Page 74 - AI for Good-Innovate for Impact

AI for Good-Innovate for Impact

14�2� Use case description

14�2�1 Description

In the era of Large Language Models led by ChatGPT, Large Network O&M Models have
become a focus of industry attention [1]. Leveraging its natural advantages in networks, data,
and technology, China Telecom developed the “Qi Ming” Network Large Model (hereinafter
referred to as Qi Ming), marking a breakthrough for Chinese technology companies.

Qi Ming integrates network knowledge, business, and operational data through large-scale
pre-training, fine-tuning, and optimization. It excels in understanding networks and operations,
meeting cloud-network operation needs intelligently.

Advantages include advanced structures and algorithms, leveraging China Telecom's vast
network data for comprehensive training, and offering various capabilities such as intelligent
network operation management. China Telecom pioneers network mega-model AI agent
technology, driving processes from solution generation to self-learning optimization.
In specific applications, Qi Ming has excelled in fault localization, audit efficiency improvement,
and home broadband fault handling, enhancing operational efficiency and service quality. It
offers Model as a Service (MaaS), deploying mega-model services via cloud computing for
rapid integration without deep model understanding.

Looking ahead, Qi Ming is expected to further enhance cloud-network operations and empower
industries with continuous technological progress.

Repository Link: Data is private and not public
Partners: Huawei.

UN GOALS:
• SDG 9: Industry, Innovation, and Infrastructure

The “Qi Ming” Network Large Model of China Telecom has significantly enhanced the quality
and sustainability of communication infrastructure through its intelligent application, thereby
promoting the achievement of United Nations Sustainable Development Goal 9. Its efficient
training and precise tuning capabilities make cloud-network operations more intelligent and
efficient, increasing resource utilization efficiency and reducing operating costs. Additionally,
it adapts to various scenarios, solving problems in complex business scenarios, and driving
innovation and development in the communication industry.

The “Qi Ming” Network Large Model relies on China Telecom's massive network data and
knowledge corpora for training and optimization. It acquires telecom domain expertise and
characteristic data, enabling features such as intelligent scheduling, rule understanding, and
decision-making. Additionally, it incorporates multiple model toolchains and model capability
algorithm features, pioneering network mega-model AI agent technology in the industry. This
technology achieves full-process driving from solution generation to task decomposition,
logical arrangement, perception analysis, automatic execution, and self-learning optimization,
thereby contributing to the advancement of scientific research and technological capabilities.

58