• AI KATANA
  • Posts
  • Google DeepMind Introduces Gemini Robotics for Physical World Interaction

Google DeepMind Introduces Gemini Robotics for Physical World Interaction

Also: AI Driving Urgent Need for Worker Reskilling Amid Job Disruption

Author

AI KATANA
April 16, 2025

Hello!

This week's AI developments showcase significant strides in embodied intelligence and highlight the growing societal and infrastructural impacts of the technology. Google DeepMind unveiled Gemini Robotics, a specialized model designed to enable robots to interact physically with the world, aiming for greater generality, interactivity, and dexterity. Similarly, XPENG updated its strategy to focus heavily on AI, launching its 2025 X9 MPV with advanced AI features and showcasing its humanoid robot, while Korean startup RLWRLD secured $15 million in seed funding to advance its physical AI models, indicating strong industrial interest in robotics. Concurrently, the accelerating AI boom is driving urgent calls for worker reskilling due to job disruption across various sectors and is placing immense pressure on data center infrastructure, significantly increasing power demand and raising sustainability concerns, as highlighted in a Seagate report. Financial regulators are also taking note, with the Bank of England intensifying its scrutiny of AI adoption risks within the financial system. In the development landscape, OpenAI signaled a potential shift by preparing to release a new open-source model, its first since 2019, responding to the growing influence of open-source alternatives.

Sliced just for you:

  • 🤖 Google DeepMind Introduces Gemini Robotics for Physical World Interaction

  • 🚗 XPENG Upgrades AI Strategy and Launches 2025 X9 Flagship

  • 💼 AI Driving Urgent Need for Worker Reskilling Amid Job Disruption

  • ⚡ AI Boom Fuels Data Center Power Demand Highlighting Sustainability Needs

  • 🏦 Bank of England Intensifies Scrutiny of AI Risks in Financial Sector

Google DeepMind has unveiled Gemini Robotics, a specialized AI model derived from its powerful Gemini 2.0 foundation, designed specifically to enable robots to interact with the physical world. This advanced vision-language-action (VLA) model incorporates physical actions as an output, allowing it to directly control robotic systems. Gemini Robotics aims to enhance three crucial qualities for helpful robots: generality (adapting to novel tasks and environments using Gemini's world understanding), interactivity (understanding conversational commands and dynamically adjusting to environmental changes), and dexterity (performing complex, multi-step manipulation tasks requiring fine motor skills). By integrating these capabilities, Google DeepMind seeks to bridge the gap between AI and physical embodiment, paving the way for more versatile and capable general-purpose robots in various settings.

Chinese smart mobility company XPENG has announced a significant upgrade to its strategy, centering on an "AI Tech Tree" that integrates AI, energy solutions, and embodied intelligence to build a future ecosystem encompassing smart electric vehicles, humanoid robots, and flying vehicles. This strategic shift coincides with the global launch of its 2025 XPENG X9 flagship MPV, which boasts 496 technical upgrades. Key features include an 800V ultra-fast charging architecture adding 405 km range in 10 minutes, an AI-driven suspension system, a self-developed 40-core Turing AI Chip capable of running large models locally (set for mass production in Q2 2025), and enhanced safety features that achieved top scores in major crash tests. XPENG also showcased its "IRON" humanoid robot powered by the Turing chip. This strategic AI focus and product launch underscore XPENG's ambition to lead innovation in future mobility across multiple domains.

The rapid advancement and adoption of generative AI are causing significant job disruption across various industries, creating uncertainty for workers and highlighting an urgent need for adaptation and continuous upskilling. Sectors involving repetitive, structured tasks such as manufacturing, technology, finance, accounting, and customer support are particularly vulnerable, with AI increasingly handling tasks like automated support, credit risk assessment, and compliance surveillance. While tech leaders and reports like the World Economic Forum's Future of Jobs suggest AI will also create new roles, potentially leading to a net job gain (projecting 11 million created vs. 9 million displaced between 2025-2030), the immediate impact involves layoffs as companies replace roles or redirect resources towards AI investments. Workers are advised by labour economists to proactively upskill to remain relevant in a labour market rapidly being reshaped by AI capabilities.

