Google launches two new Gemini AI models, and they will make Robots ‘See, Think, and Act’ in real-time
Alphabet Inc’s research division, Google DeepMind, announced the launch of two groundbreaking artificial intelligence models aimed at transforming the robotics industry. The new models, Gemini Robotics and Gemini Robotics-ER, are designed to tackle a persistent challenge in robotics: Enabling machines to adapt to unpredictable, real-world scenarios. Gemini Robotics, an offshoot of Google’s flagship Gemini AI model, focuses on enhancing robots’ dexterity and interactivity. Meanwhile, Gemini Robotics-ER emphasizes spatial reasoning, offering developers advanced tools to create robots capable of navigating complex environments using Gemini’s powerful reasoning abilities.
Google CEO Sundar Pichai on ‘AI for Robots’
Introducing the AI models, Google CEO Sundar Pichai wrote on Twitter, “We’ve always thought of robotics as a helpful testing ground for translating AI advances into the physical world. Today we’re taking our next step in this journey with our newest Gemini 2.0 robotics models. They show state of the art performance on two important benchmarks – generalization and embodied reasoning – which enable robots to draw from Gemini’s multimodal understanding of the world to make changes on the fly + adapt to their surroundings. This milestone lays the foundation for the next generation of robotics that can be helpful across a range of applications.”
Google ‘wants’ intelligent robot to deal with messiness
According to a report by Bloomberg, DeepMind engineer Kanishka Rao said that these advancements bring Google closer to realizing “general purpose robotics” — machines versatile enough to handle a wide range of tasks. “Our worlds are super messy and dynamic and rich,” Rao reportedly said during a media briefing. “A general purpose intelligent robot needs to be able to deal with that messiness.”
The announcement comes amid a resurgence of interest in robotics across Silicon Valley, with heavyweights like Meta Platforms Inc, Tesla Inc, and OpenAI intensifying their efforts, and startups securing massive funding rounds. Google’s latest push builds on this momentum, showcasing its technology in a pre-taped demonstration.
The highs and lows of Google’s robotics journey
Google’s robotics journey has seen its ups and downs over the years. Over a decade ago, the company acquired eight robotics firms to pursue co-founders Larry Page and Sergey Brin’s vision of consumer-friendly robots powered by machine learning. Those efforts eventually shifted to Google X, Alphabet’s moonshot lab, and spawned Everyday Robots in 2021, a unit focused on mundane tasks like trash sorting. However, Alphabet shuttered Everyday Robots in 2023 amid widespread budget cuts, though it vowed to integrate the technology into ongoing robotics projects. Now, with generative AI at the forefront, Google appears to be reinvigorating its ambitions.
Google: Robotics AI still in ‘early phase’
The new models remain in an “early exploration” phase, according to DeepMind research scientist Vikas Sindhwani. Safety is a priority, with the Gemini models built to exhibit “common sense safety” in physical settings. Google plans a cautious rollout, starting with robots operating at a safe distance from humans and gradually increasing interaction as safety improves.
To accelerate development, Google is collaborating with robotics companies like Apptronik, with whom it is crafting humanoid robots, as well as Agile Robots and Boston Dynamics — the latter a former Alphabet subsidiary sold to SoftBank in 2017.
