Topic 6: Representation

[Back to the course overview]

Lecture 12: We introduce a suite of topics under the broad heading of representation. We first hone our intuitions using the familiar example of maps, and the various means by which they represent things. We contrast this with the case of an imprint, which also stands in a relation to something not present, but where the relation is causal. We then consider cortical maps and find that they have a causal, topographical relation to the associated primary sensory surface. We turn then to theories of more abstract knowledge, contrasting propositional and imagistic representations. Finally, we show how the way in which knowledge is stored greatly influences what we can ask of it.

Lecture 13: Continuing the theme of knowledge representation, we look at the use of propositional representations in artificial intelligence (of the 1970s and 1980s – the meaning of that term has changed greatly!). We explore the way categories work, noting their grounding in a particular socio-cultural setting, and we contrast basic level categories with specialized and generalized categories. Turning to robotics, we find that getting by in the world may not require representing it, which leads us to Moravec’s paradox: highly intellectual tasks such as playing chess turned out to be easy to solve (relatively speaking) whereas mimicking the behavioural skills of a child or an animal turns out to be much more difficult.

Required reading:

van Gelder, T. (1998). Into the deep blue yonderQuadrant42(1-2), 33.