This article investigates the concept of ‘complex systems’. While not searching for some necessary and sufficient conditions that are valid for all of them, it acknowledges that complex systems can take different shapes, mainly depending on the features of their internal organization and how they interact with their environment. It then advocates a networked notion of complex systems that can accommodate their rich phenomenology and the various circumstances making them, focusing on two types of these systems: (i) one that is (...) mainly characterized by the generation of stable patterns through self-reinforcing dynamics at the lower levels (Bénard convection) and (ii) a distinct one characterized by a more complex organization that makes them ‘minimally decomposable’ and showing autonomy (living systems). The article also assumes that the complexity of a system is analyzable by focusing on two distinct yet interrelated aspects: (i) the features of the system itself and (ii) the relationship between the system and an observer. Its final part discusses how complex systems cannot be adequately represented by a single model or description and how this is another distinctive aspect of their complexity. (shrink)