At an exhibition in Fuzhou, Fujian province, an employee of China Unicom showed that by leveraging extensive data analysis and advanced AI large language models (LLMs), companies can automate workflows, thus reducing labor costs and enhancing production efficiency. In one demonstration by China Unicom, tea enthusiasts can customize their tea orders via a tablet, specifying the type, quantity and packaging. Within minutes, robotic arms precisely select, seal, engrave and deliver the tea to a storage unit in an entirely automated process. This human-free production line exemplifies the remarkable advancements in artificial intelligence and automation. This innovation is part of the telecom carrier’s in-house LLM, which is roughly translated as Yuanjing, designed to empower various sectors like clothing, car and equipment manufacturing with robust AI capabilities.
This initiative reflects a broader trend among China’s top telecom operators – China Unicom, China Mobile and China Telecom – which are aggressively investing in AI. As AI technology rapidly evolves, these giants are enhancing their capabilities in computing power, data and applications, thereby fueling a new wave of industrial modernization, triggering innovation in areas such as 5G, industrial internet and smart manufacturing. For instance, China Unicom’s “AI plus” strategy aims to build LLMs across more than 10 industries — such as equipment manufacturing and light textiles — to boost efficiency and quality. Additionally, they focus on strengthening internal security to safeguard this new industrial landscape.
China Mobile has introduced its Jiutian AI model, collaborating with government agencies and enterprises to co-develop over 20 industry-specific AI models. These models span a wide range of applications including network management, customer service, smart homes, public services, transportation, healthcare and industrial sectors.
In another engaging showcase, a dual-armed six-axis robot prepares and serves coffee, even cleaning the cup afterward – a testament to the fusion of networking, computing power and AI. This robot integrates the latest technologies in network, computing and AI. Based on data accumulation and intelligent algorithms, it has developed multimodal capabilities to meet needs in fields such as entertainment, household chores and family care, China Mobile said. The surge in digital innovation is underpinned by a robust data infrastructure. According to a report by the National Data Bureau, China’s data production reached 32.85 zettabytes (1 ZB equals approximately 1 trillion gigabytes) in 2023, a 22.44% year-on-year increase.
To harness the full potential of these data elements, the NDB, along with other departments, launched the “Data Elements plus” three-year action plan (2024–26). The first batch of 20 typical application cases spans industries like industrial manufacturing, modern agriculture, trade circulation and transportation, showcasing best practices in data utilization. Multiple departments have expressed their commitment to deepening reform of market-based allocation of data elements, optimizing the digital infrastructure and unlocking the value of data elements. Liu Liehong, head of the NDB, emphasized the need for innovation in data infrastructure. He highlighted efforts to promote data resource property rights, circulation, transaction, benefit distribution and security governance. The goal is to build a dynamic, categorized, and hierarchical data resource management system that facilitates data development, openness and efficient flow.
In a groundbreaking development, TeleAI, the artificial intelligence research division of China Telecom, has unveiled the industry’s first advanced speech recognition model capable of understanding and processing over 30 Chinese dialects. This large language model (LLM), known as Xingchen in Chinese, addresses a significant barrier in customer service, where millions of daily calls, especially from elderly users in remote regions, are made in local dialects that traditional AI systems and customer service representatives struggle to understand. With approximately 20% of China’s population not fluent in spoken Mandarin Chinese, this advancement aims to bridge the communication gap, ensuring equitable access to digital services for all, China Telecom said. TeleAI has tackled the challenge of creating an LLM that breaks away from conventional approaches, which typically limit recognition to a single dialect. By incorporating the dialects of Guangzhou, Shanghai, Sichuan and Wenzhou, among others, the Xingchen model marks a significant leap forward in natural language processing technology, China Telecom said. This is particularly crucial for elderly residents and those in less-developed regions, who rely heavily on their native dialects for communication, the company added.
Traditionally, dialect research has been a labor-intensive process, reliant on subjective human assessment and annotation, making comprehensive documentation difficult. AI, with its ability to systematically and efficiently catalog and analyze linguistic data, offers a transformative solution. TeleAI said it has already amassed a database exceeding 30,000 hours of high-quality dialect speech, positioning itself as a leader in this field, the China Daily reports.
