AI News Flash | Google Unveils Gemini Robotics On-Device: Controlling Robots Without an Internet Connection

Hello from the world of AI! 🚀In today's newsletter, we bring you the following topics:　・Google unveils Gemini Robotics On-Device: controlling robots without an internet connection　・Grok's new "spreadsheet editing" feature leaked: xAI building a conversational productivity tool　・Alexa+, powered by generative AI, surpasses 1 million test users

GOOGLE

Google Unveils Gemini Robotics On-Device: Controlling Robots Without an Internet Connection

Google DeepMind has unveiled "Gemini Robotics On-Device," an AI model that controls robots without any internet connection.Built on Gemini Robotics (released March 2025), the model controls and adjusts robot movements via natural-language commands, matching cloud-based performance — with internal benchmarks reportedly outperforming existing on-device AI models.In a demo, a robot opened and closed a bag and folded clothing. Originally designed for the ALOHA line of robots, it now also supports multiple other machines, including the dual-arm Franka FR3 and Apptronik's humanoid robot "Apollo".

Source: Gemini

Google says the Franka FR3 handles unfamiliar situations and unknown objects well, and can even perform tasks on industrial assembly lines.Google has also released the Gemini Robotics SDK, letting developers efficiently train robots using the MuJoCo physics simulator with just 50 to 100 demonstrations.Beyond Google, several major tech firms and startups are also active in this field.　・Nvidia is building a foundation-model development platform for humanoid robots.　・Hugging Face is developing open-source AI models and datasets for robots, as well as conducting its own robot R&D.Startup RLWRLD (backed by Mirae Asset) is also aiming to build a foundation model specifically for robots.

GROK

Grok's New "Spreadsheet Editing" Feature Leaked: xAI Building a Conversational Productivity Tool

A recently discovered code leak reveals that "xAI," the AI company founded by Elon Musk, is developing a new feature that lets its AI assistant "Grok" edit spreadsheets. The feature is expected to let users modify file contents in real time while chatting with Grok.This reflects xAI's strategy of deeply embedding AI into everyday work tools, positioning it against productivity features from major players like OpenAI, Google, and Microsoft. According to information shared on social media by engineer Nima Owji, the goal is an intuitive, efficient experience of "editing files while chatting."

Evolving Into a Multimodal, Interactive Workspace

xAI is also investing in building a multimodal workspace centered on Grok. No official strategy has been announced, but the direction is clear from the "Grok Studio" and "Workspaces" features launched in April 2025.Grok Studio offers a split-screen view for real-time collaboration with Grok — writing, coding, reports, even browser games. Workspaces lets users manage chat history and files together, improving efficiency.This resembles Google's "Gemini Workspace," though the approach differs: Gemini Workspace stays within Google's ecosystem, while xAI favors greater flexibility and independence.xAI has not officially detailed its strategy for an interactive, multimodal AI workspace, but the direction is clear from recent announcements.In April 2025, xAI released "Grok Studio," a split-screen tool for collaborating with Grok on documents, code, reports, and browser games. It also introduced Workspaces, letting users organize files and chat history in one place.OpenAI and Microsoft now offer similar features, but Google's "Gemini Workspace" is considered closest to xAI's approach, integrating Docs, Sheets, and Gmail editing with Gemini.However, while Gemini Workspace works only within Google's ecosystem, xAI appears to be aiming for a more independent solution that flexibly integrates with other tools and platforms.It remains unclear whether xAI's new file editor will support formats beyond spreadsheets, or evolve into a full office suite rivaling Google Workspace or Microsoft 365.If Nima Owji's hints prove accurate, this new feature could be a key step toward Elon Musk's vision of turning "X" into a "super app" combining social media, chat, payments, and document tools.

ALEXA+

Alexa+, Powered by Generative AI, Surpasses 1 Million Test Users

Amazon is gradually expanding testing of "Alexa+," a new version of its virtual assistant integrating generative AI, and has announced that over 1 million users are now trying it. The service was first announced in February 2025 and has yet to launch publicly.Access to Alexa+ is currently offered via a waitlist, with invitations sent progressively to those who signed up. In recent weeks, many users have shared on social media that they received invitations, praising its personalized experience, smart-home integration, and rich new AI features.Alexa+ is completely free during Early Access, and Amazon Prime members will continue to use it for free after the official launch, while non-Prime users will pay $19.99 per month.Amazon has said it is rolling out access gradually; in May 2025, CEO Andy Jassy said only about 100,000 users had access at that time. However, the user count has grown sharply in recent weeks.

Source: Amazon

Redefining Alexa With Generative AI

Alexa+ represents Amazon's attempt to build a new generative-AI-centered smart assistant and diversify its revenue model.Alexa has long shipped on devices like the Echo speaker, but Amazon struggled to profit from that ecosystem, and as ChatGPT rose, traditional Alexa began to look dated and inflexible.With Alexa+, commands feel more natural — saying "it's cold in here" auto-adjusts the temperature, no explicit command needed. It can also launch routines, pull Ring camera footage, and switch topics smoothly mid-conversation.

Source: Amazon

A Smart Experience That Learns and Adapts to You

Alexa+ learns preferences and habits to personalize the experience — remembering favorite music, recipes, and smart-home usage for smoother daily interactions.With generative AI, it also offers features such as:　・Automatically summarizing long emails shared by users　・Creating bedtime stories for children on the spot　・Generating practice exam questions from study materials　・Automatically building travel itineraries　・Consolidating logs of smart-home device activityAlexa+ also supports practical actions such as:　・Booking restaurants through OpenTable　・Purchasing tickets on Ticketmaster　・Ordering food via Uber Eats or Grubhub　・Notifying users when a wish-list item drops in priceIt already integrates with a wide range of brands, including OpenTable, Ticketmaster, Uber Eats, Tripadvisor, Yelp, Priceline, Viator, Thumbtack, Atom, and Fodor's.Alexa+ is currently limited to certain US Echo Show devices (8, 10, 15, 21), with future support planned for other Echo models, Fire TV, and Fire tablets.Among test users, some call Alexa+ "better than Siri," with high satisfaction, though some say it's "not yet complete." Amazon says about 90% of announced features are already implemented, suggesting launch is near.

In Closing

We hope today's content helps you make the most of AI technology and stay informed. If you found this newsletter useful, please feel free to share it with colleagues or friends who may be interested. We will continue to carefully select and deliver the topics that matter most. We hope you look forward to our next

newsletter

. Thank you for your continued support.