Alibaba Cloud Advances Its Gen AI Capabilities with Serverless Solution
January 30, 2024

Alibaba Cloud Advances Its Gen AI Capabilities with Serverless Solution.jpg


  • The cloud computing platform's serverless solution cuts inference costs by half
  • Alibaba Cloud hosts its first AI & Big Data Summit in Singapore


Alibaba Cloud unveiled a serverless solution on Tuesday that offers individuals and companies more cost-effective access to generative AI model services, including model deployment and inference.

 

The company is transforming its product lines for the generative AI era, widening access to large language models (LLMs) and custom applications.

 

Alibaba Cloud is also harnessing generative AI by integrating its vector engine technology into an expanded suite of products, including its data warehouse solution Hologres and search services Elasticsearch and Opensearch.

 

"Our technology updates underscore our commitment to empowering enterprises with the latest Intelligence-driven solutions for heightened efficiency and performance," said Zhou Jingren, Alibaba Cloud's Chief Technology Officer, at its inaugural AI & Big Data Summit in Singapore.

 

The one-day summit brought together Alibaba Cloud executives, thought leaders and industry experts with over 2,000 registered attendees. The crowds can browse the product showcases from more than a dozen of Alibaba Cloud's partner companies.

 

But Alibaba Cloud's serverless product, the serverless version of its PAI-Elastic Algorithm Service (EAS) platform, was the belle of the ball.

 

It allows users to tap computing resources as needed and only pay for what they use, cutting inference costs by up to half of traditional pricing models.

 

The serverless solution is in beta testing but is available for image generation model deployment. From March onwards, it will also support the deployment of open-source LLMs and models from Alibaba's own AI model community, ModelScope.


Vector Engine


Integrating Alibaba Cloud's vector engine into a wider variety of products is transformative for AI performance.

 

Vector engine technology converts text and data into a high-dimensional space to enable the embedding of massive volumes of structured and unstructured context to streamline tasks.

 

After integrating its proprietary vector engine into its full range of database solutions last year, Alibaba Cloud saw a substantial boost in the solutions' performance and capabilities.

 

With the vector engine technology, LLMs serving and training services, Alibaba Cloud is also enhancing its ability to support the Retrieval-Augmented Generation process, which will help enterprises enhance LLMs with their knowledge bases to improve the accuracy and efficiency of information retrieval and insight generation.

 

Enabling the retrieval of information that is contextually relevant helps streamline the search process and delivers more accurate and comprehensive answers to users' queries, enhancing the overall quality of information retrieval in a range of applications from business intelligence to customer service.


Supercharged Design


Serverless solutions and vector engines weren't the only novelties during Alibaba Cloud's summit in Singapore, as the company also unveiled a tool for designers to bring their creations to life.

 

PAI-Artlab simplifies the process of model training and image generation, from interior design and product promotional posters to creating characters and scene development for gaming.

 

Its seamless integration with PAI-EAS allows developers to concentrate on model training and refinement while removing the burden of deployment concerns.

 

By harnessing Alibaba Cloud's machine learning technology, the platform can expedite training and cut the time needed for image generation.