Core Insights - GSMA has launched Open Telco AI, a global initiative aimed at accelerating the development of telco-grade AI through collaboration among operators, vendors, AI developers, and academic institutions [1] - The initiative addresses the performance gap in current AI models, which struggle with telecom-specific tasks, with only 16% of telecoms' GenAI deployments applied to network operations [1] - Founding supporters AT&T and AMD are contributing open telco-models and compute capacity, respectively, to enhance the development of telco-focused AI [1] Group 1: Initiative Overview - Open Telco AI provides a new portal for telco open models, data, compute, and tools to facilitate the development and evaluation of AI models tailored for the telecom industry [1] - The initiative aims to create a shared foundation for telco-grade AI models, data, compute, benchmarks, and community engagement [1] - The Telco Capability Index will track progress by measuring model performance across various telecom-specific tasks [1] Group 2: Contributions and Support - AT&T is releasing a family of open telco-models that are hardware and cloud-agnostic, demonstrating AI's value across different project sizes and compute resources [1] - AMD is providing compute capacity for model training and evaluation through its GPU platforms and cloud partner TensorWave [1] - The initiative is supported by various community programs, including competitions like the AI Telco Troubleshooting Challenge, which attracted over 1,000 registrations [1] Group 3: Building Blocks of Telco-Grade AI - The portal will support the co-creation of essential components for telco-grade AI, including community resources, benchmarks, compute access, open data, and telco models [1] - High-performance open weight models will be developed for tasks such as network troubleshooting and standards interpretation, with contributions from various academic and industry partners [1] - The initiative has garnered support from numerous organizations, including Huawei Technologies, NVIDIA, and various universities, contributing data, models, and use cases [1]