Microsoft Courts Developers with AI-Enhanced Tools at Build 2018

Days after Facebook F8 wrapped up in San Jose, the Microsoft Developer Conference Build 2018 opened in Seattle. The conference runs May 7-9 at the Washington Convention Center, and is an opportunity for Microsoft to court developers.

In 2016 Microsoft ranked number one among open source project contributors on GitHub with 16,419 contributors. Monday’s keynote by Microsoft VP of Cloud and Enterprise Group Scott Guthrie announced a partnership with GitHub on two key products: Visual Studio App Center and Outlook.

To encourage developers to build Windows apps, Microsoft also announced a generous revenue sharing model for developers that is considerably above the industry standard of 70 percent. Developers will get 95 percent of revenue earned through app sales, in-store purchases, and new subscriptions; or 85 percent if the buyer is directed to the Microsoft Store app via Windows or Surface platforms.

AI-Enhanced Dev Tools Make Programmers’ Lives Easier

With seven million monthly-active developers, Microsoft Visual Studio is one of the world’s most popular developer tools.

A key new addition to Visual Studio is IntelliCode, a set of AI-assisted capabilities that improve productivity when coding. As you type, a built-in “IntelliSense” function will recommend the most likely API. The model also learns to provide more accurate suggestions as you write more code. The machine-learning model that makes the suggestions is trained with over 2000 GitHub repos. The extension is currently only available for C#.

Another function coming to Visual Studio is Live Share, which provides coders with a simple and streamlined real-time sharing environment, with changes updated on both parties’ screens as they are written. The function also smoothens the collaborative debugging process, as both parties can set breakpoints and advance the debug cursor.

Enlarging the .NET Ecosystem

In an attempt to pool AI engineers into their .NET ecosystem, Microsoft Executive Vice President of Artificial Intelligence & Research Harry Shum announced the opensourcing of ML.NET. Originally developed by Microsoft Research, ML.NET is a cross-platform, open source machine learning framework that has evolved into a powerful tool over the past decade and is used in multiple Microsoft product groups such as Windows, Bing, PowerPoint, Excel, etc.

Shum said the new open source ML framework will enable .NET developers to build their own machine learning models and applications without prior machine learning or parameter tuning experience.

The preview version of ML.NET can perform machine learning tasks such as classification (supporting text categorization, sentiment analysis) and regression (such as price forecasting). As a framework, ML.NET can accommodate popular ML Libraries such as TensorFlow, Accord.NET, and CNTK.

Microsoft’s Dev Platform Embraces Opensourcing

Whether you are a fan of Visual Studio or .NET development or not, it’s what Microsoft offers developers. Corporate Vice President Julia Liuson, who leads a Microsoft team of more than 1,000 for the Visual Studio and .NET Framework, tells Synced that it was back in 2011 that Microsoft made a top-level decision to open up the developer system.

She tells a story of when former Microsoft VP Robert Wahbe — who had earlier left the company to form his own startup Highspot.inc — visited Microsoft offices to do demos at the request of CEO Satya Nadella. To everyone’s surprise, Wahbe arrived with a Mac.

“He told us that Microsoft didn’t support open source, so although he wanted to use Microsoft products, he had to run codes on Linux using a Mac. His said his company only wrote 10 percent of their codes, the rest were opensource,” Liuson recalls. “That was when we knew opensource and the cloud was the way to go. We want you to code locally and have cloud functions.”

Liuson tells us machine learning has always empowered Microsoft’s development platforms, but in the background. One example is the “report a bug” function in Visual Studio that receives up to 70,000 responses per year. A machine learning system translates and filters the feedback messages, dispatching them to appropriate teams. Now, all that tech is getting its time in the spotlight. Says Liuson, “To make ML-based functions means putting AI up in the front.”