The limits of learning to learn
- Steven Phillips, Human Informatics Research Institute, National Institute of Advanced Industrial Science and Technology, Tsukuba, Ibaraki, Japan
AbstractLearning to learn is a reduction in the amount of training needed for task attainment across a series of similar tasks. Transfer differentiates (adult) humans from other species, portending a window into unique aspects of human learning. However, its unclear whether such differences are quantitative, or qualitative and what it means for the evolution/development of cognition. In this paper, learning is regarded as a (categorical) limit. A limit is a universal construction, and so transfer follows from a (generalized) optimization process. This result provides a formal basis for comparison/contrast of learning transfer in humans and other species---another step to bringing the empirical question into sharper relief.