Julie C. Rutkowska
Significant new work in AI sees perceptual and motor skills as the 'hard' problems solved by intelligent systems, solutions to which impose important constraints on remaining components of natural intelligence. This paper outlines some ways in which an ontogenetic focus may prove relevant to attempts to understand adaptive behaviour in mobot/robot-environment systems. Synchronic accounts that seem to work for basic abilities may fare less well when required to accommodate diachronic processes of change; and understanding what kinds of change those might be can contribute to bridging the explanatory gap between sensory-motor and conceptual systems. Three main issues are discussed that focus on if and how scaling-up from basic to supposedly higher abilities is possible: (1) Whether early infant abilities are better explained in terms of conceptual representations or a computational account of action; (2) Reciprocal constraints between cognitive and physical-motor mechanisms, the role they play in paradigmatic cases of adaptive change, and the need to take seriously the internal structure of behaviour; (3) How far typical self-organizing connectionist networks take us towards understanding a system that is capable of mapping recurrent viable patterns of activity into more permanent adaptive changes.
This paper is not available online