From a purely technical point of view, Windows 10 is in a transitional phase. Behind the scenes, Microsoft is reorganizing its system so that it can run on all sorts of devices. Because of this, the ARM architecture is getting a lot of attention as the old Windows mainstay of AMD PCs and Intel laptops is slowly but surely dwindling. We talked about Windows 10 on mobile devices in detail in this article.
The share of mobile devices is growing rapidly, and ARM has all the trump cards in this category. The best proof is Microsoft’s HoloLens mixed reality glasses. The first version of HoloLens uses modified Windows 10 on Intel’s Atom processor. More recently, information was leaked about the second version, which is scheduled for release in 2019. According to this information, HoloLens 2 will be equipped with an ARM chip and special hardware for artificial intelligence.
In the near future, we can also expect the release of a new mobile device from Microsoft. Whether it will be a Surface Phone is still unknown. Patents filed by the corporation speak of a foldable tablet suitable for phone calls. An ARM processor is also hidden inside.
System core for all devices
The Surface Phone will indeed be the first device to run Windows Core OS, marking the next major milestone in Windows 10. While Microsoft hasn’t explicitly announced its master plan yet, if you summarize a few hints, you can paint the following picture: Windows 10 should turn into a universal operating system that works on absolutely all devices and hardware platforms. To do this, it will be redesigned.
The core of the system, as well as additional system functions in the future will represent Core OS. This root system can be implemented on all sorts of devices, from the screen of mixed reality glasses to an 80-inch TV.
To display various interfaces, Microsoft extends the Windows Shell. This tool defines what Windows will look like and, among other things, includes all interface elements – from the window structure to the Start menu. This extension is known as Composable Shell. To prepare it, several projects are being worked on, the purpose of which is to adapt Shell to various output devices – for example, the Andromeda project for Surface Phone.
Desktop programs in this concept play only a secondary role. They are closely engaged only in the framework of the Polaris project for PCs and laptops. According to Microsoft, the future belongs to UWP applications, since they are already designed for various hardware platforms by their code.
You can’t run away from AI
Support for new hardware is the buzzword of another big project: artificial intelligence (AI) integration. Until now, all relevant functions have worked through the cloud – both the voice assistant Cortana and the identification of faces in images in the OneDrive storage. In the spring update, Microsoft basically frees AI from being tied to the Internet.
The AI platform in Windows 10 will allow you to embed neural networks directly into applications
Cam VedBrother, Team Leader for Windows AI Team
How computers learn to see
Cam VedBrat leads the Windows ML (machine learning) support team that is part of the Windows AI team. At the Developers Day event in May, he explained that the “ten” with the Redstone 4 update will also support special equipment such as the Intel VPU Movidius (VPU – machine vision processor).
Movidius’ Myriad X processor runs a neural network that analyzes visual data for content. Face recognition, identification of objects and people – such tasks are performed by the AI processor in real time. The neural network, coupled with a webcam, is theoretically capable of reading even emotions.
The advantage of AI equipment lies in the combination of high processing power and low power consumption. For example, a neural network on the Myriad X processor no larger than a fingernail performs a billion operations per second.
Windows ML acts as a link for the direct use of neural networks in Windows 10. Both the system itself and the application can initiate a task. We are not talking about some distant plans: Adobe and Corel have long integrated neural networks into their products. The brilliant idea behind Windows ML is that it’s not only big companies with the right resources that can manage it, but virtually any software developer. The neural network is integrated into the code with just three commands.
Neural networks as a component of Windows
According to Vedbrat, within a year, the ML interface for desktop programs and UWP applications will become available in all versions of Windows 10. Thanks to this, it will be possible to import already trained neural networks. Windows ML uses the Open Neural Network Exchange (ONNX) format supported by Microsoft, Facebook, and Amazon to transfer data. With the exception of Google, all the giants of the industry are in favor of this format. However, there is a corresponding plugin for Google TensorFlow. Of the hardware developers, this standard is supported by AMD, ARM, Huawei, IBM, Intel and Qualcomm.
Almost everything that Microsoft wants to be implemented in Windows 10 leaves an ambivalent impression – from the feeling that the “big brother” is watching you, to the excitement of discovering new opportunities. Longtime users, however, don’t want to mess with the new features and would rather go back to XP. At the same time, Microsoft’s plans show that a decision has been made to go even into the distant future with Windows 10.
Interesting materials about Windows:
- What awaits us in the new Windows Redstone 5
- How to test a new version of Windows
Photo: manufacturing companies, pixabay.com