Chuangda: Riding on the cloud to control intelligence, generative AI reshapes the manufacturing industry with the 'OS+AI' dual-engine power to start a journey

Currently, generative artificial intelligence is entering all walks of life at an unprecedented speed. How does the manufacturing industry need to deepen its digital capabilities? It will combine digital technology with artificial intelligence as the core and business, and use advanced data analysis solutions to improve the company's capabilities. Operational level and continue to promote business changes? Chuangda provides end-to-end artificial intelligence solutions, as well as a technology "base" that runs through end, edge, and cloud, providing "China speed" for the manufacturing industry to move into the era of large models!

Yang Xinhui is the vice president of Chuangda Internet of Things Business Group

“In the next 15 or even 30 years, the development of general artificial intelligence driven by large language models will become one of the biggest needs. All products can be made again with general AI, and all processes can be redefined. ." Yang Xinhui, Vice President of Thunderstar Internet of Things Business Group, believes that although the big language model is a new wave of technological development, it cannot just seize a wave of short-term dividends, but must look at this industry from a long-term perspective.

Zhongke Chuangda is an operating system company founded in 2008, focusing on B2B business. Initially, the company provided smartphone operating system services, but starting in 2014, the company began to expand into the field of smart cars. In 2015, it further expanded into smart hardware and gradually developed into the entire smart industry. The company is committed to improving the capabilities of manufacturing enterprises in production continuity, business global expansion, technological innovation and safety compliance, and providing them with continuous support

Currently, Chuangda is committed to using "OS AI" as a new engine for two-wheel drive growth. In this process, Chuangda will leverage its strength and cooperate with partners like Amazon Cloud Technology to embark on a journey of harnessing intelligence through the cloud.

Behind dual-engine driven growth

Zhongke Chuangda has a deep understanding of the development needs and trends of traditional manufacturing. However, if it wants to serve the manufacturing industry well and fully penetrate into the manufacturing value chain scenario, without a strong technical base, it is obviously unable to meet the digital challenges of manufacturing enterprises.

"Nowadays, a big obstacle that affects the implementation of large language models is the cost of inference, which is a key factor affecting the implementation of products. Training a general large language model with hundreds of billions of parameters requires close to 80 to 90% of human knowledge. , but it is actually redundant in a certain vertical scenario and will also cause a lot of reasoning costs, so it is very important to find a balance between model accuracy and reasoning costs." Yang Xinhui said in the interview that for specific vertical Making the most efficient product selection based on the scenario is the development direction for companies to solve manufacturing problems, and it is also one of the directions for cooperation with Amazon Cloud Technology.

As an Amazon cloud technology partner, the two parties have been cooperating since 2017 to jointly develop industry solutions in artificial intelligence. In 2020, Chuangda became the first partner to use the machine learning service Amazon SageMaker in the Amazon Cloud Technology China region, and integrated Amazon SageMaker into the Chuangda Smart Industrial ADC (Automatic Defect Classification) system, allowing the manufacturing industry to Customers can easily obtain AI quality inspection capabilities in industrial production. At the 2022 re:Invent global conference, Chuangda became the first partner in the world to obtain the Amazon SageMaker Service Readiness Plan, allowing its TurboX Inspection industrial vision platform to better provide intelligent upgrade services to many global Amazon Cloud Technology customers. In 2023, Chuangda used the technology and services of Amazon Cloud Technology to develop a large language model "Rubik's Cube Model". Through large-scale training data and complex neural network structures, this language model can more accurately understand and generate human Language, with extensive natural language processing capabilities.

In May 2023, Amazon Cloud Technology and China Science and Technology Thunder announced the establishment of a joint artificial intelligence innovation laboratory, which marked an important step in the cooperation between the two parties. With the professional resources of Amazon Cloud Technology and China Science and Technology Thunder, the two parties will carry out in-depth cooperation and exploration in the fields of generative artificial intelligence, enterprise knowledge base, digital transformation assistant, and large-scale language model empowered smart devices

The joint innovation laboratory covers the complete user experience process from creativity, design, evolution, implementation to optimization. The two parties will jointly explore innovative experience scenarios and sort out the user experience journey. At the same time, we introduce the latest AI technology to build industry-wide model design, development and application methods, develop AI model prototypes in innovative business scenarios, and quickly iterate into digital tools to create new industry solutions and help more industry customers accelerate cloud computing. digital transformation and business innovation journey.

