Huawei Introduces Pangu Large Model for AI Industry Transformation


Currently, the development of Artificial Intelligence (AI) is being intensively carried out by world technology companies. The development of AI itself cannot be separated from making various human activities more simple and efficient. Not only that, the development of AI technology also helps the industrial sector to respond to the challenges of a fast-paced era.

One of the leading technology companies that has contributed to the development of AI technology is Huawei.

In a written statement received on Monday (12/16/2021), Huawei has officially launched a large AI model called Pangu Large Models in 2021. This innovation was deliberately launched to serve customers through Huawei Cloud.

Huawei explained that Pangu Large Models have many advantages, namely Super-large neural network which allows users to absorb large amounts of data, has a stronger network structure, excellent generalization capabilities, and can increase efficiency by more than 10 times.

Meanwhile, the pre-training models released by Huawei Cloud include various Pangu NLP (Neuro-linguistic programming) models, Pangu CV models, multimodal Pangu models, and Pangu models for scientific computing.

Huawei's own use of NLP uses the Encoder-Decoder architecture by considering the excellent Natural Language Generation (NLG) and Natural Language Understanding (NLU) capabilities. The use of this model is also inseparable from the quality of NLG and NLU which have been tested at the NLPCC international conference by winning first place.

In addition, the multi-task learning method of Huawei's Pangu NLP model makes it more stable, especially the hint-based fine tuning performance that can exceed the GPT series in small sample learning. Huawei's Pangu NLP Large Model training uses 40TB of text data that contains Huawei's accumulated general knowledge and industry experience over the years.

Meanwhile, the Pangu CV Large Model is the first to be realized in this industry. The size of the extracted model also varies depending on the use on the edge or other devices. And the dynamic range can reach three times.

Over the past two years, Pangu's model has won more than 10 industry challenges, more than 30 patent applications, published more than 70 IEEE and ACM journals, and more than 120 Class A CCF conference papers.

The Pangu pre-training model has been successfully verified in more than 100 scenarios from various industries, including energy, retail, finance, manufacturing, health, environment, and logistics industries.

In the manufacturing industry, for example, Huawei makes it easy for the public to be able to control energy use in air conditioners. Through this technology, people can save electricity up to 50%.

Meanwhile, for the financial sector, the Pangu Large Model technology is able to detect anomalies in finance to facilitate accuracy of up to more than 20%. Meanwhile, for the health sector in particular to detect COVID-19, this technology is able to increase identification accuracy by 22%.

According to Huawei, the Pangu Large Model technology can support various sectors because it has a very strong AI algorithm and is accompanied by powerful computing power and data capacity as its foundation.

Unfortunately, this technology has only been launched by Huawei Cloud and AI Gallery customers in China. However, Huawei ensures, in the coming year, Huawei Cloud will release the model for the world market.

To get further updates on the latest AI technology, please follow Huawei Cloud updates.

Previous Post Next Post