AWS announced at re:Invent 2023 a custom processor for AI model training.Trainium 2, and also updated its own processor Graviton4 designed with Arm architecture, and announced the launch ofA chatbot named "Q", simplifying workflows through automatically generated artificial intelligence while also enhancing the human-computer interaction experience.
Trainium 2 boasts a quadruple improvement in inference performance and a two-fold improvement in power efficiency, accelerating AI model training and reducing overall training costs. Anthropic, an AI technology company invested in by Amazon, has already decided to introduce Trainium 4 to accelerate its AI model training.
The Graviton4, designed for general-purpose artificial intelligence operations, is based on the Arm architecture and boasts lower power consumption than Intel and AMD processors. It also boasts a 30% improvement in overall execution efficiency, allowing businesses to reduce costs and handle larger amounts of data and workloads.
Graviton4 will currently be available in preview form and is expected to be made available to more enterprises in the coming months.
As for the chatbot "Q" launched by AWS at the same time, it is a digital assistant service that combines automatically generated artificial intelligence. It allows enterprises to customize its design and even change its original code. After connecting to the company's internal systems or integrating Microsoft 365, Dropbox, Salesforce, Zendesk and other services, it can serve as a "personal assistant" to assist internal work in the company, helping corporate staff find suitable information more quickly or assist in arranging work tasks.
AWS previously stated that it would expand its development of artificial intelligence technology. In addition to strengthening various services by introducing automatically generated artificial intelligence, it also continues to update its customized processors. On the other hand, it also maintains cooperation with processor suppliers such as NVIDIA, thereby widening the competitive gap between itself and Microsoft Azure, Google Cloud, and even Alibaba and Tencent in the cloud application service market.
