Robotics startup patents artificial neural networks that run on GPUs

The U.S. Patent Office today issued to Neurala a patent that covers brain-based computational models, often called Artificial Neural Networks running on graphic processing units.  The invention is seen as an important foundation for real-time artificial intelligence and robotics applications. Humans outperform computers in many natural tasks, including vision and language processing, because the brain efficiently processes many inputs, learns, and recognizes patterns.  

A robotics startup called Neurala has received a patent (No. 8,648,867) for a GPU-based system designed to run artificial neural network models. The patent covers the physical architecture of the system, which Neurala calls an “accelerator,” as well as aspects of data processing and user experience. It’s not clear whether the patent, which dates back to 2006, will affect others artificial intelligence efforts currently underway. Neurala’s business revolves around providing computer vision and navigation intelligence for robots, but GPUs are the computers of choice for many deep learning projects.