User:CharlesGillingham/Notes/McCorduck

general quotes
writes "Artificial intelligence in one form or another is an idea that has pervaded Western intellectual history, a dream in urgent need of being realized." ✅ She continues "Work toward that end has been a splendid effort, the variety of its form as wondrous as anything humans have conceived; its practitioners as lively a group of poets, dreamers, holy men, rascals, and assorted eccentrics as one could hope to find—not a dullard among them. Its visionaries have lifted our spirits and made us trascend our own species, its poets have told us things about ourselves we never suspected, and its fast talkers have set everybody's teeth on edge."

"'Yet for all its absurdity, we find the idea irresistible. Our history is full of attempts—nutty, eerie, comical, earnest, legendary and real—to make artificial intelligences, to repreduce what is the essential us—bypassing the ordinary means. Back and forth between myth and reality, our imaginations supplying what our workshops couldn't, we have engaged for a long time in this odd form of self-reproduction."

"'I like to think of artificial intelligence as the schientific apotheosis of a veneralbe cultural tradition, the proper ssuccessor to the goldern girls and brazen heads, disreputable but visionary geniuses and crackpots, and fantastical laboratories during stormy November nights. Its heritage is singularly rich and varied, with legacies from muth and literature; philosophy and art; mathematics, science, and engineering; warfare, commerce, and even quackery.' pp=34-35"

Myth

 * Golden robots of Haphaestus, Talos, Pandora, Pygmalian and Galatea pp=4-5 ✅

Sacred statues
pp=6-9 ✅
 * old Testament prohibition of sacred statues p=8
 * pope Sylvestor's bronze head. p=9
 * brazen heads p=12-13

Automata

 * Hero of Alexandara 7 ✅
 * automaton: Haroun al-Rashid's gift to Charlemagne 10 ✅ (footnote)
 * in general, the duck 16 ✅ (footnote)
 * the Turk 17 ✅
 * 1915 Leonardo Torres y Quevedo's chess automaton, publishes speculation about thinking and automatons. 59-60 ✅ (footnote)

Zairja and Ars Magna
zairja pp=10-12, 37 "The Ars Magna was remarkable because, along with its Arab forerunner the zairja, was based on the assumptionat that human thought could be mechanized."

Alchemy

 * parcelsus 13-14 ✅
 * golem 14-15 ✅

Fiction
fiction: pp=17-25 "if machines hadn't lost their mystery, they certainly had lost their novelty" p=18
 * Faust repeats Parcelsus p=18
 * Frankenstein pp=19-25 ✅
 * Capek p25 ✅
 * Asimov pp-25-26
 * 1727, Gulliver's Travels: there is a thinking machine on Laputa "a Project for improving speculative Knowledge by practical and mechanical Operations " by using this "Contrivance", "the most ignoratn Person at a reasonable Charge, and with a little bodily Labour, may write Books in Philosophy, Poetry, Politicks, Law, Mathematicks, and Theology, with the least Assitance from Genius or study." Q in M p. 317  This is no doubt a paradoy of Hobbes and Leibniz's ideas about mechanizing thought. ✅

Babbage
Babbage pp=26-34 Lord Kelvin characterizes it "substituting brass for brain"

Philosophy
38-46 ✅
 * Descartes 36-40
 * Leibniz 41-42 ✅
 * Hobbes 42 ✅
 * Humes "atomic impressions 43
 * Le Mettrie "L'Homme Machine" 43-44
 * Kant whatever ...

Logic
48-51
 * Boole 48-51
 * Advice Taker 51 ✅
 * Robinson's resolution 51 ✅

Zuse
Zuse 61-62 ✅

Turing
62-76
 * Turing Machine 63 ✅
 * Enigma and Ultra 64-66 ✅
 * Intelligent Machinery 67-70
 * Computing Machinery and Intelligence 70-71 ✅

Von Neumann
76-81

Meat Machine
The fate of brain simulation research in the 40s, 50s and 60s.

Neural networks/Gen 0
51-57, 88-104 ✅
 * Cybernetics 52-55 ✅
 * Connection to biology 53-54
 * Pitts and McCullough 51-57, 88-94 ✅
 * Ratio Club 95 ✅
 * Teleological Society 83 & 94
 * Turtle (robot) 98 ✅
 * SNARK 102 ✅

Perceptron
104 - 107 ✅

Pamela McCorduck writes"'He was a press agents dream', one scientist says, 'a real medicine man,' Quoted in M, p. 105

I, J. Goode (now a professor at Virginia Polytechnic Institute "IN fact, it was fairly clear that Rosenblatt's original work did not contain real proofs of the results that he claimed. It was later work by H. D. Block and also by Seymour Papert that showed there were certain theorms that could be proved." Quoted in M, p. 105

