Artificial Intelligence is leveraged to find the next communication algorithm
Communication algorithms are the workhorses of the modern digital era. Everything from cell phones, satellite communication, cable modems, DVDs, flash disks, and magnetic storage solutions routinely employ these codes to combat channel noise and errors. The problem is these codes have traditionally relied on human ingenuity for their design, sometimes requiring highly sophisticated mathematical techniques, so their development has been sporadic, sometimes with decades passing between conceptual leaps. Now, Sreeram Kannan and Himanshu Asnani, from the UW Department of Electrical Engineering, have developed DeepCode, a system that leverages artificial intelligence technology and, in particular, modern deep learning and reinforcement learning to automatically invent communication schemes.