Project Vision

A new computing paradigm that exploits uncertainty to design systems that are energy-efficient and scale gracefully under hardware errors by operating below the nominal operating point, in a controlled way, without inducing massive or fatal errors.

This vision is realistic because not all computations and not all data of an application are equally critical, requiring to be performed or maintained at 100% accuracy or correctness. Several application domains offer the opportunity to trade off quality of service (QoS) for significant improvements in energy consumption. For such applications, it may be possible to only approximate the final output (or part of it), rather than computing the exact output result.