AI at the edge: Intel’s Irish acquisition Movidius reveals $79 AI stick

21 Jul 2017135 Shares

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

The new Movidius Neural Compute Stick. Image: Intel

Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on PinterestShare on RedditEmail this to someone

AI stick delivers deep neural networking capabilities to devices at the edge.

Almost a year after being bought by Intel, Movidius has revealed a tiny $79 USB tool that brings advanced AI to ordinary hardware devices.

The Movidius Neural Compute Stick is described by Intel as the world’s first USB-based deep learning inference kit and self-contained AI accelerator.

‘This enables a wide range of AI applications to be deployed offline’
– REMI EL-OUAZZANE

It delivers dedicated deep neural network processing capabilities to a wide range of host devices at the edge – think of it as supercomputing in your pocket.

AI at the edge: Intel’s Irish acquisition Movidius reveals $79 AI stick

The new Movidius Neural Compute Stick. Image: Intel

Last year, prior to Intel’s acquisition, Siliconrepublic.com reported how Movidius made a major breakthrough with the Fathom Neural Compute Stick, which would allow powerful neural networks to be moved out of the cloud and deployed on new products such as robots and drones.

When connected to a PC, the Fathom Neural Compute Stick behaved as a neural network profiling and evaluation tool.

AI on a stick

With the new Movidius Neural Compute Stick, developers can perform such actions as train artificial neural networks on the Intel Nervana cloud and optimise existing workloads for AI as well as virtual and augmented reality, and even automated driving.

The stick also makes it possible to bring the same Movidius vision processing – seen on devices such as the DJI Spark drone – to devices on the edge.

“The Myriad 2 VPU housed inside the Movidius Neural Compute Stick provides powerful, yet efficient performance – more than 100 gigaflops of performance within a 1W power envelope – to run real-time deep neural networks directly from the device,” said Remi El-Ouazzane, vice-president and general manager of Movidius.

“This enables a wide range of AI applications to be deployed offline.”

Editor John Kennedy is an award-winning technology journalist.

editorial@siliconrepublic.com