HomeWinBuzzer NewsMicrosoft Simplifies AI Deployment with New Azure Models-as-a-Service

Microsoft Simplifies AI Deployment with New Azure Models-as-a-Service

The Models-as-a-Service in Azure AI Studio is debuting with 160 AI models, including Stability AI and TimeGen-1.

-

has announced a new service within its Azure AI Studio called Models-as-a-Service (MaaS), aimed at simplifying the deployment of artificial intelligence (AI) models for developers. This service offers a streamlined approach, allowing developers to bypass the usual complexities associated with AI model deployment. By providing access to a curated catalog of AI models, MaaS enables developers to activate and utilize these models with ease, significantly reducing the technical barriers.

Expanding the AI Model Library

Central to the MaaS offering is an extensive library of over 1,600 AI models, covering a wide range of functionalities. Recent additions to this library include TimeGen-1 from Nixtila and Core42 JAIS, with further expansions expected from AI21, Bria AI, Gretel Labs, NTT Data, Stability AI, and Cohere. This diverse array of models underscores Microsoft's commitment to providing developers with a broad selection of AI tools to meet various needs.

Pay-As-You-Go Model for Flexibility

The MaaS framework is designed to be highly inclusive, allowing developers to use AI models for inference and fine-tuning on a pay-as-you-go basis. This eliminates the need for direct interaction with underlying hardware or extensive configuration, making the AI implementation process more accessible. Seth Juarez, Microsoft's principal program manager for the AI platform, emphasizes that this service abstracts away the intricate details of deployment, enabling developers to focus on the creative aspects of their projects.

Microsoft envisions a future where developers can choose between owning their AI models and infrastructure or opting for the MaaS model, similar to the choice between renting and owning a home. Each option offers distinct advantages, catering to different requirements and preferences. For those who choose MaaS, Microsoft promises continuous maintenance and support, alleviating the burden of managing the infrastructure.

While the MaaS model is designed to be highly flexible, Microsoft acknowledges that certain specialized or unique models may not fit this framework due to their specific requirements. These models may need to be deployed through more traditional means, highlighting the company's commitment to providing solutions that cater to the diverse needs of developers.

Enhancements in Azure AI Services

In addition to the MaaS offering, Microsoft has unveiled several new capabilities within its Azure at its annual Build developer conference. These include enabling greater database access, automatically dubbing videos into multiple languages, and rapid training of large language models to understand complex document structures. The company has also enhanced its integrated development environment for AI, Azure AI Studio, to include Azure Developer CLI, a set of templated commands used to deploy applications to the cloud.

Microsoft is introducing a new type of AI model called “custom generative,” which allows rapid development of language models to process complex documents using templates to define the document structure. This model reduces the number of labels a developer needs to craft, using large language models to extract fields, with users only needing to correct the output when necessary.

Updates to Azure AI Search and Database Offerings

Azure has been updated to enhance the way it scores results stored as vectors and to include the ability to turn images into vectors. The service now includes a connector that routes data contained in the OneLake data lake, enhancing the ability to connect to corporate data. Additionally, Microsoft has added features to its database offerings, including vector search and embeddings, to support large language model deployment. Azure Cosmos DB for NoSQL now performs vector search, making it the first cloud database with lower latency vector search at cloud scale without the need to manage servers. Azure Database for PostgreSQL now includes in-database embedding updates to automatically compress input data into representations the LLM understands.

SourceMicrosoft
Luke Jones
Luke Jones
Luke has been writing about all things tech for more than five years. He is following Microsoft closely to bring you the latest news about Windows, Office, Azure, Skype, HoloLens and all the rest of their products.