Hot out of the circle! ChatGPT brings an industrial whirlwind, Ziguang Digital specially launched NVIDIA AI computing product matrix.


ChatGPT is an AI dialogue model developed by OpenAI in the United States in November 2022. Due to its excellent text generation and dialogue interaction capabilities, it has rapidly become popular worldwide, with more than one million active users in 5 days and more than 100 million active users in 2 months.

ChatGPT is essentially an AI model with a large number of complex computing requirements. It can be intelligent by connecting a large number of real corpora to carry out in-depth and multi-stage training and learning. It still belongs to the application category of AIGC (production mode using AI technology to automatically generate content). Some experts pointed out that the continued popularity of ChatGPT has brought new changes to AIGC. The profound impact and changes it has triggered are based on computing power, and higher requirements are put forward for computing power support required for AI model training.

ChatGPT requires a large number of high-performance GPU computing power for deep neural network training. It supports stand-alone and multi-machine environments and can take advantage of multi-GPU training to improve computing power. ChatGPT also require a good network environment for large-scale data distribution, while taking up a lot of storage space to store training data and model parameters. In addition, ChatGPT need an advanced IT infrastructure to support the management, visualization and optimization of the training process.

There are already many companies in the country that have developed into ChatGPT and AIGC. It is reported that Baidu's Wen Xin will directly access the search. Jingdong plans to integrate ChatGPT methods and technical points into the product service. The Dharma Academy version of Ali's internal test is ChatGPT having a man-machine dialogue in nailing. Tencent continues to invest in previous technical reserves in AI big models, machine learning algorithms, NLP and other fields; byte Beat's Artificial Intelligence Laboratory (AI Lab) has carried out research and development related to ChatGPT and AIGC, and may provide technical support for PICO in the future.

At present, domestic companies are moving closer to concepts such as virtual humans and AIGC, but there are still some bottlenecks in technological development. The reason is five:

1) Lack of original models in China, no model iteration accumulation. ChatGPT depends on InstructGPT,InstructGPT on GPT-3.5 and GPT-3.

2) Domestic real data volume. In addition to Baidu has natural user search question and answer training samples, for other companies may be relatively lacking.

3) Lack of domestic technology accumulation. In the process of ChatGPT development, there are technical difficulties in data processing, cleaning, labeling, model training, reasoning acceleration and so on, and have a great impact on the results. And including domestic large factories, the intensive learning framework has not yet appeared in large-scale use scenarios.

4) Domestic innovative soil needs to be developed. The overall business environment is impatient, but inputs and outputs take some time.

5) As we all know, at present, our computing platform has obviously lagged behind:

• Hardware platform computing power is limited

• Model maturity and scalability, robustness

Cluster management is difficult, debugging is difficult, research and development results are difficult to convert into production applications.

• Data parallel, model parallel and other multi-machine multi-card management challenges

It is difficult to adjust IT infrastructure construction, IDC management, visualization, operation and maintenance, etc.

Under the demand of high computing power, more requirements are put forward for the computing and transmission capacity of supercomputing infrastructure, and expansion is imminent. The acceleration of the trend determination of enterprises on the cloud will precipitate more and more data in the cloud, which forms the basis of the AIGC outbreak, thus driving the related long-term benefits. At the application level, more companies will deeply integrate ChatGPT and other AIGC products with their respective vertical industries to produce greater utility, which will bring more opportunities to the industry chain in the long run.

 

>>>>Basic layer: computing power and cloud computing chip pre-training model.

• Data center chips: massive data storage and computing, NVIDIA AMD MRVL ......

• Mobile computing: Qualcomm, MediaTek (mobile XR, car, etc.)......

• Pre-trained models: Microsoft, Google, Meta, NVIDIA ......

 

>>>>Tool layer: verticalization, scenario, personalized models and tools

• Modeling tools: Adobe, Autodesk, Unity ......

• AI tools: C3AI, Microsoft, NVIDIA, Google......

