ABSTRACT
This chapter presents some difficulties with the characterisation of autonomous agents as information processing systems arising from the work on real autonomous mobile robots. It argues that the modern fashion for situatedness. The chapter also argues that the so-called symbol grounding problem is an inappropriate response to those who have pointed out that the world of an autonomous agent does not come ready categorized and neatly labeled. It presents a series of sketches based on the experience of trying to build real autonomous systems: simple mobile robots. In each sketch, it identifies some problems with characterising these robots as information processing systems. The chapter concludes with some speculations on what might be a more satisfactory characterisation, which has much in common with some ideas of Humberto Maturana and Francisco Varela.
