In this textbook the author endeavors to cover the large and growing field of artificial intelligence (AI) in some detail. While there are books that examine and discuss global perspectives on AI, they make no attempt to cover the diversity of theories and programs. A global perspective on the subject is provided by this volume, but in conjunction with an exhaustive survey of the field. It covers all recognized AI work in sufficient detail to allow a critiquie from general concerns to be anchored, whenever possible, in the structure of specific AI programs. It can be used as a supplement to other AI texts, providing broader perspectives on the wealth of details that such texts contain. It can also be considered as a companion to the current AI literature for it is only in conference proceedings and journals that these up-to-date details are usually found.
Table of Contents:
Preface xviii
1 AI: what is it? 1
definitions: what would it look like if I saw one? 1
True AI Story: 1.1-ELIZA meets PARRY:
the syntax is willing but the semantics is weak 7
a history of scaling down 8
categorizations of AI work 14
the goals of AI research 15
the heuristic programming approach 18
the Samuel phenomenon 22
2 AI and the Science of Computer Usage: The Forging 23
of a Methodology
how to use the essential tool? 23
first specify, then verify 26
the nature of AI problems 27
a methodology of incremental exploration 30
rapid prototypes to the rescue? 32
supportive environments 33
forging a new methodology 34
is AI so different? 37
3 The Major Paradigms 39
symbolic search spaces 40
planning intelligent solutions 45
SSSP infrastructure 54
the pivotal role of searching strategies 56
heuristic pruning 59
connectionism: a possible alternative? 62
connectionism: the second coming 64
on not losing their inhibitions 66
the need for decay 67
subsymbolic connectionism: the good news 68
when is an AI system like a piece of fine china? 70
subsymbolic connectionism: the real news 71
reasoning with amorphous complexity 72
the myth of empirical guidance 73
what's the stopping rule? 77
single-minded models 78
philosophical objections 79
potential solutions to the dilemma 81
formal analysis 81
software support systems 83
approximate translation-the truth about mendacity 84
the SSSP and the CP: integration, bifurcation,
or annilation? 86
simulated evolution: guess and try it out 89
'bad' paradigms 90
4 The Babel of AI Languages 98
it's all done by manipulating symbols 98
LISP 100
flexibility 100
the magic of recursion 101
code-data equivalence 103
the special assignment 105
lists of properties 106
PROLOG 108
the independence of declaration 109
loss of control: better or worse? 111
extralogical pollutants 116
negation as failure 120
verify or compute 120
bidirectionality 121
pattern matching 121
the promises of PROLOG 122
parallelism 122
a specification language 123
heuristic controls 124
object-oriented programming 125
programming environments 131
LISP environments 131
LOOPS 132
POPLOG 134
True AI -Story: 4.1. DIMWIT (Do I Mean What I Tell):
A PA (Programmer's Assailant) System 136
5 Current Expert Systems Technology (CEST) 139
experts with tunnel vision 140
the basic assumptions and the criticisms 141
what can be CESTed? 145
explanations and context sensitivity 146
updating knowledge bases and machine learning 150
let's dig deeper 155
logical decision making 159
human and computer decision making 161
classes of human decision making 163
connectionism: a possible answer? 165
knowledge elicitation 167
knowledge engineers and the third degree 167
automatic learning from examples 168
empirical techniques 168
CEST: where is it and where is it going? 169
6 Knowledge Representation: A Problem of Both 171
Structure and Function
why networks? 171
why neurons? 176
pandering to evolution: beware of classical reconditioning 178
neural architectures: in the beginning 179
knowledge representation: structure and function 183
the SSSP and the CP: representational issues 188
knowledge representation in the CP 189
functionally distributed representations 189
symbolic connectionist representations 189
winner-takes-all subnets 194
hybrid connectionism 195
totally distributed representations 196
path-like architectures in the CP 202
bath-like architectures in the CP 204
knowledge representation in the SSSP 209