"He did irritate a lot of people but he also charmed at least as many, and I count myself among them. Just when you were thinking that Frank didn't have another trick up his sleeve, along he'd come, and he'd be so darn convincing, you know, he just had to right." W. W. Bledsoe of the University of Texas. Quoted in M, p. 106

The end of Neural network research
"And so much for the hope of making a machine think by trying as literally as possible to imitate the brain, the meat machine, at the cellular leel. It didn't work, as John von Neumann had said it wouldn't." M, p. 107 (John von Neumann argues that you can't imitate the brain on M p. 79)

Shannon
Shannon's work 120-123, Chess Playing p. 120, Computers and automata p. 121, His mouse 123

Dartmouth conference
111-120, 123-136 ✅
 * Organizing it p=111-112
 * The name "artificial intelligence" p=114-115 ✅
 * Debut of the logic theorist 123-129 ✅
 * Rhe "invisible college" that descended from Dartmouth; the clannishness of the participants. Everything was done by Dartmouth attendees and their students. 129-130 ✅

The Money
The money: to Stanford, CMU, MIT. p. 131 (McCorduck connects this to the Dartmouth Conference and its "invisible college"). ✅

Newell & Simon
139-161
 * Newell
 * The way RAND worked p. 139
 * Newell at RAND, etc p. 140-145,
 * Oliver Selfridge's work inspires Newell, Newell's epiphany 157-158 "It all happened in one afternoon." p. 158 ✅
 * Newell's career, 159-161
 * Simon 146-151, 153-156.
 * Simon appears "unsettlingly inhuman",
 * Simon sees the map: 148 (Simon's epiphany) ✅
 * Newell & Simon The nature of their partnership, 159

Logic Theorist
p. 161-170 ✅ ✅
 * January 56: to graduate class " over christmas, Allen Newell and I invented a thinking machine." p. 138 ✅
 * Shaw, 164-165 p. 169. (Newell says "Cliff himself was the genuine computer scientist of the three.") ✅
 * List processing, IPL, 167-168 ✅
 * Wife and children simulated it. 168 ✅
 * Debut at Dartmouth 123-129	 "the evidence is that nobody save Newell and Simon themselves sensed the long-range significance of what they were doing." 124 ✅
 * A quote: The first AI program, the Logic Theorist, proved theorems in logic, "proof positive that a machine could perform tasks heretofor considered intelligent, creative and uniquely human." M, p, 167 ✅

Information processing / Physical symbol system
On the information processing model, or physical symbol system hypothesis: "This view would come to be central to their later work, and in their opinion, as central to understanding mind in the twentieth century as Darwin's principle of natural selection had been to understanding biology in the nineteenth century." p. 153 ✅ ✅

Nathan Rochester
'''Simon on Rochester" Simon speaks fondly of Rochester's assembler for the 701, p. 156

Chess and checkers
Games in AI 171-192
 * Starchey's program 73
 * Von Neumann, game theory p. 172
 * Samuel's checkers p. 173-180
 * Early chess players: in fiction "Moxon's master", Torres y Quevedo, Baron von Kempelen 182
 * Bernsteins' chess p. 180-186
 * Other chess games, to 1975 or so 187-192

IBM's abandonment of AI
p. 186-187 Hype, p. 186, backlash 187

Hebraic/Hellenic
The Other, Hebraic/Hellenic p. 197-199 McCorduck's categorization of objections, p. 199

Emotional reactions/ the other
Emotional reactions The other p. 197, Snow white's wicked stepmother: who's the smartest? p. 202 Hiller & Isaacson on creativity p. 203

Mortimer Taube
Mortimer Taube, p. 204-209 the book is Computers and Common Sense: The Myth of Thinking Machines, "He seems to have been the first of a series of critics of the field who's emotions were as deeply engaged as their intellect" M, p. 209, Italics hers.

Dreyfus
p. 211-239 ✅ ✅

Newell & Simon's exagerrated predictions
p. 218-225

(the same ones, of course)

Simon, defending his predictions, blamed two things: man-hours and the commonsense knowledge problem. "There we just vastly underestimated two things: first, how little, how few man years would go into this; and second how much very specifci knowledge had to be poured into it. Maybe we left out some other things, but those are the only things I'm willing to admit we left out!" S, quoted in M, p. 220

Herbert Simon said that they had "vastly underestimated" the commonsense knowledge problem and the man-hours. In reference to the Chess prediction, Minksy also blames man-hours. p. 220.

Simon also blames the timidity of some researchers. p. 221 (really!)

He complains that his predictions are no worse than others made in other fields. Everybody's doing it. "You someone can go around with the smallest scintilla of evience and make a new kind of universe that expands or contracts or is permanently in one state or another. Cosmologists go around doing this all the time, and tey're regarded as good scientists in astronomy because that's part of the mores of that filed.... Biologists on the whole are much more careful, in that sense of careful" (I don't buy this for a second, by the way.)

