Iván Hernández Dalas: Boston Dynamics and Google Deepmind are using Gemini to make Spot smarter

Boston Dynamics' Spot quadruped equipped with AIVI-Learning.

Boston Dynamics’ Spot quadruped equipped with AIVI-Learning. | Source: Boston Dynamics

Boston Dynamics today announced it is partnering with Google Cloud and Google DeepMind to integrate Gemini and Gemini Robotics ER 1.6 into Orbit AIVI-Learning. The robotics developer said the integration will allow it to deliver a more sophisticated, intuitive, and powerful AI experience.

Industrial environments are incredibly complex, and the assets Boston Dynamics’ customers manage require more than just basic object recognition. The company said Gemini will bring better reasoning and adaptability to AIVI-Learning.

Boston Dynamics Spot and Orbit now continuously learn about the facility they’re deployed in with unprecedented depth. This, the company said, allows for higher-order reasoning and more complex visual analysis.

DeepMind released Gemini Robotics-ER 1.6 today. The upgraded reasoning-first model features enhanced reasoning and multi-view understanding, DeepMind said. It specializes in reasoning capabilities critical for robotics, including visual and spatial understanding, task planning, and success detection. It acts as the high-level reasoning model for a robot, capable of executing tasks by natively calling tools such as Google Search to find information, vision-language-action (VLA) models, or other third-party user-defined functions.

This year, Boston Dynamics has been deepening its relationship with Google DeepMind. In January, it announced that DeepMind would help make the Atlas humanoid smarter with its Gemini Robotics foundation models.

AIVI now delivers total site intelligence

AIVI-Learning delivers insights across multiple stakeholders in an operation to give them a holistic view of what is happening on the site.

This allows users to accumulate value across multiple areas of their business:

  • Safety & Security: Execute EHS checks, such as looking for dangerous debris or spills. Reduce fines, risk, and potential liability.
  • Asset Monitoring: Perform inspections of key assets like conveyor belt damage, sight glass levels, and gauges to prevent critical failures on your site. Monitoring key assets prevents downtime.
  • Materials, 5S: Identify material movement throughout a facility. Automate manual inspections previously conducted by multiple people across many shifts.

As a facility’s needs evolve, Boston Dynamics said Orbit AI can evolve with them. Because Gemini understands complex concepts right out of the gate, the company said it can instantly roll out new, advanced capabilities. Additionally, AIVI-Learning now supports gauges.

AIVI-Learning handles advanced tasks like 5S compliance audits, measuring sight glass fullness from 0-100%, accurate pallet counting, and detecting puddles of standing liquid.

Boston Dynamics sees performance gains and downtime decreases

Customers can also expect improved performance across all of their visual inspections, Boston Dynamics said. By tapping into Gemini’s advanced intelligence, AIVI-Learning delivers better baseline performance and higher accuracy on your most critical tasks, including 5S checks, lever detection, object presence, and sight glass readings. For key inspections, including digital display readings, accuracy will be significantly improved.

Furthermore, the system gets smarter behind the scenes with Zero-Downtime Upgrades. Boston Dynamics continuously updates and refines these sophisticated AI models in the cloud, meaning inspection accuracy improves automatically without requiring end-users to run manual Orbit software updates or schedule downtime.

Additionally, AIVI prompts will now feature transparent reasoning, allowing users to see exactly how the model interprets a prompt and the logical steps it took to reach its final answer.

SITE AD for the 2026 Robotics Summit save the date.

More about AIVI-Learning, Boston Dynamics’ core intelligence tool

Boston Dynamics designed AIVI-Learning to be the ultimate visual inspection tool for a facility. The company said it has made it easier than ever to access this advanced intelligence. Whether an Orbit deployment utilizes a Site Hub, Virtual Machine (VM), or Cloud hosting, users can access the company’s servers to utilize these powerful AIVI-Learning models.

By utilizing AIVI-Learning, facilities can unlock access to specialized expert models that continually evolve. Trained on industrial data sets by the best researchers in the world, AIVI-Learning delivers true site-wide intelligence, Boston Dynamics claimed.

Looking ahead, in its next launch, Boston Dynamics will use AI on Site View, meaning Orbit will alert users of puddles, unauthorized people, and a growing number of supported assets. Turning AIVI-Learning on now will provide access to this feature immediately when available.

Editor’s Note: Alberto Rodriguez, the director of robot behavior for Atlas at Boston Dynamics, will be sharing his insights on the state of humanoid robotics at the Robotics Summit & Expo on May 27-28. The keynote panel, which will take place at 10 a.m. ET on the first day of the show, will also feature insights from Pras Velagapudi, the CTO at Agility Robotics, and Aaron Prather, the director of robotics and autonomous systems programs at ASTM International.

The post Boston Dynamics and Google Deepmind are using Gemini to make Spot smarter appeared first on The Robot Report.



View Source

Popular posts from this blog

Iván Hernández Dalas: 4 Show Floor Takeaways from CES 2019: Robots and Drones, Oh My!

Image
We’ve had some time to recuperate from the madness of CES, in which more than 150,000 attendees and thousands of exhibitors displayed their latest gadgets, gizmos and yes, even robots, on the CES show floor. Last week’s reporter’s notebook focused on the collection of startups around the Eureka Park area of CES, but several of the same trends were evident on the main South Hall 2 area of the Las Vegas Convention Center, where drones, robotics, AR/VR, and gaming took center stage. While most of our time was spent at CES hosting the Service Robots Arrive in Daily Life event, we did have some time during the week to visit the show floor and record observations. Here’s what we saw on the show floor, as well as some of our own individual meetings: 1. Underwater drones make a big splash On the main floor, there was no busier aerial drone booth than the DJI booth, but that is expected given the company’s dominance in the space. But once you got past that, you noticed a lot more drone co...
Read more

Iván Hernández Dalas: How automation and farm robots are transforming agriculture

Image
Source: EOS Data Analytics Agriculture is becoming the most high-tech industry today. All the top-notch technologies are finding a place in the way we produce, harvest, and distribute our crops. And in all cases, farm automation and robots are not an affordable luxury, but a necessary minimum due to tackling the rising global food demand. Agricultural robots can mitigate labor shortages by automating tasks like planting, harvesting, and weeding, while also supporting sustainable practices. Combined with satellite imagery tools for data-driven farming, these innovations can make precision agriculture more efficient, resilient, and environmentally responsible. According to the United Nations report , by 2050, the world will need 70% more food to feed nearly 9.7 billion people. Considering this demand, farm robots and remote sensing are the cornerstone of the proper work of the supply chain. Smart machines arise for the farm Farm robots are no longer experimental concepts — they ...
Read more

Iván Hernández Dalas: Physical Intelligence open-sources Pi0 robotics foundation model

Image
Physical Intelligence, the San Francisco-based startup that has raised more than $400 million , has open-sourced its Pi0 robotic foundation model. Pi0 was introduced a few months ago and can be tuned to a range of tasks, including folding laundry, cleaning a table, scooping coffee beans, and more. Physical Intelligence has released the code and weights for Pi0 as part of its experimental openpi repository on GitHub. It also provides checkpoints for a few “simple tasks” on available platforms such as ALOHA and DROID, example code to run inference on real-world and simulated robot platforms, and the code for fine-tuning the base π0 model for your own tasks and platforms. The company said between 1-20 hours of data was sufficient to tune Pi0 to a variety of tasks in its own experiments. HuggingFace has also prepared a PyTorch port of openpi for those developers who prefer PyTorch over JAX. “We believe that general purpose models that can control any robot to perform any task will b...
Read more