RM新时代有限公司

On December 7, the 10th Plenary Meeting of the Artificial Intelligence Industry Alliance (AIIA) and the 2023 General Artificial Intelligence Innovation and Development Forum (hereinafter referred to as the Forum) was held in Chongqing Liangjiang Collaborative Innovation Zone - Mingyue Lake.

At the conference, Li Xiafeng, Supervisor of the Board of CloudWalk Technology, delivered a sharing on "CloudWalk CongRong Large Model All-in-One Solution Based on Domestic Computing Power".

As a software-hardware integrated product of CloudWalk Technology, the CongRong Large Model Training and Inference Appliance is fully compatible with domestic chips, featuring a built-in integrated large model production platform with fully independent and controllable software and hardware. A single unit can achieve 2.5 PFLOPS of computing power and a training speed of 8000 tokens/s, enabling private deployment for enterprises. It supports flexible networking to achieve a closed loop of large model training and inference, helping enterprises establish an information, digital, and intelligent business value closed loop, significantly improving business efficiency and information security. It can be widely applied in industries such as smart transportation, smart finance, smart commerce, smart factories, and smart government.

The conference also released the list of Chongqing's AI application scenario projects and software products. CloudWalk Technology's Digital Human Service Platform and CongRong Basic Large Model were selected for the list due to their leading technology and application achievements.

The CongRong Basic Large Model is a multimodal large model built by CloudWalk around the strategic goal of "human-machine collaboration". It can handle input and output of language, images, and video, representing the latest trend and progress in the field of large models—multimodality.

In addition, CongRong uses a large amount of unsupervised learning data for pre-training, allowing knowledge sharing across different tasks, thereby improving performance.

If the vision of artificial intelligence is to create a robot with a brain, nerves, torso, and limbs, then the hardware foundation is the torso and limbs. More importantly, to enable the robot to see and think, hear and speak, it is necessary to build nerves and a brain.

Multimodal large models are the core technology for building the "brain".

Undoubtedly, multimodality will greatly expand the application scenarios of AI.

CloudWalk is actively promoting the integrated application of AI agents, vision, and cross-modal large models, as well as industry large model cases in various sectors such as government, manufacturing, and finance. Examples include the world's first port large model PortGPT, jointly developed with Tianjin Port Group, Zhejiang Seaport Group, Huawei, and Tianjin Supercomputing Center; the Zhixing Urban Transportation Industry Large Model released with PCI Technology and Chongqing City Transportation Development & Investment Group; the CongRong Large Model Training and Inference Appliance launched with Huawei Ascend, laying the foundation for private deployment of large models. Meanwhile, CloudWalk Technology's document multimodal large model and medical large model have begun to be applied in the finance and pharmaceutical industries.

CloudWalk Technology's Digital Human Service Platform is a digital human production line based on multimodal large models. The AI agent (AI-Agent) Yunyue created on this platform not only has a high degree of human-like presentation, being closer to real human levels in appearance, movement, and intelligence, but can also understand, see, remember, self-learn, and interact naturally with people.

AI agents can serve as "digital employees", coordinating control and data flow between multiple different applications and systems. Their application scenarios are extensive, covering fields such as government, manufacturing, finance, education, entertainment, and customer service.

For example, the Tianjin Port large model initially plans six major application scenarios: port production, logistics services, customer management, customer service, video analysis, and office assistance. The AI agent "Tiantian", created based on the Digital Human Service Platform, is already capable of serving port enterprise production handover meetings and answering general knowledge questions about Tianjin Port.

The inclusion of these two products holds significant meaning and impact for the entire industry, marking that large model technology has moved from pure language models into a new multimodal stage in industrialization.

In addition, the successful application of large model technology will encourage more upstream and downstream enterprises to join the wave of intelligent transformation, driving the development and progress of the entire industry.