00100	                GENERALITY IN ARTIFICIAL INTELLIGENCE
00200	
00300	
00400		Much research in artificial intelligence has gone into making
00500	programs that carry  out  particular  intellectual  activities,  e.g.
00600	playing  chess  or  proving  theorems.   Another approach has been to
00700	identify and investigate particular intellectual mechanisms  such  as
00800	tree  search or learning.  This effort has been certainly worthwhile,
00900	but attention also has to be  paid  to  the  creation  of  a  general
01000	artificial  intelligence  that  can  tackle  any  problem doable by a
01100	human. This problem is far from solved, and this paper is devoted  to
01200	discussing   the  various  approaches  that  have  been  studied  and
01300	identifying the ones I consider most hopeful.
01400	
01500		In the Turing lecture, it is appropriate to refer to Turing's
01600	work, but in this case it would be necessary even if this were not a
01700	Turing lecture.  Modern work in artificial intelligence started with
01800	Turing's 1950 paper "Can a Machine Think".  In this paper, Turing
01900	accomplished the following:
02000	
02100		1. He separated the technical problem of artificial intelligence
02200	from quasi-theological philosophical considerations by posing as a
02300	sufficient condition for intelligence the ability to pretend successfully
02400	to be human over a teletype line.
02500	
02600		2. He emphasized the universality of the stored program
02700	digital computer and proposed that future effort in AI be based on
02800	it rather than on building special purpose machines.
02900	
03000		3. He proposed as worthwhile AI problems game playing and the
03100	creation of a child-program that could reach adult intelligence by
03200	learning from its experience and from what it has been told.
03300