"It is the point of Drefys's book that human and artificial intelligence are in fact quite different---in particular, human intelligence is unique. Not only that, but a great misunderstanding accounts for public confusion about thinking machines, amisunderstanding perpetrated by the unrealistic claims researchers in AI have been making, claims that thinking machines are already here, or at any rate, just around the corner." M p. 212

The Critique
211-216

Alchemy and AI p. 211, What computers can't p. 212

Based on Heidegger and Merleau-Ponty, p. 213 ✅

Four assumptions: p. 213 ✅

She mentions a few ideas, p. 214. Fringe consciousness "that allows us to zero on the important aspects of, say, a chess game, without losing total awareness of other possible moves." The phenomenological view: the role of the body, the situation, the purpose (for seeing what's relevant). P.M characterizes all this as: "Computers, unlike human beings, have no means to discriminate between the imporatant and the unimportant," p. 215

History, Reaction
All this material is in: ✅

GPS & Cognitive Simulation
p. 245-250

reasoning as search / she calls "means ends analysis" p. 247-248 ✅, heuristic 246 ✅

cognitive simulation / use of psychological experiments of O. K. Moore, and "similar experiments underway at RAND 1955 and 1956" p. 246 "GPS was ... the first program ever developed as a detailed simualtion of human symbolic behavior; as such it clarified." p. 250

McCarthy & Logical AI
McCarthy up to 1959: 251-259


 * Advice Take 51 & 251 ✅ "In order to do [Advice Taker], you have to be able to express formally that information that is normally expresssed in ordinary language. As far as I'm concerned, this is the key unsolved problem in AI. I uncovered the problem in 1958 and it's still unsolved." McCarthy, Q in M, p. 254
 * Lisp 252
 * Time Sharing 252 & 253
 * Robinson's resolution 51 & 255 ✅ Robinson's resolution "[Robinson] felt heavy obligation to extract the AI researchers from the pit into which they were falling, the pit of the combinatorial explosion" Feigenbaum, Q in M, p. 255

Knowledge Revolution
✅

In general
"As the 1970s drew to a close, knowledge representation was perhaps the most hotley debated topic in artificial intelligence" M, p. 314

"Researchers had come to beleive that the great lesson fron the 1970s was that intelligent behavior depnded very much on dealing with knolwedge, sometimes quite detailed knowleddge, of a doman where a given taks lay" p. 421

From the perspective of robotics research at MIT
For Joel Moses, the turning point was as early as 1967, when he began to realize that the approach of AI's first generation was not going to work. "The word you look for and you harldy ever see in the early AI literature is the word knowledge." Moses emphasizes that the turning point came with robotics research, and projects like Bobrow's STUDENT: problems which were specifically concerned with the real world, in all it's complexity. M, p. 266.

A few researchers, such as Joel Moses, had been arguing for a knowledge-based AI since the sixties. Moses claims he was arguing for the "primacy of experties" (Q in M p. 266). Seymour Papert was also an early advocate. p. 266 & 276. He cites resistance from AI's founders such as Minsky and Newell (p. 267) Moses remembers Papert's frustration with AI's first generation. "How can we can get these guys to listen" P, quated by Moses, quoted in M p. 267

Edward Feigenbaum's group at Stanford also pursued knowledge

In the context of NLP and disambiguations
p. 298-99 McCorduck argues that disambiguation requires context, and that context requires knowledge.

"We understand in context ... we've agreed upon a topic of discourse, that we talk about it in aspecific setting, and that we have some knowlege about the world and each other's ideas."

AI researchers were beginning to suspect--reluctantly, for it viiolated the science canon of parsimony--that intelligence might very well be based on the ability to use large amounts of diverse knowledge in different ways." M p. 299

Common sense knowledge problem
"it was impossible as well to give a computer program all the knowledge a human being brings to a conversation" M p. 300 ✅

Symbolic Robotics: Shakey, Cart, Freddy, Mit's arm
Planning robots, using symbolic reasoning, descended from GPS.


 * SRI: SHAKEY p. 259, 268- Nils Nillson, project leader
 * Stanford: CART p. 455 Hans Moravec
 * Edinburgy: FREDDY p. 268
 * MIT's Hand Eye Radio p. 267

SHAKEY / ARPA withdraws support for SHAKEY
268-271 ✅

Nils Nillson (in charge), Bert Raphael

How it illuminated the need for uncertain reasoning. p. 270 STRIPS p. 271

ARPA withdraws support when????

Charles Rosen (director of SRI's artificial intelligence program in the late 60s and 70s) "I think the reasons were more political than technical". We were told later that there were too many people raising issues of Shakey being a dangerous thing to have." Q in M p. 271 and he add "which seems to me a little silly." Given that the military has developed and stockpiled an enormous number of profoundly dangerous machines, and Shakey was nothing but HUMOROUS DESCRIPTION OF RATTLY OLD SHAKEY.

Nils Nillson suggested that they withdrew support because "no immediate military application could be seen.

