Embedded objects featuring artificial intelligence used to require a cloud connection. But Nvidia’s upgraded Jetson deep learning module featuring the new TX2 chip aims to change this. It provides built-in intelligence for robots, drones, cameras and other devices “at the edge,” or embedded. The chip enables the device to infer from its experience—to learn and modify its behavior independently, without a cloud connection.
A cloud connection can suffer from bandwidth limitations, latency and security concerns. Processing data at the source is one way out. The TX2 consumes just 7.5 watts, twice as efficient as its predecessor. Alternatively, it can double computing power at 15 watts. For example, a TX2-equipped vehicle camera designed to identify pedestrians and other vehicles on an urban street offers twice the performance of earlier models. Other potential applications include an indoor search-and-rescue drone, personal assistance robots and virtual reality helmets.