The Computing Earthquake: Neural Networks, Cognitive Layering and Energy-Driven Compute in Electronic Systems

A set of profound changes are clearly underway in computing. New computational models like convolutional neural networks for pattern recognition are replacing programming with training. Responsive systems are evolving new hardware and software layers that shift the most commonly executed functions to low-power subsystems – even while driving richer functionality into less-frequently functions into the cloud.

Distributed applications are not just spread across multiple processors on a chip, or multiple boards in a rack, but distributed across mobile and IoT devices, network gateways and servers sometimes on the opposite side of the globe from the user. These radically distributed systems drive a dramatic readjustment of trade-offs between computing and communication. Taken together, these changes represent a seismic shift in the computing landscape, with a new imperative to closely understand the flow of data, the energy cost of computing and alternatives to conventional software development.

Speaker: Chris Rowen, Cadence