Towards a Functional Architecture for Intelligence Based on Generic Information Processing Tasks

Abstract

C o l u m b u s, OH 43210 USA T h e level o f abs t rac t ion of m u c h of the w o r k in know ledge-based systems ( the ru le, f rame, logic level) is too low to prov ide a r ich enough vocabu la ry for knowledge and con t ro l. I prov ide an overv iew of a f r amework cal led the Generic Task approach that proposes t h a t knowledge systems shou ld be b u i l t out of b u i l d i n g blocks, each of wh ich is app rop r i a te for a basic t ype of p rob lem so lv ing. Each generic task uses forms of knowledge and cont ro l strategies that are character is t ic to i t, and are general ly conceptua l ly closer to doma in knowledge. Th i s fac i l i ta tes knowledge acqu is i t i on, and can produce a more perspicuous exp lana t ion of p rob lem so lv ing. The re la t ionsh ip of the const ructs at the generic task level to the ru le- f rame level is analogous to t h a t between h igh level p r o g r a m m i n g languages and assembly languages. 1 describe a set of generic tasks t ha t have been found pa r t i cu la r l y useful in cons t ruc t i ng d iagnost ic, design and p lann ing systems; d iagnost ic reasoning is used to i l lus t ra te the approach. 1 describe the Generic Task Toolset for cons t ruc t i ng knowledge systems, w h i c h embodies the Generic Task app roach. I conclude w i t h the imp l i ca t i ons of this approach for the func t iona l a rch i tec ture of in te l l igence. 1. Overview of the Paper