Titanium Media Exclusive | Huawei Version of ChatGPT will be released in early July, titled "Pangu Chat"

(Image source: edited and shot by Titanium Media App)On June 3rd, it was reported that the Titanium Media App had recently learned exclusively from within Huawei that Huawei would release a multi-modal, billions level device that directly targets ChatGPTA large model product called "Pangu Chat".It is expected that Huawei Pangu Chat willAt the Huawei Cloud Developer Conference (HDC

(Image source: edited and shot by Titanium Media App)

On June 3rd, it was reported that the Titanium Media App had recently learned exclusively from within Huawei that Huawei would release a multi-modal, billions level device that directly targets ChatGPTA large model product called "Pangu Chat".

It is expected that Huawei Pangu Chat willAt the Huawei Cloud Developer Conference (HDC. Cloud2023) held on July 7th this year, the product was released externally and tested internally, mainly targeting ToB/G government and enterprise clients.

This means that in the domestic large-scale model arms race, after Alibaba and Baidu, another important technology giant has entered. Based on Huawei's technological capabilities, Pangu Chat is expected to become the strongest ChatGPT product in China, and Huawei's ecological industry chain enterprises will also benefit from this.

It is reported that,The Pangu Big Model was successfully established within Huawei Cloud in November 2020. For the positioning of the Pangu large model, Huawei's internal team has established three key core design principles: firstly, the model should be large enough to absorb massive amounts of data; The second is to have a strong network structure that can truly unleash the performance of the model; The third is to have excellent generalization ability, which can truly be applied to work scenarios in various industries.

In April 2021, the Pangu model was officially released to the public. Among them, the Pangu NLP large model is the first Chinese pre trained large model with 100 million parameters, while the CV large model reaches 3 billion parameters for the first time. In April 2022, Huawei Pangu upgraded to 2.0 and released a hierarchical development plan (LO, L1, L2) to create an industrial grade Pangu large model.

According to the speech PPT information of Huawei Cloud executives, the basic layer of Huawei's "Pangu Series AI Big Model" mainly includes NLP Big Model, CV Big Model, and Scientific Computing Big Model. The upper layer is the Huawei industry big model developed with partners.

Tian Qi, the chief scientist in the field of artificial intelligence at Huawei Cloud, stated in April this year that the Pangu NLP large model covers areas such as intelligent document retrieval, intelligent ERP, and small language large models. In 2022, Huawei just delivered a 100 billion parameter large model in Arabic. In terms of application, Pangu large model can be applied to intelligent customer service, machine translation, speech recognition and other fields to provide AI technical support.

Last year, Huawei implemented the Pangu model in the vertical industry field. One example is that based on the Pangu AI big model, Huawei has created the first "artificial intelligence training center based on the big model" in the mining field, achieving the "industrial" development of Al applications. This not only reduces costs - saving 80% compared to manual verification, reducing costs by 8 yuan per ton (annual production capacity of 2.6 million tons), but also increases benefits - saving 80% compared to manual verification for anti impact and anti pressure, and improving efficiency - improving intelligent customer efficiency by 30%.

Compared to other manufacturers, Huawei's advantages may lie in having a complete industrial chain and strong computing power allocation capabilities. According to previous disclosure by Zhejiang Securities,During the training of the Pangu large model with hundreds of billions of parameters, the Huawei team called on over 2000 Ascension 910 chips,Conducted over 2 months of data training capabilities. Huawei internally stated that,Every year, over 4000 GPU/TPU cards are used for large model training,The computational cost of the 3-year large-scale model is as high as 960 million yuan.

According to a research report released by Dongwu Securities on March 27th, the advantage of Huawei's Pangu Big Model lies in its talent reserves and independently controllable computing power, which is expected to become a leading big model in China. Its ecological industry chain targets are expected to accelerate development, including Huawei ecological companies such as Tuowei Information, Sichuan Changhong, Qilin Software (China Software), Tongxin Software (Chengmai Technology), and Qilin Xin'an. Guosheng Securities believes that Huawei Pangu is the first multimodal model with a scale of hundreds of billions, which is expected to empower Baiye.

It is worth noting that due to the large scale parameters of the Pangu model, the training cost is relatively high. Although the Pangu model has high accuracy and quality in processing Chinese language, it still has certain limitations in semantic understanding, especially in dealing with complex language structures and semantic reasoning. At the same time, the Pangu Chat large model will also face fierce competition pressure in the industry, requiring continuous optimization and improvement of the model's performance and quality in order to maintain its leading position in the market.(This article premieres the Titanium Media App, author | Lin Zhijia)

Disclaimer: The content of this article is sourced from the internet. The copyright of the text, images, and other materials belongs to the original author. The platform reprints the materials for the purpose of conveying more information. The content of the article is for reference and learning only, and should not be used for commercial purposes. If it infringes on your legitimate rights and interests, please contact us promptly and we will handle it as soon as possible! We respect copyright and are committed to protecting it. Thank you for sharing.(Email:[email protected])