AI-RAN Network Demoed at MWC

- The Large Telecom Model (LTM) is a generative AI foundational model developed by SoftBank, trained on its own extensive network data and operational expertise to automate complex network design, management, and optimization tasks. - "Intent-driven" networking allows operators to state a high-level goal, such as "provide low latency for autonomous vehicles," which the AI-RAN then automatically translates into the specific, low-level network configurations required to achieve that outcome. - zTouch Networks is a spin-off from Northeastern University's Institute for the Wireless Internet of Things, specializing in zero-touch automation that converts operator intents into AI-driven actions across the Open RAN ecosystem. - Keysight Technologies provides the crucial testing and validation tools for AI-RAN, creating digital twins of networks to train AI models and benchmark their performance under realistic, high-scale conditions without impacting live networks. - This collaboration is part of a broader industry movement, the AI-RAN Alliance, which was co-founded by SoftBank in 2024 to bring together telecom operators, AI firms, and research institutions to advance the integration of AI into cellular networks. - In previous trials, specialized models fine-tuned from SoftBank's LTM achieved over 90% accuracy in predicting optimal base station configurations, reducing a process that typically takes days down to just minutes. - Northeastern University is a key research partner in the Open RAN space, leading federally-funded initiatives like Open6G+AI and developing open-source testing platforms such as Colosseum, a massive wireless network emulator. - The AI-RAN architecture aims to use the same GPU-accelerated hardware to run both RAN functions and AI applications, improving infrastructure utilization and enabling new AI-based services directly at the network edge.