Catching up on Build 2021: AI and Arm


Microsoft’s annual developer event Build mixes education with new product announcements across its three main developer platforms: .NET and Windows, Azure, and the Power Platform. There’s always a lot of other news, too, with previews going into general availability and new features being added to existing platforms and services. Then there’s the bigger picture, with a snapshot of its developer strategy and its product road map.

What were the key announcements at Build this year? I’ve already looked at the Azure data platform updates, so it’s worth a quick dive into what I think are the most important pieces of news for anyone building applications on Microsoft platforms, whether they’re experienced coders or someone picking up a low-code development tool for the first time.

Azure got the biggest share of Build news this year, cementing its position as the centerpiece of Microsoft’s modern development strategy, with a focus on cloud-native development tools, machine learning (ML), and bringing Azure services to hybrid cloud deployments via Azure Arc and Azure Stack.

Machine learning with a supported PyTorch

One of the more interesting Azure ML announcements is its release of a supported build of PyTorch Enterprise. Originally developed by Facebook but now used by many different organizations (including ones helping develop Microsoft’s own ML products), PyTorch is an open source deep-learning framework that can create and share custom ML models, using either its own inferencing runtime or exporting models as Open Neural Network Exchange (ONNX) for use in Windows.

PyTorch Enterprise builds on the open source tooling and adds a long-term support model, as well as allowing Microsoft to integrate it more closely with Azure tools and services. It’s an approach that should help you move from using PyTorch as a research and development tool to one that’s suitable for production, with Microsoft offering troubleshooting and support for users with existing support contracts. It’ll also be built into the Azure Machine Learning platform, giving you a managed instance that you can use to develop and test models before moving them into production.

Using AI to help low-code developers

We’re finally seeing the first fruits of Microsoft’s partnership with OpenAI, which developed the GPT-3 natural language model. Microsoft is using it in conjunction with its new low-code Power Fx language, going from simple statements to code. Using these tools, you’ll be able to type a basic query in natural language: for example “show me all the customers in the eastern United States,” and the model will generate a set of possible Power Fx statements that can be used in your code. It’s an approach that should make query design available to a greater audience, with no need to learn SQL; all you need to do is ask a question.

Copyright © 2021 IDG Communications, Inc.



Source link