Alibaba Cloud, the cloud unit of Alibaba Group, has unveiled a serverless iteration of its Platform for AI (PAI)-Elastic Algorithm Service (EAS). This solution is developed to provide individuals and enterprises with a cost-effective means of deploying and inferring models.
Alibaba Cloud has also integrated its vector engine technology into key offerings such as Hologres, Elasticsearch, and OpenSearch, streamlining access to large language models (LLMs) for the creation of tailored generative AI applications.
“Our technology updates underscore our commitment to empowering enterprises with the latest intelligence-driven solutions for heightened efficiency and performance,” said Zhou Jingren, chief technology officer, Alibaba Cloud.
Read:
Alibaba Cloud deploys cloud technology innovations at Asian Games
Alibaba Cloud open-sources more LLMs
The PAI-EAS platform enables users to leverage computing resources on-demand, eliminating the need for overseeing physical or virtual server management. This results in a 50% reduction in inference costs compared to traditional pricing models, offering a compelling solution for efficient model deployment.
Currently undergoing beta testing for image generation model deployment, the serverless version aims to expand its capabilities in March 2024. It will support the deployment of open-source LLMs and models from Alibaba’s AI model community, ModelScope, catering to tasks such as image segmentation, summary generation, and voice recognition.
Large Language Models
Alibaba Cloud’s technology suite, including LLMs, training services, and vector engine technology, facilitates a Retrieval-Augmented Generation (RAG) process. This process empowers enterprises to enhance LLMs with their knowledge bases, leading to improved accuracy, faster retrieval of relevant information, and nuanced insights across various applications.
Alibaba Cloud’s MaxCompute MaxFrame, an upgraded big data service, addresses the growing demand for data preprocessing and offline/online analysis in AI-related computing tasks. It enhances data processing efficiency, particularly for tasks like LLM training.
To enhance creativity, Alibaba Cloud introduced PAI-Artlab, a platform for model training and image generation. This solution caters to designers, enabling them to swiftly produce professional-grade designs for various applications without coding. The platform is currently operational in mainland China and will soon launch in Singapore.
Optimizing AI performance
Alibaba Cloud integrated its vector engine technology into its entire range of database solutions, boosting performance and capabilities. These vector engines optimize AI performance by embedding large volumes of context, benefiting LLMs and advancing various AI functionalities.
Alibaba Cloud elevated its entire range of database solutions, including the cloud-native database PolarDB, cloud-native data warehouse AnalyticDB, and cloud-native multi-model database Lindorm, integrating its proprietary vector engine technology to significantly enhance performance and capabilities.
“By open-sourcing our proprietary language models, we are well-equipped to offer powerful computing solutions and cutting-edge AI innovations to support clients in developing customized generative AI applications,” said Selina Yuan, president of International Business at Alibaba Cloud.