The path to implementing AI is guided by cloud vendors.
Nowadays, generative AI is experiencing explosive growth, with increasingly rich and diverse applications in artificial intelligence, such as text generation, image generation, knowledge Q&A, product design, and more innovations emerging continuously. However, the question arises: how can we better promote the implementation of generative AI applications and reshape future industrial value?
As a key player in industrial innovation and capability building, cloud vendors need to leverage their capabilities. After all, the widespread adoption of cloud computing during the digital transformation process has provided a new upgrade path for many enterprises constrained by IT infrastructure; today, it is evident that cloud vendors are comprehensively reconstructing the foundation of generative AI, laying a solid foundation for application implementation.
In Gartner’s “2023 Cloud AI Developer Services Magic Quadrant” report, Amazon Web Services has been listed as one of the “leaders” for the fourth consecutive year, ranking highest in the execution capability dimension. In the changing landscape led by generative AI, Amazon Web Services has also taken the lead, accumulating a number of exemplary implementation cases across various industries, setting a new benchmark for enterprise-level generative AI advancement.
Breaking Data Silos, Building Enterprise-Level Intelligent Knowledge Bases
Innovation is a key factor for the resilient growth of enterprises. Siemens, as a long-established leader in electronic and electrical engineering, has long had insights into this. Discovering data value has become a crucial direction for their continued efforts.
The Deyu team at Siemens China, responsible for IT data analysis and artificial intelligence, has explored building an enterprise knowledge base. This not only facilitates better excavation and protection of enterprise knowledge assets but also allows employees to quickly and accurately access relevant knowledge during business activities, enhancing their professional capabilities and making enterprise knowledge “live,” thus increasing knowledge reuse rates.
However, during the building process, the Deyu team gradually discovered that the retrieval and utilization of internal resources had long been plagued by issues such as structural disarray, slow retrieval speeds, and inconvenient interactions. Moreover, given that it spans multiple different fields and involves various business units, traditional methods of building a knowledge base would always encounter these problems. Therefore, the Deyu team decided to apply a large database and generative AI to a new “intelligent knowledge base,” fundamentally enhancing the usability of the knowledge base.
In response to this demand, Amazon Web Services, with a long-standing good cooperation foundation, was once again the first choice for Siemens. After in-depth research and comprehensive consideration, Amazon Web Services ultimately provided the Deyu team with a smart knowledge base and intelligent conversational robot solution guide, capable of achieving about 80% of the intended knowledge base functionalities, while Siemens China would customize the remaining 20% based on internal needs, ultimately forming a complete solution.
The overall solution has three major highlights:
First, it adopts a “RAG architecture + vector database” design, where the core knowledge base is constructed in a vectorized manner, capable of storing extremely large-scale vector data. Additionally, the RAG architecture ensures that the knowledge base possesses near-infinite scalability without affecting access speed, greatly expanding the usability of large models;
Second, the serverless nature of the open-source search engine Amazon OpenSearch Service allows developers to avoid managing clusters or worrying about production scale, enabling rapid deployment;
Third, the machine learning service Amazon SageMaker provides a rich set of model development and training tools, ensuring that customers can easily fine-tune large language models in the cloud and test various types of open-source models.
Ultimately, with the support of Amazon Web Services, the Deyu team completed the development, deployment, and launch of the generative AI conversational robot “Xiao Yu” in just three months. Compared to traditional robots, “Xiao Yu” not only generates responses faster but also has a higher hit rate for search keywords, providing a superior user experience. After the launch of Siemens China’s exclusive intelligent knowledge base, over 4000 internal users participated in its use in the first week, with more than 12000 questions posed and answered.
Accelerating Online Education, Triggering a Complete Transformation of Teaching Models
Throughout history, every technological revolution has profoundly changed educational forms.
Generative AI is also triggering deep changes in educational teaching models.
As a vehicle for promoting digital transformation at Foreign Language Teaching and Research Press, FLTRP Online is actively investing in the construction of generative AI to achieve intelligent enhancement of foreign language teaching. However, conducting extensive exploration and testing of generative AI requires flexible, easily deployable, and efficient underlying infrastructure support.
Therefore, the artificial intelligence team at FLTRP Online first conducted a detailed assessment of the impact and challenges of generative AI on their business; furthermore, they hope to conduct extensive evaluations and trials of mainstream foundational large models on the market to select the most suitable foundational model tools for their teaching environment. Meanwhile, to meet long-term continuous reasoning demands, FLTRP Online also needs to carry out data cleaning, model customization, quantization, fine-tuning, and compression optimizations to reduce the usage costs of large language models.
Considering the above needs, Amazon Web Services, which has years of successful cooperation experience with FLTRP Online, stood out. When expanding the application of generative AI, FLTRP Online also chose to build on the infrastructure and product services of Amazon Web Services.