The Darrach article
The infamous Life Magazine artice, written by Brad Darrach, which contained misquotes, material taken out of context and out right fiction, in an attempt to dramatize Shakey's capabilities and potential military applications. M p, 272-274 ✅

Languages, Scenes, Symbols, and Understanding
NLP: p. 277-286

Why NLP is important: "If we could figure out how to make a computer 'understand' our language, we'd finally know just what understanding and even language is all about." M p. 280.

Automatic translation
"It would be be impossible to provide a machine with a world encyclopedia, Bar-Hillel went on, much less the capability to infer new concepts from the facts in such an encyclopedia. Therefore, mechanical translation was impossible; therefor, case closed. (Bar-Hillel, 1964)." M, p. 281
 * Automatic translation p. 280-281 ALPAC report, p. 281 ✅
 * Disambiguation and commonsense knowledge
 * Yehoshua Bar-Hillel's ciritique of machine translation. The problem is disambiguation: "the box is in the pen" etc. p. 281.
 * quote on the commonsense knowledge problem (M describiing Bar-Hillel

Early NLP systems

 * Conversation machine Could muse about the weather. I.E.S.Green, Edmund Berekeley, Calvin Gottlieb 1959. p. 282
 * Oracle (program) ??Not clear A.V. phillips (under John McCarthy (computer scientist)) 1959. p. 283
 * Baseball conversation program Bert F. Green 1963 (in Computers and Thought) p. 283
 * SAD SAM Kinship relations Robert K. Lindsay (under N&S) 1963 p. 284
 * ?? Geometric analogies (in english, I assume) Thomas Evans p. 285-286
 * Semantic information retriever (SIR) Bert Raphael 1964 p. 286
 * STUDENT (computer program) Algebra word problems Daniel Bobrow 1964?? p. 286 ✅

Minsky
Minsky's management style: p. 287-288. "his students feelings for him verge on adoration" Mp. 288 Compares it to weeding; he just gets rid of bad ideas. Mp. 287 he imbues them with "a self confidence that borders on arrogance."

On learning (is this important?)--- "It took us a long time to realize-- and people elsehwere still haven't-- that in a sense, once you have the right kind of descriptions and mechanisms, learning really isn't important.. the peoplme in learning now as we see it is how do you decide what you want to have in your memory," says Minksy. McCorduck observes "The answer to the problem of induction was again description, as it was to the problem of understanding." M p. 288-289.

Vincent Giuliano's critique of NLP in 60s
Guiliani's crtique is comment attached to Simmons R.F. 1965 "Answering English questions by computer" British Journal of Psychology, 58, May

"I tend instead to see evidence mainly of motion, with littel real evidence of progress" Q in M p. 289. "The existence of procedures of alchemy does not create a science." So Dreyfus' analogy had already been used.

ELIZA, DOCTOR, PARRY
p. 291-296 ✅

DOCTOR passes the Turing Test (p. 281) ✅

Weizenbaum helped COLBY to build DOCTOR (p. 292) Inspired by STUDENT and SIR (p. 292) Weizenbaum recalls coming up with the idea (p. 292-3)

Weizenbaum's paper announcing ELIZA: "ELIZA---A Computer Program for the Study of Natural Language Communication between Man and Machine" A title which, McCorduck notes, "squelching anybody's idea that his work had been about psychotherapy... He was dismayed to discover that in the publicatiosn lag that take place in nearly all professional journals, somebody else had jumped the gun. Kenneth Colby published a short note in the ''Journal of Nervous and Mental Diseases" stressing the therapeutic aspects of the program... What additionally irked Wiezenbaum, and helped accelerate the split between him and Colby, was the feeling that Colby had seized Eliza and made it his own, under the name of DOCTOR, without giving due credit to Weizenbaum". p. 295-296

Wiezenbaum doesn't think it has therpeutic use, Colby thought it did. p. 296

Blocks world
p. 299-305 ✅

Winston's vision work p. 299 (David Waltz too p. 300)

Microworld motivated by commonsense knowledge problem. p. 300.

SHRDLU
p.300-305 ✅

She quotes the published conversation (also in Crevier) in a nice paragraph (rather script form) --- this would make a better blockquote for the SHRDLU article. p. 301

symbolic. p.302

Winograd didn't assume that "categorization generally neither consistent, nor parismonious, nor complete" -- that it is pragmatic. p. 302

"Some scientists even argued that SHRDLU was essentially unextendable." p.

