Limits of Artificial Intelligence

Whilst the greatest effort has been made to ensure the quality of this text, due to the historical nature of this content, in some rare cases there may be minor issues with legibility. This philosophical debate concerns the existence of fundamental limits to the artificial intelligence enterprise, which however is only one of several kinds of potentially significant limit that need to be considered. Even if no such fundamental limits existed, i.e. Even if a hypothetical infinitely fast computing engine possessed of infinite amounts of memory could in principle duplicate all aspects of human mental capability, it would still remain necessary to ask just how much computation and data storage such duplication would require. Suppose, for example, that it could be shown that the minimum computational resource required to duplicate some human mental function is implausibly large, relative either to the extreme limits of physically realizable computation, or to the largest computers likely to be constructed over the next decades or centuries. In this case, construction of significant artificial intelligences would be blocked by inescapable practicallimits, even if fundamental limits did not exist. Finally, even if no such computational factors proved to limit the possibility of artificial intelligence, one would still want to assess the existing state of the field and project the rate of progress likely to result from application of its present intellectual tools to the profound problems with which it must wrestle. The next five sections of the present article develop points relevant to the three kinds of limits defined in the preceding paragraph. A final section discusses certain other concerns, implicit in the debate between the enthusiasts of artificial intelligence and their opponents, which may explain some of the vehemence which has crept into this debate.