Connectionism and the multiplicity of mind

摘要

Some connectionists (e.g. Smolensky, 1988) argue that classical AI models offer at best a good approximation to the fine-grained pychological truths revealed by connectionism. Likewise, some conventional AI theorists suggest that connectionism at best displays a new way of implementing the insight embodied in classical models. But the terms of the debate, I suggest, are by no means as cut and dried as such polarizations suggest. Instead, the mind may require explanation in terms of a multiplicity of virtual architectures. Both different tasks, and different aspects of that same tasks, may call for computational explanations quantifying over the operations and data-structures of different virtual machines.