Huawei’s annual developer conference kicked off on July 7 in Dongguan, a manufacturing base in southern China’s Guangdong province.
The burgeoning AI industry is the focus of this year’s event, gathering developers from 30 regions and 10 countries across China to discuss new opportunities and applications of technology, as well as review their previous experiences in the field. rice field.
During the keynote speech, Huawei Cloud Executive Chairman and CEO Zhang Ping’an announced the company’s latest contributions: Huawei Cloud Pangu Models 3.0 and Ascend AI cloud services.
Pangu is a large-scale model, the same technology that powers ChatGPT. Pangu 3.0’s layered architecture means it can be quickly adapted to a wide range of downstream tasks. Customers also have the option to train the model on their own datasets, making it highly generalizable across industries.
In a subtle jab against ChatGPT, the product that has taken the Open AI world by storm (and other AI chatbots such as Baidu’s Ernie), Zhang said Pangu 3.0 is not designed to “write poetry,” but to “empower and restructure” the industry.
“The Pangu model was born to meet industry-specific needs. We help customers in all industries develop and use large-scale models to solve their problems in ways never before possible.” said Zhang.
Huawei’s self-developed AI models have already been introduced in a wide range of industries, including finance, manufacturing, and pharmaceutical R&D. Pangu’s model has already been put into practical use to improve the safety of China’s 155,000 kilometers of railways. Pangu can detect critical faults with 99.99% accuracy, thus defeating labor-intensive and relatively inefficient manual fault detection.
Huawei has also authored a research paper on the application of its model to weather forecasting, which was published in the world-renowned scientific journal Nature. According to Huawei, his Pangu Meteorology model, which generates hourly high-resolution 3D weather forecasts, The first AI that can predict the weather more accurately than traditional methods.
The newly launched Ascend AI cloud service aims to underpin the Pangu model with long-term and stable computing power. These platforms are optimized for AI model development and execution, and can train AI models faster than common processing units.