Windows 10’s next major update will include Windows ML, a new AI platform

Windows 10
Photo by Amelia Holowaty Krales / The Verge

Microsoft is planning to include more artificial intelligence capabilities inside Windows 10 soon. The software giant is unveiling a new AI platform, Windows ML, for developers today, that will be available in the next major Windows 10 update available this spring. Microsoft’s new platform will enable all developers that create apps on Windows 10 to leverage existing pre-trained machine learning models in apps.

Windows ML will enable developers to create more powerful apps for consumers running Windows 10. Developers will be able to import existing learning models from different AI platforms and run them locally on PCs and devices running Windows 10, speeding up real-time analysis of local data like images or video, or even improving background tasks like indexing files for quick search inside apps. Microsoft has already been using AI throughout Office 365, inside the Windows 10 Photos app, and even with its Windows Hello facial recognition to allow Windows 10 users to sign into PCs and laptops with their faces.

Microsoft’s Windows machine learning model is designed to run across a number of different devices, including laptops, PCs, Internet of Things devices, servers, datacenters, and the HoloLens headset. AI processors, like Intel’s Movidius VPU, will also be supported, and Microsoft’s platform will optimize tasks for the hardware available. Essentially, Microsoft is promising that Windows will take care of the hassle of worrying about older hardware processing machine learning models, instead of developers having to consider performance impacts in their apps.

Developers will be able to get an early look at the AI platform on Windows with Visual Studio Preview 15.7, and they’ll be able to use the Windows ML API in standard desktops apps and Universal Windows Apps across all editions of Windows 10 this year. Microsoft is planning to reveal more about the platform at its Windows Developer day today and at the company’s Build conference in May. We’d expect to see Microsoft demonstrate some of the direct AI improvements for future versions of Windows at Build, to show consumers exactly how this new platform will improve apps.


So give us an example of how this could be used by a developer.

If you have to ask, it’s not for you.

It’s not going to be for hardly anybody.

So, it’s going to be for everybody?

OK, I’ll admit that was funny.

Not really sure what to say here. I’m not a developer, but surely the products developers create will be for me. I’m just trying to understand if this is a big deal or not. We’ve seen the word "AI" plastered all over the place over the last year or so, but we’re not really told what it’s doing or how it’s doing it. I was hoping to get a real answer.

If you can’t answer, I guess it probably not for you either.

Microsoft’s "AI" might not be much when compared to Google’s, but in my opinion, MSR does a much better job at developing real-world applications that rely on ML. For example, the Seeing AI app for iOS is pretty cool. And they’ve also developed a near-real time translation feature that provides subtitles in the language of your choice within a 2-3 second delay, and is something that’s already being used in some universities that invite guest lecturers from different parts of the world.

Coming to Windows ML in particular, according to Microsoft’s blog, it already powers Windows Hello (which if you haven’t used, is an amazing feature to be missing out on), and the new Photos app is said to use that as well. No 3rd Party Apps use it as of yet (since it was just released), but it will be pretty exciting to see how developers incorporate it.

Does anyone know definitively who’s AI is ‘better’ and across what dimensions? I bet AI is a catch-all for a lot of things. Maybe Google is better in some aspects, MS in others. But that won’t make for a sexy headline.

>Surely products developers create will be for me.

Wrong. Most software is written for other developers. This is an example that was poorly explained by the author, probably because he or she doesn’t know what the Microsoft ML is for either.

it’s a platform, meaning it’s not there to be one specific product. It enables development. It’s probably just a set of APIs that make certain ML tasks easier.

E.g. right now as an ML developer I have to manually check how many GPUs the computer has and their specs. That’s idiotic, and it sounds like this will abstract that away.

I understand that it’s a platform for use by developers. However, eventually those developers will create a product for end users?

They certainly will, but this is as much a product for end users as a video game engine like Unreal is for gamers— It’s not, it’s a product for game developers.

You know I totally misunderstood your initial comment. Sorry for the rant. I think the only applicable thing I said was this:

>E.g. right now as an ML developer I have to manually check how many GPUs the computer has and their specs. That’s idiotic, and it sounds like this will abstract that away.

Hopefully making it easier for developers to deploy ML applications on Windows allows them to find bugs more easily and worry about arbitrary non-ML shit less. That’s all I got. Have a nice day!

Hey, thanks for apologizing. It’s a rarity on the interwebz and should be commended. And thanks for your input.

I was mentioning AI to my grandfather the other day, who was a farmer for most of his life. He thought I was talking about artificial insemination (you know, for cattle) and was quite confused why I was so excited about it.

A trivial example is image recognition.

Isn’t there talk of moving Cortana into the OS, as opposed to the search box (where it currently is, in a discoverable sense).

Well, Cortana is already part of the OS but yes, there have been several reports that she’s going to move to the Action Center.


All will be revealed!

Anyone with the capabilities of utilising it would already know how it can be utilised; ML is not a standard developer tool. If you needed it, you will have sought out frameworks that supported it.

Adding it natively simply means that its a lot easier to deploy ML applications intended for regular consumers into Windows. Right now, deploying ML applications is a bit of an ass because its rather picky about the environment. Specific GPU drivers, specific CuDNN versions, and so on. I had to deal with my Linux machine crashing the desktop environment every 20 minutes because that particular GPU driver just performed a lot better with the version of Tensorflow we were using.

Object recognition is the typical entry-level example.
Advanced examples include things like a DOTA 2 AI, or the Go playing Deepmind.

currently, Azure ML studio, which I’m assuming works similarly, as is is meh.

you essentially drag and drop data into a learning path and get an outcome. it spits out an arbitrary P value and poof! you have predictive results. the issue is that most people lie with stats so I can only assume their ML structures will be ill informed at best. additionally, most normal humans wouldn’t have an in depth awareness of predictive modeling/statistics. so all of these models could be garbage from the start. seems like Microsoft is just continuing their habbit of both making and chasing vapor trails…

… after that January update I’m so used to Microsoft putting out thoughtless, lazy, poorly tested product that my expectations for them lie far far far below what I expect a functional consumer tech company should be.

You’ve stated nothing exclusive to Microsoft. Google and Amazon’s systems do the same. If you want smarter ML…wait 5 more years.

Now that only thing missing is their solution to Google’s AR Core and Apple’s AR Kit, seriously they already have Mixed Reality on their HoloLens and VR devices, but they need to bring it down even to the mobile PCs with cameras (considering they don’t have smartphone foothold today) or they’ll miss another train again

Windows Mixed Reality came out last fall, how are they ‘missing a solution’? The problem for Microsoft is lack of presence on smartphones which has nothing to do with Windows MR/ML, but Ballmer’s failures.

I mean mobile AR solution through the cameras of mobile PC (e.g. 2-in-1s, tablets). Not needing a greater GPU hardware or some AR/VR device

View All Comments
Back to top ↑