Amazon Cloud Technology fully utilizes generative AI technology to further improve the cloud computing platform

Generative artificial intelligence has become a battleground for cloud service providers, and Amazon Cloud Services, the leader in the global cloud computing market, is also comprehensively promoting generative artificial intelligence

At the 2023 re:Invent global conference, Amazon Cloud Technology announced a series of new services and features, including the launch of underlying infrastructure, generative artificial intelligence (AI), and data strategy. These new services and features include Amazon Q, a new generative AI assistant designed to reshape the future of work; Amazon Bedrock, offering more model choices and new powerful capabilities; and Amazon SageMaker with five new features , assisting in large-scale development of application models. The launch of these services and functions helps enterprises build and apply generative AI more easily and securely

Chen Xiaojian, General Manager of Amazon Cloud Technology Greater China Product Department said: "Amazon Cloud Technology will release many new services, new functions and new applications at the annual re:Invent global conference. In terms of infrastructure, , computing, storage, data and other fields continue to reshape cloud computing, and launch blockbuster new services and functions around today's most transformative technology, generative AI. We hope that through these technological innovations, we can help more companies accelerate innovation and take advantage of Generative AI comprehensively reshapes the future.”

Amazon Cloud Technology 2023 re:Invent China City Tour officially starts today and will be held in 10 cities including Beijing, Shanghai, Guangzhou, Shenzhen, Chengdu, Qingdao, Nanjing, Xi'an, Hangzhou, and Changsha. This tour aims to provide Chinese builders with a comprehensive display of the latest services and technologies, cutting-edge trends and best practices at the 2023 re:Invent Global Conference

1. Fully develop generative AI

Amazon Cloud Technology provides a three-tier architecture for generative AI, including applications built using basic models, tools built using basic models, and infrastructure for basic model training and inference.

At the bottom level, Amazon Cloud Technology provides infrastructure for basic model training and inference through self-developed chips.

Amazon Trainium2 processor is a dedicated chip for generative AI and machine learning training. It is optimized for training basic models with hundreds of billions to trillions of parameters. Compared with Amazon Trainium, it has a 4x performance improvement and 65 exaflops. On-demand supercomputing performance; Amazon SageMaker HyperPod service can accelerate basic model training on a large scale, shorten training time by up to 40%, and ensure uninterrupted training processes that last for weeks or months.

Amazon Cloud Technology and NVIDIA jointly announced several latest cooperation, which is what needs to be rewritten

• Amazon Cloud Technology will provide the first cloud AI supercomputer equipped with NVIDIA Grace Hopper super chip and Amazon Cloud Technology UltraClusters technology; the first NVIDIA DGX cloud using NVIDIA’s latest chip GH200 NVL32 will soon log in to Amazon Cloud Technology; the two companies jointly Launch the "Project Ceiba" cooperation project to use the world's fastest GPU-driven AI supercomputer and NVIDIA DGX cloud supercomputer for NVIDIA AI training, research and development, and customized model development. It will have 16,000 of the latest GH200 super chips , providing an astonishing computing power of up to 65 ExaFLOPS.

Amazon Cloud Technology provides middle-tier tools that can be built using basic models

Amazon Bedrock is the easiest way to build and scale generative AI applications with large models. Amazon Bedrock supports Anthropic Claude 2.1 and Meta LLama 2 70B, as well as the Amazon-exclusive Amazon Titan model.

Rewritten content: The key to creating real value for generative artificial intelligence applications is to be customized based on the company's own data. Only through data customization can the company's differentiated competitive advantage be established. Amazon Cornerstone has three major functions: continuous pre-training, fine-tuning and knowledge base retrieval enhancement, and provides a preview function

With models and customization capabilities, they also need to be integrated with applications to serve the business. As such, Amazon Bedrock provides agent capabilities that enable generative AI applications to perform multi-step tasks across company systems and data sources.

Guardrails for Amazon Bedrock Preview, protect generative AI applications with responsible AI policies. At the same time, Amazon Bedrock ensures data security and privacy: No customer data will be used to train the underlying model; all data is encrypted during transmission and at rest; data used for custom models remains with you Within the VPC; supports standards such as GDPR and HIPAA.

At the top application layer, Amazon Cloud Technology provides applications built using the basic model-Amazon Q preview version.