Continuously iterate and accelerate the evolution of generative AI

Take the automotive industry as an example. The automotive scene is the most friendly scene for large language models. In the closed environment of the car, using the multi-round language dialogue and semantic understanding capabilities of large language models can significantly improve the car experience

For example: During the R&D process, concept drawings of some products can be quickly constructed through large language models. Only a short-term closed-loop communication with the customer is required, and the communication process of the previous R&D cycle and conceptual design stage can be transformed from Two or three months was reduced to one or two weeks.

For example, the combination of car manuals and large models is a very popular feature among car owners. Car owners no longer need to read hundreds of pages of car manuals, they only need to obtain some basic usage tips on the car through simple questions and answers

In addition, large language models can also be intelligently applied to factory production scenarios to further optimize the entire production and manufacturing process. On the production side, this large-scale model can significantly improve the quality control efficiency of the factory, reduce the work burden of general staff, and comprehensively improve their work efficiency

Faced with the needs of the world's leading Fortune 500 logistics companies, Thunderstar cooperated with Amazon Cloud Technology to jointly explore how to use large-scale language models to solve basic scenario applications within logistics groups, such as intelligent Q&A and proof-of-concept for corporate employees. Build

Zhongke Chuangda started working on expanding the model scale at the end of last year and continued until the beginning of this year. Although it started late compared with those companies that train tens of billions or hundreds of billions of parameters, judging from its application, its development has just begun

Yang Xinhui emphasized that we will gradually shift from a direction with relatively small computing power requirements to a direction with increasing computing power requirements, and will not invest in expensive hundreds of billions of model training when capacity is insufficient. This requires a process. It is expected that by the end of next year, we will complete model training with a scale of more than 100 billion. He also said that large-scale language models require continued investment, and the company will increase investment and seize new development opportunities

In the past, all businesses revolved around the end side and edge side. However, the end and edge sides cannot run models with hundreds of billions of parameters. The difference between Chuangda and other model manufacturers is that they hope to form their own core capabilities and competitiveness on the end side and edge side in the future, and apply 7B and 13B models to the end side and edge side

For example: In cars, large models are directly used as edge-side products. How can we achieve local model operation with the maximum computing power without the need for networking? If a technical practice path is found, for the majority of car owners, it will be It will bring new experience upgrades.

The question is, the manufacturing industry has huge demands, and facing customers from so many industries including automobiles, how can Chuangda take into account all aspects and ensure that the individual needs of each customer are met?

Yang Xinhui said that the fundamental reason why we chose to cooperate with Amazon Cloud Technology is that as a global leader in cloud computing, it has extensive and deep cloud services. Amazon Cloud Technology has a very clear layout and insight into the industry, understands the pain points and difficulties of the industry, and knows the key to the entire industry

For Chuangda, joint innovation with Amazon Cloud Technology is a process of growing capabilities. Amazon Cloud Technology has a team of scientists who are all very senior algorithm experts and have made many high-quality models in large language models, such as database retrieval, refined image recognition and segmentation, etc. These technologies and capabilities can well empower innovative artificial intelligence solutions and can jointly create some productized services for customer implementation.

In cooperation with Amazon Cloud Technology, we focus on upper-layer applications and do not need to worry about underlying infrastructure issues. Amazon Cloud Technology has a global infrastructure and global business system. Chuangda's business is also developing globally. Our R&D centers are located in 15 countries or regions around the world, and overseas revenue accounts for more than 30% of total revenue. Most of our customers are global enterprises, and almost all of them are using products within Amazon’s global business system

With its global business layout, Zhongke Chuangda integrates advanced technology with the industrial chain and introduces overseas superior technologies into domestic enterprises. At the same time, the company is also committed to serving domestic leading technologies to overseas enterprises and truly realizing "rooted in China and empowering the world"

The above is the detailed content of Chuangda: Riding on the cloud to control intelligence, generative AI reshapes the manufacturing industry with the 'OS+AI' dual-engine power to start a journey. For more information, please follow other related articles on the PHP Chinese website!