In this talk I will describe some of our efforts in identifying and modeling the future of information technology. I will concentrate on my work in the area of AI on modeling and simulation of complex systems with advanced adaptive algorithms like genetic algorithms, neural networks and simulated annealing. The research I will describe is part of a project dedicated to exploring, studying and creating alternative computational metaphors. In particular we have been investigating how biological processes, ranging from natural selection to self organization, can guide us in the search for alternative computational paradigms and their internal structure. As a first step in this process, in this talk I will analyze the interaction of the parameters that control the dynamics of a simple but general evolutionary system modeled with an adaptive algorithm.
We look at evolution as a process of adaptation rather than an optimization. We study the effect of the problem size, complexity of the fitness function, strength of selection and different selection regimes, on the performance of an evolutionary algorithm in terms of speed of fixation and fixation to suboptimal values. The simulation results from our experiments are justified with a complete theoretical probabilistic analysis of the dynamics of the system. The talk introduces several new insights into the interaction and roles of the different parameters of an evolutionary adaptive system. Some areas of application of these results include modeling of populations of adaptive agents for human-computer interaction (HCI), and advanced algorithms for signal processing.