Huawei's Pangu Goes Commercial: The AI Model Sweeping the Art Creation Industry Now Applied in Coal Mines
-
On July 18, Shandong Energy Group and Huawei held a press conference for the first commercial application of the Pangu AI mining model. At the conference, Jiang Wangcheng, President of Huawei's Coal Mine Business Unit Solutions, stated that the introduction of the large model essentially aims to solve the issues of AI replication across mines and the shortage of AI talent.
Shandong Energy Group operates 115 mines domestically and internationally, including 85 production mines. The group has been undergoing digital transformation for years, deploying AI technology underground and seeking to further address challenges in digital reform by introducing AI large models. The press conference revealed the latest applications of the Pangu AI model in Shandong Energy Group's coal mining operations: currently, Shandong Energy Group, Huawei, and related tech companies have validated the model's capabilities in industrial production through pilot scenarios and are developing and implementing the first batch of applications, covering nine specialties including coal mining, tunneling, main transportation, auxiliary transportation, lifting, safety monitoring, impact prevention, coal washing, and coking.
A typical scenario is tunneling operations. Wang Licai, Assistant General Manager of Shandong Energy, cited an example: traditionally, tunneling operations relied on dedicated safety inspectors for on-site supervision in a 'person-to-person' manner. Now, with the machine vision recognition technology of the large model, combined with equipment operation data, it can identify scenarios such as personnel entering hazardous areas, falls, and cutterhead landings, achieving a shift from manual supervision to automated monitoring. Tunneling operations themselves are high-risk and difficult to control in coal mines, often prone to issues like irregular personnel operations.
Wang Licai stated that one of the group's goals is to significantly reduce the workload of human monitoring.
The AI transformation of the coal mining industry is considered a more challenging reform than the previous round of mechanization, as it is not a point-to-point equipment procurement but a systemic change with longer cycles, broader scope, and higher complexity. The AI large model, as a new paradigm of artificial intelligence, significantly lowers the barrier to AI adoption.
Wang Licai noted that the AI transformation solutions introduced earlier were difficult to replicate in practice. For example, models trained for one mine production scenario often required retraining when transferred to other mines, with high costs and long cycles. In response, Jiang Wangcheng stated that large models have good generalizability and support large-scale replication. Models trained for one production unit can meet commercial requirements with only minor additional training data when transferred to new production units.
In Wang Licai's view, another fundamental issue in digital reform is the shortage of intelligent mining talent. AI applications in coal mining, transportation, and electromechanical scenarios have high professional barriers, requiring experts who understand both the business and AI. Currently, in the process of AI implementation, production workers and R&D personnel are insufficiently involved and lack autonomy.
Jiang Wangcheng stated that lowering the barrier to AI adoption is a key capability of AI large models. 'We have always considered AI to be something very advanced. If you're not a Ph.D. or postdoc, using AI can be daunting. But with the advent of ChatGPT, everyone feels that AI has entered daily life.'
Jiang Wangcheng believes that large models, as an engineering method, can significantly lower the talent threshold for enterprises, turning AI into a production factory where tasks once done by postdocs can now be handled by undergraduates.