From quickly generating logically clear lengthy articles to producing video clips without the need for video materials, large models have developed rapidly in recent times. In addition to applications such as chatbots, cultural graphics, and coding, how to further integrate large models with the industry and how to have more industry stickiness are still issues that need to be explored in the industry. "It has become a consensus that big models will empower various industries, but how to integrate them with the specific business of industries and enterprises in concrete practice still needs continuous exploration." At the recently held Amazon Cloud Technology Generative AI Media Communication Conference, Chen Xiaojian, General Manager of Amazon Cloud Technology's Greater China Product Department, believed that in the ever-changing scenarios, various industries need to make good use of the powerful technical capabilities of big models, and enterprises need a series of peripheral capabilities to use big models correctly, reasonably, safely, and efficiently. To obtain a large model that can be applied to various industries, it is necessary to first have a strong basic model, and then carry out relevant training combined with business scenarios. To this end, Amazon Cloud Technology has partnered with American artificial intelligence company Anthropic to train the Claude3 series models. Currently, the model exhibits excellent understanding ability in complex tasks. But having only a basic model is not enough. Although large models have strong capabilities, their application scenarios and tasks are constantly changing. Customized tuning is necessary before using a large model. If the basic large model is simply used, it will be difficult to achieve optimal results. Training models is the only way to increase the stickiness of the large model industry. "When selecting application scenarios for training large models, the training end needs a cluster with sufficient scalability." Cui Wei, Director of Data Analysis and Generative AI Products at Amazon Cloud Technology Greater China, suggests that a reliable, secure, and resilient environment is needed for industry large model training. After training, sufficient computing power services need to be provided in the cloud for large model operations. In the process of implementing industry models, the guarantee of talent teams is also crucial. Chen Xiaojian stated that even with a sound data foundation and good industry training, industry big models still cannot meet all the needs of the industry. This requires professional teams to provide business support, such as solution architects and business personnel jointly searching for application scenarios and solutions, and product technology experts fine-tuning the output mode of large models based on specific needs. (Lai Xin She)