Massive data processing: Rapids, snowflake, cflt ......

Automation and process reengineering: Uipath(RPA and AI deep integration)......

 

>>>>Application layer: AI industry application

C- end and B- end taps with a large number of data portals

• Autopilot: NVIDIA, Qualcomm......

• Big Data: Plantir, AYX ......

• Financial sector: Lemonade, Upstart ......

• ITSM 与网安:erviceNow 、crowdstrike、FTNT、Dynatrace、Splunk ......

 

Ziguang Digital has been rooted in the ICT field for more than 20 years, and has launched a series of AI computing products for the development needs of the digital economy and the intelligent era. This series of products covers all NVIDIA enterprise-class products, including DC GPU, DGX super AI server, professional graphics GPU, high-speed network products, cluster management and scheduling, and resource monitoring software. At the same time, it provides AI solutions such as NVIDIA AI Enterprise and AI-stack software.Ziguang Digital has strong project management and implementation capabilities, and has experience in the implementation and operation of many large AI projects. It will empower AI scenarios and applications in all walks of life with a series of AI computing products, and provide surging computing support for the development of new technologies.

It can be expected that AIGC will play an important role in the intelligent transformation. It will be widely used in various scenarios, from social media to advertising creativity, from games to film and television entertainment, from programming to deep writing, from graphic design to product industrial design... and more unknown application fields.In the future, Ziguang Digital will join hands with the industry to build a solid foundation of computing power, embrace the AI era with channel partners, focus on the development of AI technology, and share the new dividends of AI.

Special attention:

Part of the text is written by the class ChatGPT model.

Part of the content source: network comprehensive finishing

Picture material: thousands of library network

 

Recommended News


Gathering Potential Breakthrough and Sailing, Ziguang Digital Joins Hands with NVIDIA to Accelerate the Layout of the Next Station of Artificial Intelligence Development

In October 2022, Ziguang Xiaotong, a subsidiary of Ziguang Digital Group, signed an agency agreement with NVIDIA. Ziguang Digital officially became the general agent of NVIDIA Data Center GPU,NVIDIA DGX System, NVIDIA Professional Visualization, NVIDIA Virtual GPU and NVIDIA Network Products in China. Start a new chapter of cooperation with NVIDIA in the field of artificial intelligence.


Hot out of the circle! ChatGPT brings an industrial whirlwind, Ziguang Digital specially launched NVIDIA AI computing product matrix.

ChatGPT is an AI dialogue model developed by OpenAI in the United States in November 2022. Due to its excellent text generation and dialogue interaction capabilities, it has rapidly become popular worldwide, with more than one million active users in 5 days and more than 100 million active users in 2 months.


GTC23 | Huang Renxun's keynote speech will introduce the latest progress of AI computing, OpenAI and DeepMind founders reveal the secret of generative AI

The GTC 2023 Online Conference will be held March 20-23 and will host more than 650 sessions hosted by leaders in technology, business, academia and government.


NVIDIA artificial intelligence | GTC23 hot watch introduction!

The artificial intelligence audio program launched by NVIDIA China is on again. Here, you will witness the majestic power and infinite possibilities of artificial intelligence together with NVIDIA!


GTC23 | "Voice AI" Topic Club Speeches Capped

GTC23 | "Voice AI" Topic Club Speeches Capped 2023-03-20 09:34 In the past, when people mentioned synthetic speech, they often associated with monotonous robot phones and GPS navigation systems, and voice AI changed this situation. Voice AI enables people to talk to devices, machines, and computers to simplify and improve their lives, and it is being used to support virtual assistants, expand call centers, enhance AR experiences, improve medical experiences, and more.


GTC23 | Come to China AI Day to explore the cutting-edge AI applications of the Internet!

The NVIDIA GTC conference will be held from March 20 to 23, 2023. The world premiere of the keynote speech will start at 11:00 p.m. Beijing time on March 21, and the replay in China will start at 10:00 a.m. Beijing time on March 22.