Self-consciousness
Gerald Sussman's improved version of SHRDLU, called HACKER, had "the ability to exanmine its own problems solving oals and actions so that it was able to supply this debugging expertise to its own reasoning. In short, it had self-consciousness'". m p. 304

Frames & Scripts
p. 305-306 ✅

"even experts are hard pressed to tell the difference between scripts and frames and chunks." p. 305

Some schank examples: "I hit the boy with the firl with the long hair with a hammer with a vengeance." (ambiguity of with) "The old man's glasses were filled with sherry" Q in M, p 306

Speech understanding: Hearsay (S.U.R.)
p. 306-313 ✅

Bell labs attempt in the 50s ends in failure. p. 309

She writes "in 1976, the Carnegie Mellon project cam in on target. It was not only on time, it not only met the specifications, but it was withing budget." p. 309. Contrary to Crevier!!

Reddy
310-311

Procedural/Declarative
p. 313-317

As a result of Hearsay (and inspired, I am assuming, by SHRDLU).

M believes that the procedural declarative distinction ... dynamic knowledge, not static ... knolwedge as a set of actions or reactions ... To M, the procedural approach "share seom ideas with epistemological philosophies of the past, including, of all things, phenomenology. These ideas include the situation; the notion of a dynamic, flixible, and often indosyncratic response to the situation; contingency; and a sense of pirpose or goals." p. 314

"Minsky holds that most precomputer phenomenologists were philosophers in search of an information-processing vocabulary, and would probably have been in the thick of articial-intelligence research if the computer had existed to them its reich possibilities for metaphor and modeling." p. 314

(My own view would be that Minksy hear is probably talking about frames and knowledge which, I think, are a flailing attempt to create contexts and gestalts).

Feigenbaum

 * EPAM. 322-323 (a form of cognitive sim)
 * Computers and Thought p. 323 326 Defined the field for a generation. Deliberately excluded neural net research, such as the perceptron.

DENDRAL
p. 327-335 ✅

Dendral

Knowledge acquistion phase of Expert system development has roots in cognitive simulation: find out how people solve problems. What experts do to solve problems. Express this as knowledge. Feigenbaum was CIT during the heyday of this phase of N&S. p. 329

Technical aspects n DENDRAL (p. 334)
 * 1) Generate and test, from GPS
 * 2) Condition-action rules. "Situation leads to action"
 * 3) Specific knowledge
 * 4) Line of reasoning must be explicit, and thus comprehensible to human expert
 * 5) Coordinate many source of knowledge

Papert
p. 336=349 LOGO 337-350

Can a Made-Up Mind Be Moral
Ethics of AI

Ethics of AI
Solzhenitsyn's First Circle -- speech understanding program used to spy on people p. 308 ✅

Weizenbuam's Critique
p. 356-357 ✅

History (p. 362-363)
 * 1972 "On the Impact of Computers on Society" paper
 * 1973 Public debate with Colby (unfortunately gives people impression that this is personal
 * 1976 Computer Power and Human Reason the book

According to M, it has three main points: pp. 356 & 374-376
 * 1) There are domains where computer are not to intrude (jobs that require love or care)
 * 2) Where computers would represent an attack on life iteslef
 * 3) Where the effect are irreverible and not entirely forseeable (Buchanan points out anything interesting will have these properties)
 * 4) When invovles respect and love (McCorduck counters that, speaking for women and minorites "I'd rather take my changes with an impartial computer.")
 * 5) Most of the work in the field is not science but technique. (performed by pathological and megalomaniacal hackers) (McCorduck points to the Nobel, Turing Award, etc)
 * 6) We've embraced the machine metaphor as a description of ourselves and our intsitutions much too readilyu, that in this embrace we're in acute dange of uielding what is essentitally human. An "atrohpy of the human spirit'' that comes from thinking of ourselves as computers. (M replies that there wasn't any Eden where we had self-respect or respect for each other.)

M describes the books as "a cry from the heart" p. 374

My thought on three: clearly we are more awed by knowledge than mystery. Anyone who has spiritual bent tends to fill each mystery with knowledge, special knowledge, acquired in moments of insight. What we regret is not the loss of mystery, it is the loss of this special knowledge we already had. And the introduction of new mysteries where there were certainties. The mystery argument against science is stupid.

Colby dispute again. p. 363-366 "What really troubles him is what he call the con job, that DOCTOR represented having the potential to help psychiatric patients. " M compares it to the introduction of drug-based psychiatry. Colby argues that there aren't enough psychiatrists to go around, and argues "to not explore the use of the best tools and intrucments aviaable is immoral since it violates a basic principlce of the helping professions whiwh are devoted to the relief of suffereinfg of everyone." Q in M p. 365

McCarthy's reply: "When moralizing is both vehement and vauge, it invites authoriatrain abuse either by existing authority or by new political movements."

Joshua Lederberg (who worked on DENDRAL), agreed that (as M puts it) "world knowledge underlying human understanding ... needs the life-long experience of having been human" (although he disagreed that brain is unknowable). p. 371

Forging the Gods
McCorduck's thesis

Fractioning into subproblems
p. 421-425 ✅ ✅ No common language or set of problems, Newell 1982. Neat vs. scruffy (1980s)

Around various problems or tools (later, I'm guessing she means 90s or so) "The rough shattering of AI in subfieldsvision, natural language, decision theory, genetic algorithms, robotics ... and these with own sub-subfieldthat would hardly have anything to say to each other." M, p. 424 ✅

Around various knowledge representation languages (an issue in the 80s).

