This will be a short one from me, but it’s big news for Windows devs. With the next iteration of Windows 10 (which is being rumoured to be called the Spring Creators Update), Microsoft will begin to make an AI platform available, enabling developers to utilise out-of-the-box machine learning models Microsoft has been training for a while natively in their apps.
We’re entering an era where it’s not enough for your app to serve static information that isn’t specifically targeted to the user, or doesn’t surface the information they need at a specific time or context. This will be made a lot easier for us by pre-packaged models like the ones that Windows will offer natively, and even better, the platform will be offer the ability to run the models locally on users’ devices, meaning internet connectivity won’t be a pre-requisite and processing can happen a lot more quickly and efficiently. This will make full use of hardware advanced made by Microsoft’s partners, including Intel and NVIDIA’s new chipsets.
Windows will support the industry-standard ONNX for their models, and this will mean greater flexibility to develop these across different app platforms – not having to create new models for your Android apps for example.
You also don’t have to rely on the pre-packaged models if you’re an AI developer. There will be a development experience available with Visual Studio (coming with Preview 15.7) where adding an ONNX file to a Universal Windows Platform project will generate the model interfaces for you automatically, and this will be coming to Visual Studio Tools for AI soon as well. Pretty nifty.
I’m personally quite excited about what the future holds for app developers with AI becoming omnipresent across applications and devices. It’ll be a lot easier to create the user experience we’ve always wanted to without having to rely on teams of data scientists to come up with models first. More announcements at the annual //Build conference in May. Stay tuned.