Companies like Apple, Samsung, Qualcomm and Nvidia all make their own computer chips, but they sometimes get the schematics from another company: Arm.
That company, which might be the most important chip designer in the world because of its influence on the giant tech companies, is now betting big on artificial intelligence. Arm announced today (Feb. 13) a pair of chip designs for artificial intelligence, called Project Trillium.
A gadget like a smart light or an early iPhone may have a single Arm chip inside. It’s the brain of the device, running the operating system and responding to commands. These central chips weren’t designed to run AI software, like a smart security camera that can analyze video to see who’s knocking on the door. It would be like asking someone at Burger King to make you a Big Mac: They might have the parts to do it, it’s just not what they were meant to do.
So Arm is making two different chips that are specially built to run AI operations. One is aptly named the Arm Machine Learning processor, which is for any AI task from translation to facial recognition. The other is the Arm Object Detection processor, which finds interesting bits in images (like faces) and passes them onto the ML processor for processes like facial recognition.
Jem Davies, Arm VP of machine learning, tells Quartz that what’s special about these processor designs is that they should scale up—meaning they could be used in Internet of Things (IoT) devices or modified and enlarged for use in datacenters.
Amazon, Google, Apple, Intel, Qualcomm, and a host of startups have all started designing their own AI-specific chips, signaling that the best software and hardware engineers are looking towards specialty chips to enable AI functions. Arm’s entrance onto the scene can be taken as a sign that the hardware has “arrived.” It’s no longer niche or special to have an AI-specific chip for your little gadget, because now you can just buy this design from Arm.
Still, Arm doesn’t predict this technology to be mostly used in smartphones. Instead, the most growth will come from IoT devices and internet-connected cameras.