Tuesday, February 9, 2016
Ben Goertzel and others have criticized much of current mainstream work as being "narrow AI." (Artificial General Intelligence, Springer, 2007) Bringsjord and Schimanski have suggested we overcome this limitation using "one program for many tasks" as proposed in Newell's "20 Questions" paper (in Visual Information Processing, Academic Press, 1973). Where as "everyone is carrying out work on his or her own specific little part of human cognition" Newell proposed we "stay with the diverse collection of small experimental tasks, as now, but to construct a single system to perform them all." "it must be a single system in order to provide the integration we seek." This has been done to some degree by the connectionists. What a wide number of problems have been attempted using the standard backprop algorithm for example. I have been doing similar things with my Asa H.