The fully managed machine learning service Amazon SageMaker provides enterprises with a complete set of tools and frameworks, including data labeling, model training, model deployment, and automated modeling functionalities. This end-to-end model deployment solution allows for secondary development and fine-tuning of foundational models, facilitating model building, training, and deployment for all developers.
When building the generative AI platform, FLTRP Online adopted Amazon SageMaker for fine-tuning large models, while the data processing workload was handled by Amazon EC2 cloud servers, processing third-party open-source datasets and using Amazon S3 cloud storage to save various data and model files of AI models, thus gaining the capability to build generative AI applications on large models more conveniently.
For the large model deployment issue that FLTRP Online is particularly concerned about, Amazon Web Services provides rich professional technical support resources. Its professional services team collaborates closely with FLTRP Online, assisting in evaluating almost all large models and generative AI tools on the market, and completing multiple POC tests for different application scenarios. At the same time, several large models were deployed for testing and comparison using Amazon SageMaker, demonstrating real-time inference and running batch inference tasks, making complex testing simple and efficient, ensuring that the FLTRP AIGC platform goes live on time.
With the assistance of Amazon Web Services, FLTRP Online has launched new services such as iTEST, iWrite, and iTranslate, providing intelligent feedback and translation polishing, achieving new learning methods such as project-based learning, human-computer interactive co-creation learning, infinite personalized learning, and immersive learning, helping students gain a better learning experience while also reducing the workload of teachers, achieving the goal of improving teaching efficiency and quality.
Enhancing Service Upgrades, Deepening Cross-Border E-Commerce Business Scenarios
For many years, the e-commerce field has been exploring paths to achieve cost reduction and efficiency improvement with better technological tools. With the breakthrough development of generative AI, marketing content and images generated by artificial intelligence have also ushered in an explosive period in the e-commerce field.
Dianjiang Technology is a SaaS enterprise-level technology company focused on global independent station sales, concentrating on cross-border e-commerce scenarios. In conversations with merchants, Dianjiang Technology found that clothing and hat products update rapidly, requiring high-quality content materials and facing significant pressure in image production. This is because the shooting and processing of material images involve not only creative, copyright, and cost issues, but also the diversity of different countries’ consumers, models, and product displays.
To better serve clothing merchants, and address the pain points of high costs for material image production and the need for rapid listing, Dianjiang Technology decided to launch an AI-generated image application—BetaCreator.
However, to develop this new application, Dianjiang Technology faced several challenges: First, the rapid development of the e-commerce industry makes it time-consuming and labor-intensive for enterprises to build and deploy models, and flexibility is hard to achieve; Second, AI-generated images have certain algorithmic and engineering thresholds, and enterprises lack specialized algorithm personnel and mature practical experience; Third, the user experience of the new application is crucial, requiring innovative and user-friendly AI technology to meet and optimize these needs.
Considering these factors, Dianjiang Technology ultimately chose to tackle these challenges in collaboration with Amazon Web Services.
Through their joint efforts, Dianjiang Technology easily hosted their model on Amazon SageMaker, and based on this, quickly built BetaCreator, completing experiments in various application scenarios such as model generation, face-swapping, product variations, background changes, and creative hits. This not only helped merchants quickly achieve automated generation and processing of e-commerce material images but also ensured that the generated e-commerce material images exhibited vivid and realistic expressiveness in terms of details, quality, and coherence.
Additionally, the Amazon SageMaker JumpStart machine learning center provided Dianjiang Technology with hundreds of built-in algorithms and pre-trained model libraries, accelerating model building and deployment. At the same time, Amazon SageMaker JumpStart also offers responsible AI technology with features such as content moderation, automatically censoring any inappropriate input or generated content to reduce risks.
From the conception of the product prototype to the final launch, the technology team at Amazon Web Services provided comprehensive technical support and services to help Dianjiang Technology overcome technical difficulties.
Overall, based on this foundation, the constructed BetaCreator not only saved merchants time in testing different material images for advertising placements but also significantly improved the output efficiency of e-commerce material images, providing yet another representative example of the implementation of generative AI in the e-commerce field.
In the above cases, generative AI has continuously integrated with various industries, becoming a powerful force in reshaping key industries of enterprises. But it is worth pondering, if AI truly redoes all applications and products, can we imagine the changes it would bring?
The gap between prospects and reality is not to be understated. Currently, there are indeed insurmountable gaps in the implementation of generative AI across various industries. To combine artificial intelligence with business, enterprises still need to strive forward to possess this capability.
However, this year, generative AI has surged like a hurricane, and Amazon Web Services, as a global leader in cloud computing, is currently making technical route choices and practical paths that are prompting it to become a partner for more enterprises to implement generative AI strategies. The creation of numerous cases proves this. In the future, enterprises across various industries may gain insights from these paradigm cases, using them as references to form their own generative AI cloud maps.
