Google DeepMind unveils its first “thinking” robotics AI

In a significant leap for robotics, Google DeepMind has unveiled its Gemini Robotics project, which introduces two innovative models designed to enhance robotic functionality. These models enable robots to 'think' before taking action, pushing the boundaries of what AI can accomplish in the physical world. Generative AI has revolutionized various fields by creating text, images, and audio, and now it is being harnessed to power robotic actions. Traditionally, robots required extensive training for specific tasks, making them inflexible and time-consuming to deploy. Carolina Parada, the head of robotics at Google DeepMind, emphasized that current robots are often specialized, stating, "Robots today are highly bespoke and difficult to deploy, often taking many months to install a single cell that can perform just one task." The introduction of generative AI changes this dynamic, allowing robots to adapt to new environments and challenges without needing reprogramming. DeepMind's approach is anchored in the collaboration of two models: Gemini Robotics 1.5 and Gemini Robotics-ER 1.5. The former operates as a vision-language-action (VLA) model, utilizing visual and textual data to generate actions for robots. On the other hand, Gemini Robotics-ER 1.5, where 'ER' signifies embodied reasoning, is a vision-language model (VLM) that generates the necessary steps for complex tasks based on visual and textual inputs. This model represents a significant advancement, as it is the first robotic AI capable of simulated reasoning akin to modern chatbots, an ability that DeepMind refers to as 'thinking.' However, it's important to note that this is a simplification within the context of generative AI. DeepMind claims that the ER model has excelled in both academic and internal benchmarks, demonstrating its ability to make precise decisions concerning physical interactions. While it does not execute actions itself, Gemini Robotics 1.5 steps in to implement those decisions, marking a new era in AI-driven robotics.