Room U
Workshop

AI and Machine Learning in communication networks

In person
  • Date
    10 July 2025
    Timeframe
    09:00 - 17:15 CEST
    Duration
    8h 15 minutes
    Share this session

    This workshop marks the 3rd edition of the MLComm series, continuing the exploration of AI/ML in modern communication networks. As networks evolve towards IMT-2030 (6G), AI/ML is transforming telecom architectures, enabling distributed intelligence, AI-native operations, and multimodal capabilities for automation, optimization, and decision-making.

    The 2023 edition provided foundational inputs to the Focus Group on Autonomous Networks, while the 2024 edition facilitated collaborations toward the Focus Group on AI-Native Networks.

    The workshop aligns with key ITU standardization activities, rooted in fundamental specifications from FG ML5G (ITU Y.3172) and FG AN (ITU Y.3061). It has enabled contributions to the ITU Journal for Future and Evolving Technologies, ML5G Challenges, and the Discovery webinar series, promoting research and real-world AI applications in telecom.

    Several ITU ML5G initiatives, such as AI/ML Challenges and Zindi-hosted competitions, further demonstrated the importance of regional inclusivity, diverse datasets, and research collaborations around AI/ML in networks. Resolution 101 of WTSA-24 (New Delhi, 2024) mentioned standardization in AI-enabled networks, protocols, services, and applications. The ITU AI Readiness Initiative continues to explore AI integration challenges in SDGs, while Innovate for Impact programs foster global collaborations on use cases, datasets, and proof-of-concepts for AI-driven networks.

    This year’s MLComm workshop brings together innovations in AI technologies, opensource and standardization efforts, architectural paradigms, and cutting-edge toolsets for AI-driven telecom networks. This workshop aims to strengthen the bridge between AI research, industry innovation, and global standardization, ensuring that AI technologies in telecom meet real-world needs while driving next-generation network advancements.

    Schedule

    Exploring AI-driven transformations in RAN, core, and edge, including federated learning, Open RAN, semantic communications, and autonomous networks.

    Highlighting AI-enabled network simulators, digital twins, dataset repositories, and validation frameworks from partners.

    Building on the insights from the workshop sessions, this panel explores how AI will transform telecom networks—not just as an intelligent pipe but also as an evolving platform for future applications and services. So far in the workshop, we have looked at Innovations in AI models, standards perspectives and open-source implementations, architecture components, and finally tools, simulators and datasets for integrating AI/ML in networks. 

    We discussed the recent advancements in AI/ML, from federated learning to LLMs for wireless, and AI Native networks. We also had presentations on new mechanisms such as agentic architectures and open-source components. We also looked specific architecture enablers such as Edge AI and digital twins and new innovative technologies such as AI native RAN. Now is the time for panel discussion on “picking out the AI trends for future networks”. 

    1. What are the architecture changes in the network of future which can hold all these wonderful advancements in AI/ML? for e.g. how and where do you think models would be trained and hosted? Would model babysitting become one of the many responsibilities of the network orchestrator?

    2. Are there end-to-end pathways for the AI agents to work smoothly across RAN and Core and Applications? E.g. How would we use protocol innovations like OpenAI’s MCP (Model Context Protocol) and A2A to redefine interactions between applications (network functions?) and models in future networks?

    3. Do we require custom models, layer-wise, use case wise, geo specific? In this regard, is data still a differentiator? Would there be les « nantis » et les « démunis » (haves-and-have-nots) as far as network knowledge bases go? 

    4. What’s up in the application space? What are the new AI-Native application interfaces? Would the lines blur between applications and lower layers? Would that obliterate all the lower layers (as claimed)?

    AI-native networks demand advancements in agents, generative AI, and optimized hardware, alongside techniques like quantization for efficiency. Finally, data and code generators are driving automation in network management and AI lifecycle development. This panel will examine the key enablers and gaps that define AI’s role in shaping the telecom landscape.  

    Share this session
    • 19
      Days
      18
      Hours
      23
      Min
      20
      Sec