HUAWEI CLOUD Pangu Model 3.0 has been released, and a global competition of 100 models is underway

On the afternoon of July 7, Huawei Developer Conference 2023 (Cloud) was held in Dongguan. At the meeting, Zhang Pingan, Executive Director of Huawei and CEO of HUAWEI CLOUD, officially released HUAWEI CLOUD Pangu Model 3.0. He said that Pangu is to "reshape thousands of industries", not to write poetry, only to do things, focus on product research and development, software engineering, production and supply, marketing, customer operations in value scenarios, and deeply cultivate industry government affairs, finance, manufacturing, coal mines, railways, pharmaceuticals, meteorology and other fields.

Zhang Pingan, Executive Director of Huawei and CEO of HUAWEI CLOUD, delivered a speech at the conference.

Zhang Pingan introduced that Pangu Model 3.0 is a series of large models completely oriented to the industry, including a 5+N+X three-layer architecture. The L0 layer includes five basic large models: natural language, vision, multimodality, prediction, and scientific computing; The L1 layer is N industry large models, including government affairs, finance, manufacturing, mining, meteorology and other large models, and can also be based on the industry customer's own data, on the L0 and L1 layers of the Pangu large model, training proprietary large models; The L2 layer provides more detailed scenario models, focusing more on specific industry applications or specific business scenarios such as government hotlines, network assistants, lead drug screening, conveyor belt foreign body detection, and typhoon path prediction.

It is reported that Pangu 300 provides customers with a series of basic large models of 10 billion parameters, 38 billion parameters, 710 parameters and 100 billion parameters, which can meet the diversified needs of different scenarios, different delays, and different response speeds. At the same time, it provides knowledge Q&A, copywriting generation, code generation, image generation and image understanding of NLP large models, as well as image generation and image understanding capabilities of multimodal large models.

Big models accelerate the development of a new generation of artificial intelligence, and the world has set off a competition of 100 "models"

Looking back at the history of the development of artificial intelligence, from the 1960s to the 1990s, the first generation of artificial intelligence was knowledge-driven artificial intelligence, mainly represented by expert systems; From 1990 to 2020, the second generation of artificial intelligence was connectionism represented by data-driven deep learning, which was characterized by data + computing power + algorithms; Since 2020, the third generation of artificial intelligence has been driven by "knowledge + data", characterized by the ability to think like humans such as common sense, experience, and reasoning.

Since the advent of ChatGPT in December 2022, emerging technologies represented by pre-trained large models have accelerated the development of a new generation of artificial intelligence, and a global competition of 100 "models" has been set off. Zhang Pingan said that the current AI technology from multiple branches of development, into the era of large models, so that AI has become a general technology, after OpenAI released ChatGPT, a variety of large models emerged, hundreds of large models have been released around the world, China has released 80+, toC applications bloom.

Industry experts pointed out that the parameter scale of large models is getting bigger and bigger, ChatGPT-3 has 175 billion parameters, GPT4 is not disclosed, but it is expected to exceed the order of trillions, which is similar to the number of human neuron connections, which also makes the reasoning decision-making ability of large models reach the point of panic for scientists. The industry is actively exploring the future development direction of large models.

In China, representative ones include Baidu Wenxin Yiyan, Alibaba Cloud Tongyi Qianwen, Tencent Mixed Yuan Model, Huawei Pangu Big Model, etc. However, some insiders believe that there are only 5 companies in China that can train 20,000 GPU cards, and only 3 to 4 companies can make general-purpose large models. And Huawei is clearly an important seed player in this race.

According to public information, Huawei's Pangu model was officially released in April 2021 and upgraded to version 2.0 in April 2022. Pangu model includes four series in the fields of natural language processing (NLP), computer vision (CV), multimodal and scientific computing, with a multimodal Chinese database, which realizes the combination of discriminant model and generative model, and can flexibly support downstream tasks, such as text and life diagrams, text-oriented image completion, image editing, etc.

Huawei previously demonstrated that with the support of Huawei's Pangu model, AIGC has completed internal applications such as text generation, image generation, and video editing, and AIGC has cooperated with mobile phone manufacturers to automatically generate mobile phone screensavers, and cooperated with brand retailers to generate posters and advertising images for different products.

Ren Zhengfei talks about the AI big model: To pay attention to applications, Huawei focuses on making AI underlying computing power platform

On March 17 this year, Ren Zhengfei said when talking about AI big models, the future AI big models will be surging, everyone should pay attention to applications, especially the application of industrial and agricultural society, the application of models is sometimes more promising than the model itself. Huawei will focus on the underlying computing power platform of AI and strive to build a computing power platform that meets the needs of society. "What are the opportunities for ChatGPT for us? It will stretch up the calculation, stretch the pipeline flow, so that our products have market demand. ”

In terms of industry applications, Nanduwan Finance Agency reporter noted that Huawei has launched the Pangu Meteorological Model, which extracts global meteorological prior knowledge from the data operated by nature history through a more accurate and efficient learning and reasoning mechanism, replacing the timing solution of ultra-large-scale partial differential equations in traditional scientific computing, so as to quickly complete the global weather forecast for the next 1 hour to 7 days, and the prediction accuracy is more than 20% higher than that of the European Medium-term Weather Forecast Center.

In the pharmaceutical field, drug target discovery is a key issue in the drug development process. HUAWEI CLOUD Pangu Drug Molecule Model learns the chemical structure of 1.7 billion small molecules, deeply characterizes small molecule compounds, efficiently generates new drug molecules, calculates protein target matching, predicts the biochemical properties of new molecules, and quantitatively optimizes the screened lead drugs to achieve AI-assisted drug design throughout the process.

During a visit to Huawei's exhibition hall, a reporter from Nanduwan Finance Agency learned that the HUAWEI CLOUD Pangu drug molecule model has greatly improved the efficiency of drug design, reduced the R&D cost in the lead drug design stage by 70%, and shortened the R&D cycle from several years to one month.

It is reported that the Pangu model has been verified in more than 100 industry scenarios, including energy, industry, finance, medical, retail, logistics and other fields. Based on the Pangu model, Huawei has successfully trained a series of large models, including HUAWEI CLOUD Pangu NLP Chinese large model, HUAWEI CLOUD Pangu Computational Vision Model, HUAWEI CLOUD Pangu Drug Molecule Large Model, HUAWEI CLOUD Pangu Scientific Computing Big Model, Pengcheng Pangu NLP Chinese Large Model, and Pengcheng Biopharmaceutical Large Model.

Written by: Cheng Yang, reporter of Nanduwan Finance Agency