perm filename TOP10[ADM,DBL] blob
sn#188017 filedate 1975-11-22 generic text, type C, neo UTF8
COMMENT ā VALID 00006 PAGES
C REC PAGE DESCRIPTION
C00001 00001
C00002 00002 MINE
C00004 00003 NEWELL
C00005 00004 BLEDSOE
C00006 00005 WINOGRAD
C00007 00006 NILSSON (ordered)
C00008 ENDMK
Cā;
MINE
Difficulty of acheiving AI: Recognition is hard (Speech)
Intergration is hard (Natl Lang)
Search is combinatorial (Chess)
From Dendral/Mycin: importance of choosing a good task domain
From Dendral: a program that is actually useful in the real world
Heuristic Search: cataloging existing methods
re-discovering already-existing methods
discovering new methods
admissability of algs using heur guidance criteria
New languages and features; Automatc Pgmming
Need for interactive debugging of complex systems.
Need for large systms to be self-explaining.
Today's AP pgms are tomaorrow's AI languages
Property list; obj/attrib/value triples.
Modular representations of knowledge
Procedural repr.
Asociative memory nets
Need for uniformity of repr.
In a dialog, or in really any large interactive pgm, it is important for the
speaker to (i) possess a good model of the listener, (ii) maintain
a good, current model of the listener's state of mind/memory, etc.
(iii) estimate (conservatively) what the listener will wrongly infer
(if he had to guess at this moment), (iv) tell him enough to correct
these misconceptions, but (v) not tell him any more than that.
NEWELL
Computer used as a symbol processor
Intelligence can be described by sym-processing
Heur search: primacy of search in problem-solving
Goal-directd processing: goals as organizing structures
Perception demands cognition (we can see only what we can comprehend).
Resolution
Cooperating knowl sources
Induction by modelling and hyp. formation techniques
Means-ends analysis: GPS
Goal-oriented langs
Augmented Transition Net gramars
BLEDSOE
Symbolic proc = intell
Power of heur knowl
Heur search
Semantics importance
Means-ends ana.
FOL pf-checker
Auto thm-provers using sem knowl
Algebraic simplif and reduction in Thm-provers
Ineffectiveness of uniform pf procedures
WINOGRAD
Problem solving = search
ATN grammars
Multiple know sources
Guzman, Waltz stuff
Knowl-based systems (sem)
Process understanding (self-explaining)
Logic as problem
Explicit manip of control
Genl search methods
NILSSON (ordered)
Family of repr structures (PBG, Nets, ...)
Interpretation of repr
heur search
embedding key resoning abilities into pgmming langs
Knowl molecules (minitheories)
Automatic synthesis of action sequences
Global data bases = blackborad
Use assertions for control model