A new report from Seagate underscores the massive impact of AI on tech infrastructure, predicting that AI growth will significantly increase data storage requirements for nearly all organizations (97% anticipate impact) and contribute to a forecasted 165% surge in global data center power demand by 2030 compared to 2023, according to Goldman Sachs research. This escalating energy consumption is now a primary concern for over half (53.5%) of business leaders, who face the challenge of balancing infrastructure expansion, carbon emissions management, and total cost of ownership. Seagate advocates for a unified industry approach focusing on technological innovation like its new Mozaic 3+ platform (offering 3x capacity in the same footprint with 70% less embodied carbon per terabyte), lifecycle extension through refurbishment and reuse, and shared accountability via transparent environmental monitoring to ensure AI's growth is sustainable.

The Bank of England's Financial Policy Committee (FPC) has signaled closer examination of AI adoption within the UK's financial system, outlining emerging risks in its recent "Financial Stability in Focus" report. Key concerns highlighted include potential financial instability if widespread reliance on similar AI models for core decisions leads to correlated risk misestimations and credit misallocation. The report also flags the danger of autonomous AI systems learning to exploit market weaknesses or amplify volatility, the systemic risks from concentrating reliance on a few external AI service providers, and heightened cybersecurity threats despite AI's defensive potential. The FPC plans to collaborate with industry via the AI Consortium to understand deployment practices and share risk management strategies, suggesting existing guidance and regulations may need evolution to ensure safe AI integration in finance.

🛠️ AI tools updates

Google has introduced Veo 2, an advanced AI model for video generation, now available to Gemini Advanced users. This tool enables the creation of high-quality, 8-second video clips from text prompts, showcasing realistic motion and physics. Users are exploring its capabilities by creating videos like words formed by skydiving parachutes or ice cubes on a frozen lake. The technology has been lauded for its realistic depiction of physics, enhancing the authenticity of videos. Discussions focus on Veo 2's potential in visual storytelling and content creation, with users suggesting improvements like extending video length.

Anthropic has introduced a new feature called 'Research' for its AI assistant, Claude, enabling it to perform autonomous research by accessing both the web and internal documents like emails, calendars, and Google Docs. This feature, currently in beta, is available for Max, Team, and Enterprise users in the United States, Japan, and Brazil. Additionally, Claude now integrates with Google Workspace, enhancing its ability to provide context-aware information from Gmail, Google Calendar, and Google Docs for all paid plan users.

OpenAI Signals Return to Open Source AI Models 

OpenAI is preparing to release a new open-source language model in the coming months, marking its first such release since GPT-2 in 2019. The company, known for its proprietary models like GPT-4.5 and the ChatGPT service, has solicited feedback from developers and researchers to maximize the utility of the forthcoming model. This move represents a potential shift in strategy for OpenAI, occurring as open-source alternatives from competitors like Meta (Llama), Mistral, and DeepSeek gain significant traction and adoption within the AI community and across major cloud platforms. While OpenAI's advanced models developed under its partnership with Microsoft have remained closed-source, this planned release suggests a recognition of the growing importance and demand for accessible, modifiable AI foundations in the rapidly evolving landscape.

💵 Venture Capital updates

RLWRLD, a South Korean startup focused on physical AI and robotics, has successfully secured US$15 million in seed funding to advance its development of embodied AI models. The investment round saw participation from a consortium of major industrial corporations across Korea and Japan, including LG Electronics, SK Telecom, KDDI, ANA Holdings, Mitsui Chemicals, and Shimadzu Corporation, alongside AI-focused venture capital firms like Hashed and Global Brain. This significant backing from prominent industrial players highlights a growing strategic interest and investment focus on physical AI – systems capable of reasoning and acting autonomously in real-world environments. RLWRLD's funding signals the increasing importance of robotics and embodied intelligence as the next frontier in AI development, particularly for industrial applications potentially addressing workforce challenges.

🫡 Meme of the day

⭐️ Generative AI image of the day