Questions for NSF XPS Workshop 2015

Pervasive machine intelligence is bound to be a major theme behind the research and development of future computer systems, from computing artifacts embedded in millimeter-scale IoT devices to blade servers stacked in large-scale data centers. These intelligent systems will be capable of gathering, processing, exchanging, and extracting information from the environment in an increasingly autonomous manner. In this light, I believe that applications from the realm of machine perception and learning will be of high importance in defining the new characteristics of next-generation computer architectures and programming languages. Notably, software developers have increasingly utilized domain-specific languages (DSLs), which leverage novel language features and smart compilers to express and carry out optimizations that would otherwise require significantly more effort to realize with traditional languages. I imagine new DSLs (or embedded DSLs) that explicitly model uncertainty will be needed for capturing the stochastic and/or approximate nature of many important artificial intelligence algorithms.