is bullish, and the real verdict will come only when these robots leave the lab.

Continue reading https://foxvector.com/stories/019fb81f-f86c-1d55-728e-187b88134173