00100	Minsky and Papert	AI Memo 252
00200	
00300	"blocks world" robotics environment and children's stories better suited
00400		to these studies than puzzle, game, and theorem-proving environments
00500	Thinking is based on the use of SYMBOLIC DESCRIPTIONS and description-
00600	manipulating processes to represent a variety of kinds of KNOWLEDGE--
00700	about facts, about processes, about problem-solving, and about computation
00800	itself, in  ways that are subject to HETERARCHICAL CONTROL STRUCTURES--
00900	systems in which control of the problem-solving programs is affected by
01000	heuristics which depend on the meanings of events.
01100	
01200	The ability to solve new problems ultimately requires the intelligent
01300	agent to conceive, debug, and execute new procedures.  Such an agent
01400	must know to a greater or lesser extent how to plan, produce, test,
01500	modify, and adapt procedures; in short, it must know a lot about
01600	computational processes.  We are not saying that an intelligent machine,
01700	or person, must have such knowledge available at the level of overt
01800	statements or consciousness, but we maintain that the equivalent of such
01900	knowledge must be represented in an effective way somewhere in the
02000	system.
02100	
02200	REASONING BY ANALOGY
02300		
02400		Make up a description DA for figure A
02500		Make up a description Dc for figure c
02600		Change DA so that it describes figure B, ie DA=DB
02700		Make a description for the way that DA was changed to DB
02800		Use D to change DC
02900	
03000	children use schematic descriptions to represent geometric information
03100		folded-out cube drawing vs perspective drawing
03200	children's drawings are nearer to internal representation
03300	perspective drawings are nearer to external world
03400		this is a good way of communicating with the interior representation
03500		by way of simulating the external world
03600	
03650	LEARNING STRUCTURAL MODELS
03700	discussion of the pass-oriented model of human visual perception
03800		not entirely wrong, or useless
03900		introduces HETERARCHY
04000	
04100	computational theory
04200	
04300	Perceptrons
04400	
04500	Guzman
04600	Rattner
04700	Learning
04800		primary question what to remember
04900		learning is not separate from other intelligent activity
05000	
05100	do not attempt unsound approaches to learning:
05200		learning based not on bookkeeping, but exploring structure
05300	learning to learn requires describing processes and manipulating
05400		such descriptions
05500	
05600	Piaget's conservation experiments
05700		model for the performance of the non-conservation child
05800	
05900		A scenario for considering descriptions for the child in
06000		the egg experiment.  
06100			QUANTITATIVE
06200				SPATIAL
06250					EXTENT means more
06275					SPARSENESS means less
06300				NUMERICAL
06400					counting
06500					one-one correspondence
06600			HISTORICAL
06700				IDENTITY  none added or taken away
06800				REVERSIBILITY  spreading out is reversible
06900	
07000		quantitative seems inconsistent, HISTORICAL SEEMS CONSISTENT
07100		
07200	[ASIDE, I suppose that the maturing is actually in coming up with
07300	these sophisticated descriptions]
07400	
07500		the secret of learning often lies in the discovery of descriptions 
07600		that emphasize the "essential" aspects of things or events, and omit
07700		or subjugate the "accidental" features.
07800	
07900	parameter modification
08000		severe limits on class of things learnable in this way
08100		end up only with coefficients, not a symbolic analysis
08200			which another program might use.
08300		alternative for game-playing
08400			instead of returning a low numerical value for a position,
08500			return a description of why the position is bad.
08600			Then the plausible-move generator can look for moves
08700			which defend that square or pin attackers.
08800			Or return a description of why the move is good to look
08900			for enemy replies.
09000		misconception about the significance of trial and error
09100			many small qualitative changes to procedures
     

00100	advantages of robots in automation
00200	
00300	a)  direct sensor input from tools in modes not natural for humans
00305	b)  high data rate from sensors to control to effectors
00350		Humans are
00400		faster when information can be input to their visual or
00500		auditory systems which are highly parallel, but much information
00600		cannot be directly input.
01000	To bring about a transfer of technology from the laboratory to pilot
01100	production, several concurrent phases are necessary:
01200		systematize hardware (cost conscious)
01300			Scheinman arm
01400			sensor study
01500			standard interface and pan/tilt for visual devices
01600			adopt commercial visual hardware
01700		systematize software and interaction
01800			hand/eye system
01900			drawing program
02000			experience with low/cost graphics
02100		ease of problem specification
02200			visual representation
02300			geometric editor
02400			visual input
02500			perceptual strategies
02600		standardize economical computer configuration
02700			alternative is de facto standards at lower level
02800			powerful force for guiding manufacturers of minis
02900		provide powerful modules to broaden engineering choices
03000		design inherent reliability.
03100			continuous self-calibration
03200			disaster detection and contingency planning
03300			wide latitude in tolerances
03400			low incremental cost in quality control
03500			continuous monitoring of function and product quality
03600	
04000	A plant would normally have a central time-sharing system which would
04100	monitor and schedule operations of machines and subsystems.  That
04200	provides a natural subdivision of abilities between the small computers
04300	which would control one or more machines, and the central system which
04400	would provide file storage and software support for large programs.
04500	Thus, we would find that the central computer would support:
04600		interactive job specification
04700			mechanical computer-aided design (CAMD)
04800		strategies
04900		complex vision for recovery (interrupt basis from minis)
05000	while the small machines would have, in addition to the arm control
05100	and contingency planning described above:
05200		visual feedback prediction and verification (Gill)
05300		edge and line finder (Hueckel)
05400		corner finder (Gill)
05500		region finder (Baumgart)
05600		color regions
05700		2-D template matching of shape
05800		access to large machine for contingency
06000	overall significance of information sciences in industry
06100		batch production
06200		control of scheduling, resource allocation, flow routing
06300		commonality of design
06400			find common fabrication and assembly operations
06500			find more common components
06600		increased understandibility and communication
06700			availibility of information
06800				keep track of delayed effects of design decisions
06900				cuts cost of preparation and transmission
07000				increased accessibility of information