Hot on the heels of a new artificial intelligence-capable thermal imaging camera, chip maker Movidius has yet another AI implementation up its sleeve: a USB stick that can allow pretty much any Linux computer to handle advanced neural networks, one of the building blocks of AI.
The San Mateo, Calif.-based company announced the device, called the Fathom Neural Compute Stick, today along with its Fathom deep-learning software framework. Together, the two products will enable device manufacturers to move AI processing from the cloud to native deployment in end-user devices.
Such a step is critical if AI is to become as ubiquitous as its advocates claim it can be. While massive data centers with powerful GPUs can quickly churn through complex machine-learning problems, it isn't always feasible or practical to find enough bandwidth to transfer the results to user devices.
The Fathom stick isn't a standalone device. Instead, it works much the same way you'd boost your PC's memory using Windows' ReadyBoost feature. Once it's plugged into a Linux-powered device, it will enable that device to perform neural network functions like language comprehension, image recognition, and pattern detection.
Since those functions require extensive programming, the Fathom stick can save companies and researchers time as they develop a new generation of AI-capable devices by not having to build the neural networks themselves.
"As a tinkerer and builder of various robots and flying contraptions, I've been dreaming of getting my hands on something like the Fathom Neural Compute Stick for a long time," NYU computer science professor Yann LeCun said in a statement. "With Fathom, every robot, big and small, can now have state-of-the-art vision capabilities."
The stick runs on an ultra-low power Myriad 2 VPU. Myriad chips have also been used in Google's Project Tango, in addition to the Flir thermal imaging camera.