Capgemini’s Post

Cities globally are sprinting to adopt AI, with many seeing it as a driver of greater productivity and efficiency and, ultimately, economic growth and competitiveness. To learn how AI will reshape cities, Deloitte, ServiceNow, NVIDIA, and ThoughtLab have come together to conduct a pioneering study on the AI plans, investments, and practices of 250 cities in 78 countries around the world. The study examines how these cities are harnessing all varieties of AI, from machine learning and robotic process automation to Generative AI (GenAI) and now agentic AI:

So true - LLMs (and agentic AI build on them) are an intermediate step. Despite their sometimes awe-inspiring (or awful) answers they are just a specific skill within many needed to achieve world understanding.

What if the next AI breakthrough gave machines a more grounded sense of understanding? The 2025 State of AI report highlights AI2’s Molmo-Act, an “action reasoning” model that plans in 3D physical space. Given a high-level command, Molmo-Act generates intermediate visual or geometric waypoints (like depth or trajectory sketches) before translating them into continuous robot motions. It’s an intriguing step toward systems that can link abstract reasoning with the physical world. Combined with emerging work on long-term memory and planning, these advances hint at progress toward more contextually aware AI, not yet “human-like” in the way Yann LeCun envisions, but perhaps a little closer? Stay tuned for next year’s discussion as this evolution continues! Link in the comments for the full conversation.

The A* Algorithm is one of the most trusted pathfinding techniques in Artificial Intelligence, known for its ability to find the most efficient route between two points. By balancing real movement cost with intelligent prediction, A* delivers optimal paths with speed and precision. In our recent knowledge session, Sreejith K R took us through the inner workings of this powerful algorithm, explaining how heuristics enable A* to not just search, but think ahead. From robots navigating obstacles to delivery systems optimizing routes, A* exemplifies how AI blends logic, efficiency, and foresight to solve real-world challenges. Thanks to our teammate Sreejith for leading this engaging session and deepening our understanding of AI-driven pathfinding! #AI #Pathfinding #ArtificialIntelligence #KnowledgeSharing #TeamLearning #TurnB

What’s Next for #PhysicalAI by The GTM Firm Snowflake’s Silicon Valley AI Hub, the F50.AI Physical AI Summit made one thing clear: the next wave of artificial intelligence won’t live on screens — it will live in motion. Physical AI blends perception, cognition, and action, enabling robots and systems to adapt in real environments. The focus is shifting from humanoid spectacle to specialized impact — machines that weld, harvest, and care alongside people. The next frontier may come from cognitive autonomy. As AI visionary Robert Scoble shared after 1 demo, “Autodesk is f*ed.” The demo, by 20-year-old Brayden Levangie, showed an AI system that reads goals, reasons through them, and executes autonomously — a glimpse of what could follow the chat paradigm. The intelligence revolution is becoming embodied. The challenge now is ensuring it evolves with human purpose. More on Substack: What’s Next for Physical AI https://lnkd.in/geBbb3wY

🤖 🤖 🤖 AI is transforming how Europe powers its industries, manages its cities, and delivers public services. But as adoption accelerates, so does the need for trust - in how AI systems make decisions, handle data, and reflect human values. 🩷 That challenge is at the heart of a new policy brief, AI, Your Way: Trusted, Scalable, and Ready to Deliver, developed by five leading European projects, TANGO Project, AI4REALNET Project, PEER AI, and HumAIne project. The brief distills key insights from our joint workshop at the AI, Data and Robotics Forum (ADRF) 2025, highlighting pathways to build AI that is both trustworthy and impactful across sectors. READ THE BRIEF: 👉 https://lnkd.in/evfUyN3g 👀 #ARDF25 #AItrust #AIethics #AIlifecysle #EUvalues #EUregulation #humancenticAI #Hybriddecisionmaking

Introducing the HUSKYLENS 2 by DFRobot, a next-generation AI vision sensor designed to bring real-world machine vision into classrooms and makerspaces. With a dual-core processor, integrated 6 TOPS AI accelerator, 20+ built-in models, and flexible LLM interaction, HUSKYLENS 2 empowers educators and students to create, code, and innovate with confidence. Discover how HUSKYLENS 2 is redefining hands-on AI learning: https://bit.ly/4nJkB9F #STEMfinity #DFRobot #HUSKYLENS2 #AIinEducation #STEMLearning #EducationalTechnology #Innovation