Genetic algorithms
p. 426 John Holland develops.

Strategic Computing Intialitive
p. 426-432 ✅

for AI Winter article: In 1983, in response to the fifth generation project, DARPA again began to fund AI research through the Strategic Computing Initiative. As originally proposed the project would begin with practical, acheivable goals, which even included strong AI as long term objective. The program was under the direction of the Information Processing Technology Office (IPTO) and was also directed at supercomputing and microelectronics. By 1985 By 1985 it had spent $100 million and 92 projects were underway at 60 institutions, have in industry, half in universities and government labs. (Note that all this money isn't going to AI) p. 426-429

Jack Schwarz, who ascended to the leadership of IPTO in 1987, dismissed expert systems as "clever programming" and cut funding to AI "deeply and brutally," "eviscerating" SCI. Schwarz felt that DARPA should focus its funding only on those technologies which showed the most promise, in his words, DARPA should "surf", rather than "dog paddle", and he felt strongly AI was not "the next wave". Insiders in the program cited problems in communication, organization and integration. A few projects survived the funding cuts, including pilot's assistant and an autonomous land vehicle (which were never delivered) and a battle management system, which (as noted above) was successful. ✅

Rise of Expert Systems
p. 434-435 ✅

Collapse of the lisp market p. 435 (mentions moore's law explicitly). ✅

Limits of expert systems: "They came programmed in a language nobody in the mainstream part of the firm and ever heard of, and required a champions to usher them across internal borders." p. 435 ✅

Expert systems are used to approve credit p. 435

Fifth Generation
436-441 ✅

p. 436 "converse with humans in natural language" and "understand speech and pictures"\

SCI (Strategic Computing Initiative) was in direct response to FGI. "Bob Kahn and Richard DeLauer of DARPA claimed later to be skeptical of Japanese abilities to pull this off, but were not above exploiting congressional and military alarm to help acquire funds for Strategic Computing Initiative." p.437

"Two and a half decades later, we can see that the Japanese didn't quite meet all of those ambitious goals." p. 441 ✅ (in footnote).

AI Winter
"AIs major funding agnecy in the US, DARPA, was cutting support and asking for precise, measurable, and therefore incremental results." p. 442, McCorduck analyzes "AIU research didn't stop, but it becmae more "normal'', to use Thomas Kuhn's description of one kind of sicnec, as distinc from "revolutionary", the other kind."

Searle
Chinese room, 444 to 446 ✅

"By buying this image, the reader is unwittingly sucked into an impossibly unrealistic concept of the relation between intelligence and symbol mainpulation" Hofstadter and Dennett, 1981 Q in M p. 444

The Society of Mind
p. 444-447

Penrose
p. 448-449

McCarthy to penrose
"McCarthy concluded that some future programs would be able to answer what it feels like to be a computer "based on their ability to observe the reasoning process that their programmers had to give them in order that they could do their jobs. The answers are unlikely to resemble those given by people," because it won't e advantageous to give programs the kind of motivational and emotional structure we have inherited from our ancestors." (McCarthy, reply to penrose, 1990, quote in M, p. 448)

McCarthy on self awareness
"McCarthy concluded that some future programs would be able to answer what it feels like to be a computer "based on their ability to observe the reasoning process that their programmers had to give them in order that they could do their jobs.(McCarthy, reply to penrose, 1990, quote in M, p. 448)

McCarthy on emotion and AI
"it won't be advantageous to give programs the kind of motivational and emotional structure we have inherited from our ancestors." (McCarthy, reply to penrose, 1990, quote in M, p. 448

SOAR
p. 450-451

Brooks, Moravec, etc.
p, 454-462 ✅

Moravec's Paradox
Rodney Brooks points out that intelligence, according to early AI research, was "best characterized as the things that highly educated male scientist found challenging" --- chess symbolic integration ., proving mathematical theorems, and solving complicated word algoebra problems. "The things that children of four or five years could do effortlessly, such as visuually disctinguising between a coffee cup and a chair, or walking around on two legs, or finding their way from their bedroom to the living room were no thought of as activities requiring intelligence. Nor were any aesthetic judgements included in the repertoire of intelligence-based skills." p. 456 ✅ ✅

Nouvelle AI's
Allen (robot), Herb (robot) and six legged Genghis (robot). All able to do things no robot had ever done before. Non-representative, etc. p.457-8

"Genghis appeared to have intentions, but no intentions were internally represented." M p. 458.

