SoftBank Creates AI Language Model for Telecommunications Sector

SoftBank is actively developing an advanced artificial intelligence called LLM for the telecommunications sector.
SoftBank Corp Accelerates Its AI Initiatives
In the ever-evolving world of technology, SoftBank Corp has recently added a new dimension to its focus on artificial intelligence initiatives. The Japanese operator unveiled details of a large-scale linguistic model tailored for telecommunications, just a week after announcing plans to enhance its collaboration with Open AI.
The Large Telecom Model (LTM): A Specific Foundation for Generative AI
SoftBank explained that the Large Telecom Model (LTM) serves as a foundational platform for generative AI. The model was trained using diverse datasets, enabling it to perform advanced inferences for mobile network design and operation. Specifically, the LTM is designed to optimize base station configurations, calculating setups beyond its initial training scope.
Promising Results and Key Benefits
It’s notable that the calculated configurations were verified by experts to be 90% accurate, according to SoftBank. Additionally, a significant advantage of this technology is the “time saved compared to manual or semi-automated workflows”, along with the potential to reduce human errors.
SoftBank also stated: “These results demonstrate that by refining the LTM for specific use cases, it will become easier to develop AI models tailored to various operational scenarios” in mobile networks.
Future Expectations and Collaboration
SoftBank anticipates that its LTM will contribute to AI RAN initiatives and serve as a foundation for network design and optimization. The operator plans to use the LTM to manage its own network, potentially creating new services and enhancing quality.
Furthermore, the operator revealed that it used an Nvidia AI full-stack data center for distributed training of the LTM and intends to continue its close collaboration with the American chip company for future verification and validation efforts.
Ryuji Wakikawa, VP and Director of the Advanced Technology Research Institute, stated that the LTM enables the operator to “build specialized AI models for various processes and deploy them as agents”.