logic-based representations 210
procedural representations of knowledge 213
semantic networks 217
elements of structured knowledge: frames,
scripts, and schemata 226
7 Vision: Seeing is Perceiving 228
bottoming in: operators canny, uncanny,
and cannyless 233
pixel processing 233
edges and lines 237
vertices or junctions 238
texture: a truly superficial feature 239
illumination, reflectance, and other sources
of nuisance 241
the intrinsic image 241
model-based vision systems 244
True AI Story: 7.1 247
beer cans, broomsticks, etc. 248
seeing as perceiving 250
oversight and hallucination 252
the modularity of human vision 255
eyeballs and nervous optics 256
biological feature detectors 256
human perceptual behavior 261
breaking up context 262
structuring top-down information 262
a cognitive model of word recognition 264
the eye of the robot 267
general theories of visual perception 273
the vision of connectionists 278
8 Language Processing: What You Hear is What You Are 283
natural language 286
what mode of natural language? 287
the goals of AI-NLP 289
natural language: the essential ingredients 289
phonetics and phonology 290
the lexical level and above 291
generation and analysis 291
natural language generation (NLG) 293
text generation systems 297
empirical guidance for NLG 298
natural language understanding (NLU) 299
syntax, grammars, and parsing
grammars 300
furious transformational grammarians
sleep curiously 303
transition networks: augmented and otherwise 305
unification and the new grammatism 308
semantic definite clause grammars (SDCG) 309
NLP and a formal complaint 314
semantics 318
the meaning of semantics 319
the atomic struture of meaning 320
the case of the missing-blocks world 322
True AI Story: 8.1 SHRDLU and a "SORRY" story 324
revolting computational linguists 324
scripted NLU and its dependencies 326
True AI Story: 8.2 Try it again SAM 328
the conceptual dependency notation 329
a Swale of a tale 331
True AI Story: 8.3 Another SWALE of a tale 332
giving semantics preferential treatment 333
bidirectional NLP 335
pragmatics? 336
machine translation (MT) 339
natural language interfaces (NLI) 341
networks for NLP 344
9 Learning To Do it Right 351
can we have intelligence without learning? 354
can we have AI without learning? 355
learning paradigms in AI 356
learning as the accretion of symbolic structures 358
learning as the adjustment of link weights 361
external tutoring: learning by being told 363
learning on the path 370
learning in a bath - taking the plunge 381
climbing hills because the 're there 384
rote learning: if it might be useful, store it 390
learning generalities 391
induction 393
overgeneralization and refinement 395
a first guess and generalization 397
True Al story: 9.1. Underneath the arches:
an everyday story of concept learning 399
competitive learning 404
learning particularities: removal of unwanted
generalization 406
EBG, or is it EBL? 409
the EBL viewpoint 416
mechanized creativity 421
learning by introspection 423
rediscovering things 427
learning by analogy 432
learning at the knowledge level 433
soaring through search spaces 437
the more you know the slower you go 443
on finding needles in haystacks 448
when to learn and what to learn 448
giving credit where it is due 450
unlearning 453
10 Foundations of AI: Can we find any? 458
foundations: why dig for them? 459
formal foundations 460
a disinterested user's guide to the FOPC 462
the curse of nonmonotonicity 467
logical odds and ends 471
True AI Story: 10 .1 It is not a closed world after all 471
the epilogic 474
methodological foundations 477
the roles of programs in AI 478
programs as theories 479
programs as experiments 483
rational reconstructions in AI 483
sorting out AI methodologies 486
philosophical foundations 488
there's nothing special about you, or me 488
building the foundations on the CP 490
undermining the foundations of the CP 491
total disbelief: let's not be Searle-ish 492
11 Prognostications, or W(h)ither AI? 496
abstract AI and concrete AI 496
is the mind an appropriate object for scientific study? 498
True AI Story: 11.1. Sand in the works 499
AI as a magnifying glass 501
AI: can it be practically useful? 503
AI: just wait till we get into parallel hardware 503
last words 506
References 507
Author Index 531
Subject Index