Brooks writes "the software itself was certainly not profound. It was rather straightforward, in fact. The software's behavior, however, was profound." Q in M p. 458

1997: Sojourner explores mars

Later: works with Sony on the AIBO, and builds the Roomba

Brooks as scruffy
The scruffy approach was applied to robotics by Rodney Brooks in the middle 1980s. He advocated building robots that were, as he put it, Fast, Cheap and Out of Control (the title of a 1989 paper co-authored with Anita Flynn). Unlike earlier robots such as Shakey or the Stanford cart, they did not painstakingly build up representations of the world by analyzing visual information. The simply reacted to their sensors in a way that tended to help them survive and move. M, p. 454-459 ✅

2ND AI WINTER
P. 460

McCorduck describes the funding cutbacks of the late 80s as a "guillotining" of AI.

Mind Children
p. 460

"An historian might have some difficulty reconciling Moracec's intriguing speculations with the simultaneous fact of DARPA's guillotining of AI at the end of the 1980s. To be sure, some of his more conservative colleagues just rolled their eyes and groaned." M p. 461

Breazeal
Cynthia Breazeal's Kismet (robot) p. 463-464

Fuzzy logic
Lofti Zadeh unpopular in U.S. discovered by Japanese and put to work p. 470-1

Intelligent Agents
p. 471-478 In the late 1980s, Allen Newell had declared that it was time for the sub-fields of AI to merge and create 'complete intelligent agents'" M p. 478 ✅

Deep blue
p. 480-483 ✅ Is it intelligence?

Neats vs. scruffies
pp. 421-424, 486-489 ✅

Debate in 1983
p. 421-424 ✅ ✅

Nilsson, in his presidential address to AAAI in 1983. McCorduck writes "was AI empirical (the Scruffies?) or was a theory based technical subject (the Neats)? This, he said, was a nonissue; the field needed both."

Nilsson argued strongly that "much of the knowledge we want our programs to have can and should be represented declaratively in some kind of declarative, logic like formalism. Ad hoc structures have their place, but most of these come from the domain itself." Nils Nillson, presidential address to AAAI in 1983 (quoted in McCorduck, p. 421-422.)

Alex P. Pentland and Martin Fischer of MIT argued in response that "There is no question that deduction and logic-like formalisms will play an important role in AI research; however, it does not seem that they are up to the Royal role that Nils suggests. This pretender King, while not naked, appears to have a limited wardrobe." Pentland and Fisher 1983, quoted in McCorduck, p. 423-424

"Hegemony" of the neats
486-487 ✅ ✅

"It will probably take one or two more generations of students to debug the muth that AI is ad hoc," writes Victor Lesser (an AI researcher at Amherst, "For me personallty, the major change in AI over teh last 25 years is my understanding that anything I do these days must, of necessity but also for reasons of intellectual honesty, be more formally based either in connecting my research to some formal system of in its analysis through formal techniques." Q in M p. 486-487.

"Much of the progress in AI in the 1980s and 1990s was based on mathematical techniques drawn from fields such as probability, statistics and decision theory." M p. 487

"As I write, AI enjoys a Neat hegemony, people who believe that machine intelligence, at least, is best expressed in logical, even mathematical terms." M p. 487.

Some argue that scruffy methods will be needed eventually
Takeo Kanade says, for example, "it is time for us to remarry" talking about "neat" research in machine vision, and how eventually it will require knowledge, and he jokes, we were married to soon, and no wonder we got divorced. p. 488-9

"Computer scientists, with their typical backgrounds from logic and mathematics, enjoy creating elegant and powerful reasoning methods," writes Edward Feigenbaum, "But the importance of those methods pales in importance to the body of domain knowledge---the artifact's knowledge base." Q in M, p. 489

Cyc is scruffy ADD ME TO AI under Scruffy Symbolic AI
"A determinedly scruffy enterprise is Cyc'' M p. 489

Not anthropomorphic
M argues that the new neats "felt no particular obligation to be faithful to human models of intelligence". p. 487

M, reading Randy Davies AAAI presidential speech, equates Neat with logical and scruffy with psychological.

Computational Creativity
p. 490-494 AARON (computer program), Bruce Buchanan, EURISKO.

Feigenbaum test
p. 503-505 Described in Jan. 2003 Journal of Computing Machinery

Material for AI behand the scenes/under different names
"We are calling our area of interestest Cognitive systems,=== the foldks that never like ;AI' seem to be able to deal with that." Ron Brachman, head of IPTO in 2004. M p. 510

Applications, p. 413-415

National Association of Security Dealers uses AI applications to detect suspected fraud. M p. 414

Virtual players in electronic games. M p. 414

Train and airline scheduling uses "bin-packing" and "planning" algorithms developed by AI. Red Pepper Software never mentions AI; their applications are "scheduling software." M p. 414

Triology which sells the last descendent of XCON (a program called R1), does not mention the terms AI or expert system in its literature. They simpkly call it "the Sales Builder". M p.414-415

Dystopian
p, 494-501

Bill Joy's Dystopian Critique p, 494-496 Bill Joy, April 2000 Issue of Wired (magazine)

