This paper investigates the problem of completely decentralized adaptive control of large-scale systems. When the subsystems are stable, a new approach, which eliminates the assumptions about the interconnection terms, the time delay, the order, etc., is proposed. Each subsystem is modeled by an orthonormal Laguerre network put in state-space form. A simple decentralized predictive control law is proposed. The potential of such a scheme is illustrated by simulation results.