Present and the Future of Chaos Computing Behnam Kia, Vivek Kohar, William Ditto Department of Physics, North Carolina State University, Raleigh, North Carolina 27695-8202 USA Abstract We study chaos computing as a new approach for reconfigurable com- puting and present some of our latest results and discuss what this new direction to computing means and implies. We discuss the advantages and challenges that come with this new paradigm of computing and envision its future. 1 Introduction Nonlinear dynamics can be considered as a rich library of different behaviors. More specifically, by changing a bifurcation parameter of a nonlinear system, the dynamics of the system qualitatively changes. For example, one bifurcation pa- rameter value puts the nonlinear system in a periodic regime, whereas another pa- rameter value can move the system to a chaotic regime [1]. Also, when a nonlinear system is in a chaotic regime, its chaotic attractor is composed of infinite number of different unstable periodic orbits, where each can be dynamically selected and stabilized [1]. This rich dynamics of a nonlinear system, combined with our ability to control and program it, enables us to introduce fascinating applications for it that otherwise are not possible through conventional engineering or by linear sys- tems. Chaos computing is a term that refers to the methods and approaches to utilize nonlinear dynamics and chaos to perform computing [2,3]. The main advantage of chaos computing is that the same circuit can be programmed to implement differ- ent types of computation [4,5]. In this approach, since the nonlinear dynamics, which contains many different behaviors, is performing the computation, different types of computation coexist within the system and we can dynamically select and pick up different types of computation from the system [4,5]. This capability of a circuit to implement different types of functions is of cru- cial importance in today’s troubled semiconductor industry, where we cannot shrink the size of the transistors anymore to integrate more transistors into a mi-