Against various doomsday scenarios, McCarthy responds that they are all too remote to concern us. AI will not sneak up on us; we have plenty of time. "We won't know enough to regulate until we see what it looks like," (M, 2003. Q in M, p. 499) he wrote. "Correct decsions will require an intense effort that connot be mobilized to consider an eventuality that is still remote." (M, 1976. Q in M p. 380)

In general
Richard Feynmann, thinking about the unexpected consequences of technology, wrote of a man he came across in buddhist temple who told him "to every man is given the key to the gates of heaven. The same key opens the gates of hell" Q in M p. 499

"Both AI's loudest public champiaons and its loudest public enemies raise expectations or fears that are, to put it generously, premature." p. 511

McCorduck, for example, suggests dozens of scenarios that have no been suggested. An evangelical prohibition, akin to George Bush's crackdown on stem-cell research, which leads to Indian and Chinese supremacy in AI. Or, consider the possibility that extremely powerful AI systems are built, but, for economic reasons, they are only used to solve specific problems for business or government. p. 516-517.

Hope of another spring
McCorduck reports "in the 1990s, shoots of green broke through the wintry AI soil." p. 418 ✅

Raj Reddy, 1988, (in his presidential address to the AAAI, said that when he gazed about he saw not the mythical AI winter but a spring with flowers blooming. p. 434 Find original quote ✅

AI Effect
Does not refer to this by name 204, 423 ✅

"It's part of the history of the field of artificla intelligence that every time somebody figured out how to make a computer do soething---play good checkers, solve simple but relatively informal problems--there was chorus of critics to say, "that's not thinking". p. 204

She calls it an "odd paradox." p. 423

"Practical AI successes, computational programs that actually achieved intelligent behavior, were soon assimilated into whatever application domain they wer found to be useful, and became silent partners alongside other problem-solving approaches, which left AI researchers to deal only with the "failures," the tough nuts that couldn't yet be cracked." McCorduck, p. 423

MACSYMA: originally an AI program. p. 423

When Deep Blue succeeded in defeating Barry Kasparov in 1987, people complained that it had only used "brute force methods" and it wasn't real intelligence. p. 433

AI Winter/The problems/Where is HAL 9000?
Pamela McCorduck wrote in 1979: "Making machine think, designing computer programs to behave intelligently, was far harder than anynoe in 1956 thought it would be. Over two decades have imbued the field with more modesty than it had in its infancy, but the fact remains that the problems continue to be harder than anyone expected." p. 118

McCorduck, talking about AI in the late 80s: "The goals once articulated with debonair intellectual verve by AI pioneers appeared unreachable, and their methods seemed, if not exhausted, not quite scalable either." p. 417

"The grand vision held by AI's founding fathers, a general machine intelligence, seemed to contract into negligible, probably impossible dream." M, p. 424

"If they want it at all, they want AI yesterday. Resultas have been so disappointing, they pout, and I must wonder, compared to what? The 2500 yeasr of research in physics? The hundresd of year of research into biology? Medicine is only edging toward science after millennia as an art? Cosmology?" p. 511-512

Anthropomorphy issue
"two major branches of artifical intelligence: one aimed at producing intelligent behavior regardless of how it was accomplioshed, and the other aimed at modeling intelligent processes found in nature, particularly human ones." M p. 100-101

Fear of Machines
Fear of machines: good quote, p. 151

Womb Envy
Lighthill p. 134-6, 244 (but she is mostly interested in his accusation that AI researchers, unlike women, are unable to bare children, so that's why they pursue AI)

"Sir James found this womb-envy idea dubious, but couldn't resist bringing it up. Since his report is largely negative it seems a bit malicious of him to bring it up at all." p. 135

Lighthill
p. 244 McCorduck says (in footnote) that funding didn't drop! Should be quoted in a footnote of AI winter.

Imitating brains
"If we wait til the physiologists get around to give us a theory of mind, we'll be waiting forever." Herbert Simon, Q in M p. 291. May have said this in the early 60s.

Raw computer power
McCorduck mentions the issue of raw computing power on p. 421.

Simon against Connectionism
"These 'connectionist' architectures have a role to play (for instance in simulating auditory and visual sensory processes) but ... they will not replace physical symbol systems as models of higher mental processes." Simon Q in M, p. 452

Cyclical nature of AI paradigms
"[A] recurring theme in AI [is that] ideas are picked up, exploited to the maximum extent allowed by available hardware and software, and then picked up again as major improvements in hardware and software (the latter often from AI research itself) allow a new round of exploitation." p. 422, where she specifically mentions the way neural nets are a resumption of the work of Pitte, McCullough and Rosenblatt, or that Rodney Brooks work was "really a return to the principles of cybernetics." p. 422 Newell writes "Everything waits until it's time. Science is the art of the possible." Q in M p. 422