The concept of "task" is at the core of artificial intelligence (AI) --- the raison d'\^etre of AI systems. Tasks are used in system evaluation, pedagogy and decision making (through task decomposition and subtask selection). Other fields have strong theories of tasks in their domain that allows them to thoroughly evaluate their designs by methodical manipulation of well understood parameters of known importance; this allows an aeronautics engineer, for instance, to systematically assess the effects of wind speed on an airplane's stability. No framework exists that allows the kind of methodical manipulation found in other disciplines: Results on the few tasks in current use (c.f. board games, question-answering) cannot be easily compared, however similar or different. A task theory would address this by providing the appropriate formalization and classification of tasks, environments, and their parameters, enabling more rigorous ways of measuring, comparing, and evaluating intelligent behavior. Furthermore, analysis and (de)construction of tasks can help teachers devise appropriate challenges for learning systems, and help AI systems make more informed decisions about what (sub)tasks to pursue. Here we discuss the main elements of this argument and propose what such a theory might look like for physical tasks.
from cs.AI updates on arXiv.org http://ift.tt/1Vf09ht
via IFTTT
No comments:
Post a Comment