Amazon Q is a new type of generative AI-powered assistant that can be customized according to customer business and is specifically designed to meet the needs of office scenarios. Customers can quickly get relevant answers to complex questions, generate content and take action, all based on insights from their own information repositories, code and enterprise systems. Additionally, customers’ content is never used to train Amazon Q’s underlying models. Amazon Q can be built on Amazon Cloud Technology, or it can use on-premises data and systems, using Amazon Cloud Technology applications for business intelligence (BI), contact center and supply chain management. Amazon Q is already available in preview to customers, Amazon Q in Amazon Connect is officially available, and Amazon Q in Amazon Supply Chain is coming soon.

The success of generative AI is inseparable from strong data support. At the 2023 re:Invent global conference, Amazon Cloud Technology launched a number of services and features covering data infrastructure, integration and governance.

First of all, to further enrich the selection of vector databases, Amazon Cloud Technology launched the Amazon OpenSearch Serverless vector engine, the new vector search functions of Amazon DocumentDB and Amazon DynamoDB, and the preview version of Amazon Memory DB for Redis vector search, improving Performance of generative AI applications in terms of response and latency.

Launched four Zero-ETL integration features to make data access and analysis across data storage faster and more convenient.

In terms of data governance, Amazon Cloud Technology has launched a preview version of the AI ​​description suggestion function for Amazon DataZone, which can automatically generate a more understandable business description for an enterprise's data set and provide information about the data set. Recommendations.

2. Reshaping cloud computing - self-developed chips, storage, serverless

Amazon Cloud Technology released the independently developed Amazon Graviton4 and Amazon Trainium2 chips at the 2023 Global Conference

Compared with the current generation Graviton3 processor, Graviton4 has a performance improvement of up to 30%, more than 50% more independent cores, and more than 75% increase in memory bandwidth, providing the best possible performance for workloads running on Amazon Elastic Compute Cloud (Amazon EC2). Optimal performance and energy efficiency; Graviton4-based Amazon EC2 R8g instances are currently available in preview. Through cooperation with Sinnet and NWCD, Amazon EC2 C7g, M7g, and R7g instances based on Graviton3 processors are now officially available in Amazon Cloud Technology China (Beijing) Region and China (Ningxia) Region.

The Trainium2 chip is specifically designed for high-performance training, which is suitable for base models and large language models with trillions of parameters or variables. Compared with the first-generation Trainium chip, Trainium2 performance has been improved by up to 4 times, memory has been improved by 3 times, and energy efficiency (performance per watt) has been improved by 2 times. Amazon EC2 Trn2 instances use the latest Trainium2 chips, and each individual instance contains 16 Trainium acceleration chips. Trainium2 instances can be expanded to up to 100,000 Trainium2 acceleration chips, integrated with Amazon Elastic Fabric Adapter (EFA) PB-level network interconnection, providing up to 65 exaflops of computing power. Customers can get supercomputing-level performance on demand

The second new product launched by Amazon Cloud Technology is storage service

Since its launch 17 years ago, Amazon Simple Storage Service (Amazon S3) has become one of the most popular cloud storage services, with millions of customers across the world from all walks of life. At this conference, Amazon Cloud Technology announced that Amazon S3 Express One Zone is officially available. Compared with Amazon S3 Standard, the data access speed is increased by up to 10 times and the data request cost is reduced by 50%, providing machine learning training, inference, and interactive analysis. and request-intensive workloads such as media content creation to provide the highest performance storage.

The last new product is serverlessServerless.

Amazon Cloud Technology pioneered serverless technology 17 years ago, providing customers with ultimate elasticity and automatic expansion capabilities. At the 2023 re:Invent global conference, Amazon Cloud Technology launched three serverless service innovations to help customers analyze and manage data at any scale and significantly simplify operations. Customers do not need to spend time and energy to configure, manage and expand their data foundation. facility.

The rewritten content is as follows: Among them, Amazon Aurora Limitless can automatically distribute and query data across multiple Amazon serverless instances, and can scale to millions of transaction-level writes per second , manage petabyte-level data. Amazon ElastiCache Serverless can help customers create highly available caches in a minute and scale vertically and horizontally in real time to support customers' complex applications without the need to manage infrastructure. Amazon Redshift Serverless uses artificial intelligence (AI) to predict workloads and automatically scale and optimize resources to help customers achieve cost-effective goals

The above is the detailed content of Amazon Cloud Technology fully utilizes generative AI technology to further improve the cloud computing platform. For more information, please follow other related articles on the PHP Chinese website!