Assembly And Differentiation

There are two key challenges to creating the next generation applications: assembly and differentiation. Robust, inexpensive assembly is required at the very lowest levels (assembling nanoscale low-power economical computing devices) and the highest levels (assembling distributed applications from reusable low-cost software components). Differentiation is less obvious, but in some sense a more basic primitive that will enable low-cost assembly. By differentiation I mean the ability to create an aperiodic complex system from relatively homogeneous and possibly defective parts. Robust inexpensive assembly is the key to accomplishing next generation applications. Assembling nanoscale wires and molecules into computers is the key to low-power low-cost high-density computers. Assembly of sensing and actuating components with computing elements is essential for the low-cost computation and communication needed to realize ubiquitous computing. Moving up the hierarchy is the challenge of assembling many processing elements into a unified distributed computing system. The means of organizing billion-processor networks into a coherent whole is a challenge containing many challenges. Reducing the cost and increasing the reliability of software is essentially a challenge of assembling components. Producing the high-fidelity sensor networks required for many ubiquitous computing tasks is a challenge in assembling information from many low-resolution sensors into a high-resolution picture. Computer science has produced, or at least understands, a powerful set of tools and stand-alone components. It now faces the task of assembling them together into coherent ensembles. Here, I briefly outline some of the assembly challenges facing computer science, from basic computing technology to a grand challenge application: smart matter. At the base of any computer science application is the computing element itself. For the last twenty years the computing substrate of choice has been CMOS. The cost per compute element has dropped as manufacturers have learned to assemble ever more components on a single chip. With the advent of MEMS, this assembly process has begun to include sensor and actuator technology as well. The beauty of this process has been that components (e.g., wires and transistors) have been manufactured and assembled into a device in a “single” step. However, that “single” step is now more than 200 processing steps—each of which requires a fine degree of precision. The challenge is to find a new method of assembling components into devices; a method that requires less precision, a method that works with elements at the nanoscale, a method that is inexpensive, and yet robust. Researchers have already demonstrated computing components that are smaller than ten nanometers. The goal is to assemble these components into a useful computing device. Many believe that we should be able to create inexpensive, low-power computing devices with densities of