It looks like Microsoft is quite aware of the fact that artificial intelligence needs to work with info as it happens. So, it unveiled Project Brainwave, its own hardware acceleration platform, promising faster, real-time AI in the cloud. And the best part is the company has selected Intel® Stratix® 10 FPGAs in this deep learning platform. Its FPGA is capable to sustain a hefty 39.5 Teraflops on a large gated recurrent unit, with no need to batch tasks together.
Compared to many of its hard-coded rival, it’s considerably more flexible. It depends on a soft dynamic neural network processing unit put into the off-the-shelf FPGA chips where opponents require their technique locked in from the outset frequently. Besides handling Microsoft’s own AI framework, it’s able to work with Google’s TensorFlow, as well as other systems. Instead of allowing the hardware to dictate your technique, you can create a machine learning system the way you want and make it run in real-time.
Microsoft is also planning to make its Project Brainwave available through Azure, company’s own cloud services. It’s not surprising as the company wants all the companies using the Azure cloud service can make use of live AI. Microsoft hopes it will increase the use of Azure when people know they can control the way their AI runs.