There is still much to learn about that variety types of learning and development in nature and the genetic and epigenetic mechanisms responsible for that variety. This paper is one of a series exploring some ideas about how to characterise that variety and what AI researchers, especially robot designers can learn from it. It is proposed that whereas some robots, like most animals, will need to be pre-programmed with all the competences they will ever need, others, like humans, will need powerful learning mechanisms. However, contrary to popular assumptions, instead of those mechanisms being totally general it is likely that, as in humans, they will start with some deep, but widely applicable, assumptions about the nature of the 3-D environment, about the nature of other information users in the environment and about good ways to explore that environment, e.g. using creative play and exploration. One feature of such learning could be learning how to learn. Besides learning things that are expressible in terms of an innate ontology, using innately determined forms of representation, the process of extending learning capabilities will include extending the learner's ontology and the forms of representation available. Further progress may require close collaboration between AI researchers, biologists studying animal cognition and biologists studying genetics and epigenetic mechanisms.
Subjects: 11. Knowledge Representation; 9.4 Philosophical Foundations
Submitted: Sep 14, 2007