Are general-purpose AI models missing the mark for your specialized needs? RunwayML is championing a pivotal shift in generative AI, moving beyond "one-size-fits-all" video to offer custom fine-tuning for enterprise and scientific partners. This addresses the critical limitation of broad AI, which often struggles with the nuanced, high-fidelity demands of fields like robotics, life sciences, and architecture due to its "opinionated" generic training data. By enabling partners to adapt foundational models with their own proprietary datasets, Runway is transforming generic tools into precision instruments. We're already seeing this with Lionsgate, leveraging custom models for pre-production savings, and Getty Images, ensuring brand-aligned, commercially safe content. This pivot signals a maturation of generative AI, accelerating its adoption as an indispensable problem-solving and discovery tool far beyond traditional creative industries. What does this move towards specialized AI mean for your sector? How do you envision custom models reshaping your workflows or driving new breakthroughs? #GenerativeAI #AI #RunwayML #CustomAI #IndustryInnovation #AIForEnterprise Read more: https://lnkd.in/gDP85U2T

Is it Artificial Intelligence… or just Intelligent Application? 🤖💡 There’s a lot of buzz around AI — and rightly so. Big Data, Machine Learning, Generative AI, Digital Twins… it’s everywhere. These technologies can indeed solve real challenges and improve industrial performance — but only when applied thoughtfully. At Nucomat, we believe success lies not just in having the right tools, but in how you implement them. Sometimes, the smartest solution is a hybrid one — using the right technology for the right task. Meet Robin, a digital twin and real-time dynamic scheduler to control your automated laboratory systems. Not built by generative AI, but by human intelligence, Robin simulates and optimizes robotic lab workflows — before a system is even built, or before you run your valuable samples. And Robin runs the systems using the same control logic, adapting in real-time to the live situation ensuring accuracy and efficiency every step of the way. Curious to see Robin in action? 🎥 👉 Take a look at our video or get in touch at info@nucomat.com to learn more. #DigitalTwin #LabAutomation #AIinIndustry #IndustrialInnovation #Robotics #AutomationSolutions #DataDriven #Nucomat #FutureOfLabs

🚀 𝑼𝒏𝒗𝒆𝒊𝒍 𝒕𝒉𝒆 𝒗𝒊𝒔𝒊𝒐𝒏 𝒐𝒇 𝒕𝒉𝒆 𝑰𝒏𝒕𝒆𝒍𝒍𝒊𝒈𝒆𝒏𝒕 𝑬𝒓𝒂 — 𝒘𝒉𝒆𝒓𝒆 𝒊𝒎𝒂𝒈𝒊𝒏𝒂𝒕𝒊𝒐𝒏 𝒎𝒆𝒆𝒕𝒔 𝒓𝒆𝒂𝒍-𝒘𝒐𝒓𝒍𝒅 𝑨𝑰 We’re entering a new era of intelligence — one that unites 𝒗𝒊𝒔𝒊𝒐𝒏, 𝒍𝒂𝒏𝒈𝒖𝒂𝒈𝒆, 𝒂𝒏𝒅 𝒓𝒆𝒂𝒔𝒐𝒏𝒊𝒏𝒈 into seamless real-world deployments. From 𝑹𝑨𝑮 𝒑𝒊𝒑𝒆𝒍𝒊𝒏𝒆𝒔 to 𝒔𝒂𝒇𝒆𝒕𝒚 𝒎𝒐𝒅𝒆𝒍𝒔, from 𝒐𝒑𝒆𝒏 𝒅𝒂𝒕𝒂 𝒆𝒄𝒐𝒔𝒚𝒔𝒕𝒆𝒎𝒔 𝒕𝒐 𝒂𝒖𝒕𝒐𝒏𝒐𝒎𝒐𝒖𝒔 𝒔𝒊𝒎𝒖𝒍𝒂𝒕𝒊𝒐𝒏𝒔, AI is evolving beyond theory — it’s learning to understand the world. With breakthroughs like 𝑵𝒆𝒎𝒐𝒕𝒓𝒐𝒏 2-𝑽𝑳 𝒂𝒏𝒅 𝑶𝑪𝑹𝑩𝒆𝒏𝒄𝒉𝑽2 (12𝑩 𝒎𝒖𝒍𝒕𝒊𝒎𝒐𝒅𝒂𝒍 𝒓𝒆𝒂𝒔𝒐𝒏𝒊𝒏𝒈 the frontier of AI assistants is being redefined. What once lived in imagination is now becoming part of our daily experience — a step closer to building intelligent systems that see, reason, and act. ✨ 𝘛𝘩𝘦 𝘐𝘯𝘵𝘦𝘭𝘭𝘪𝘨𝘦𝘯𝘵 𝘌𝘳𝘢 𝘪𝘴𝘯’𝘵 𝘤𝘰𝘮𝘪𝘯𝘨 — 𝘪𝘵’𝘴 𝘢𝘭𝘳𝘦𝘢𝘥𝘺 𝘶𝘯𝘧𝘰𝘭𝘥𝘪𝘯𝘨 𝘣𝘦𝘧𝘰𝘳𝘦 𝘶𝘴. #AI #MachineLearning #VisionLanguageModels #RAG #GenerativeAI #Innovation #OpenData #AIAssistants #DeepLearning #TechFuture