The robotics world is experiencing a debasing earthquake that could radically question the very human/machine interaction. For the past few years, robots have been confined to serving as preprogrammed tools; they follow only very precise instructions to perform their duties in a repetitive manner.
Today, that paradigm will give way to a new generation with the introduction of Gemini Robotics 1.5 and Gemini Robotics-ER 1.5. Leading the charge is Google Deepmind, which has been at the cutting-edge of AI research for as long and has been working around the clock on more versatile robots that can think, learn and even solve real world problems without constant human intervention. The target is clear: to transform robots from tools in which you place your trust as it can aid in the work that humans do, into helpers.
The Smarter, More Flexible Robot Revolution
The big tech players are pursuing the power of technology to its limit and robots appear to be the next one. Whereas robots were previously restricted in what they could do: for the most part, carry out repetitive movements within structured environments, like a manufacturing line, Google DeepMind is keen to see its new models take on complicated and multi-step tasks. And here is where Gemini Robotics steps firmly into the spotlight.
The search giant unveiled these two latest significant AI models — the Gemini Robotics 1.5 and Gemini Robotics-ER 1.5 — in a recent update on its robot strategy. This is not one big monolithic system, but rather a clever two-model setup that approximates a human-like flow of operations: plan, then act.”
Gemini Robotics-ER 1.5 (Embodied Reasoning): This is the high-level brain, or the thing that orchestrates. It’s all about logic and being able to deconstruct a large, complex job into smaller tasks that you can do. Notably, the ER model can search information on-line when required and the robot has a large up-to-date knowledge source.
Gemini Robotics 1.5 (Vision-Language-Action): The chance that one is destined to act. It takes the commands from the ER model and generates a set of physical motor commands to execute them on (i.e., it deals with low-level execution).
It allows the robots to plan multiple steps ahead, as opposed to thinking of single commands in a vacuum, said Carolina Parada, the lead robotics head at Google DeepMind.
Transferable Knowledge And Real World Applications
These reworked Geminis are just so full of use. Take, for example, the not-so-simple activity of packing for a vacation. With the usual preprogrammed robot, you’d be lucky to get help putting suitcase in. Now, with the new models, your robot would be able to help you decide what to pack, look up the weather at your destination and in general assist in planning out a trip if not itself serving as an advertisement for how good Gemini Robotics has become with its deep reasoning. Because going on behind that two-model system is a pretty human way of processing: We assume we plan first, and then we act.
Among other radical improvements, knowledge transfer is a MAJOR one. If a skill or behavior is learned on one robot, it can now be taught to another—no matter if the “recipient” robot, built in a different country and with radically different hardware capabilities—are equipped. This capability for Gemeni’s core intelligence to generalize across multiple embodiments will be a major accelerator for the whole domain. The fact that the underlying Gemini Robotics intelligence is universal means a single model can power different types of robots.
Transforming Tools into Assistants
The ramifications of this new Gemini-driven bot are massive. In the health sector, for instance, assistance robots could provide customized support catered to individual patient requirements by taking on different actions in real-time scenarios using gemini robotics strong models. For your home, the Gemini Robotics system may become a very helpful in-home assistant as well, more than simple automation into true cognitive assistance.
But no major technological leap is without its hiccups — and virtual and augmented reality are no different. The lightning-fast development of AI models, like Gemini Robotics’, also raises important questions around data privacy and trust and safety. Before releasing it at scale, Google would have to run extensive tests and develop strong safety protocols. One thing’s for sure: Google DeepMind is on a mission to harness its game-changing Gemini Robotics technology to turn robots – currently limited tools of the trade – into intelligent, adaptive assistants… that can work with us.
