Microsoft is designing its own artificial intelligence chip.
The company’s new chip will be used so that HoloLens, its mixed reality headset, can process AI tasks like speech and image recognition within the device instead of processing that data through the cloud.
Currently, the HoloLens contains a custom multiprocessor called the Holographic Processing Unit, which is responsible for processing the information coming from on-board sensors, including head-tracking cameras and the infrared camera. The new chip is a majorly upgraded version of this.
“The second version of the HPU, currently under development, will incorporate an AI coprocessor to natively and flexibly implement deep neural networks,” says Marc Pollefeys, Director of Science, HoloLens. “The chip supports a wide variety of layer types, fully programmable by us.”
The chip concept was unveiled by Harry Shum, executive vice president of Microsoft’s Artificial Intelligence and Research Group, at CVPR 2017.
“Mixed reality and artificial intelligence represent the future of computing, and we’re excited to be advancing this frontier,” added Pollefeys.
Microsoft is not alone in custom-building AI silicon; Apple is planning to do the same for its iOS devices to bolster Siri’s capabilities.