Map
Index
Random
Help
Group
Quotes
Related
Subtopics
th

Group: artificial intelligence

topic root > Group: computer science


--related--
computer science
coordination system
relationship between brain and behavior
-
actor machines
artificial neuron nets
evolutionary systems
limitations of robots
logic programming
motion planning for robots
networks of relays
problem solving
psychology
self-regulating systems
software models of reality
theorem proving systems

--referenced--
all groups
map of the Thesa web site
topics a-e

Summary

Artificial intelligence (AI) is the construction of intelligent, computer-based systems. An AI system may play a game, solve problems, or act as a pet. AI systems tend to use heuristics, a search strategy, or lookup tables to avoid an exhaustive search of the possible solutions.

Chess is the classic game for AI. Attempts at Go are less successful. (cbb 6/06)

Group members up

Topic: computer as an intelligent agent
Topic: people vs. computers
Topic: expert systems
Topic: frame problem
Topic: heuristic-based systems
Topic: intelligent machines
Topic: knowledge representation
Topic: knowledge representation by frames
Topic: limitations of artificial intelligence and cognitive science
Topic: reality is a machine
Topic: people better than computers
Topic: production systems
Topic: thought is computational
Topic: Turing test
Subtopic: emergent behavior up

Quote: the real mystique of computers is their elaborate behavior from limited capabilities [kentW_1978]

Subtopic: game-theoretic solutions up

Quote: a game is ultra-weakly solved if it has a game-theoretic value for initial positions; weakly solved if there is a strategy for both players; strongly solved if there is a strategy for all legal positions [bowlM11_17]
Quote: heads-up limit Texa hold-em poker is essentially weakly solved and the dealer wins; first nontrivial imperfect game played competitively; variation of Counterfactual Regret Minimization of Nash equilibrium [bowlM11_17]

Subtopic: AI as problem solving up

Quote: one form of artificial intelligence consists of solving problems that have required human intelligence [parnDL12_1985]
Quote: we need AI because intelligence is used to reduce intractable problems to manageable size; how do we do this? [strzT5_1988]

Subtopic: AI as imitating intelligence up

Quote: another form of artificial intelligence uses heuristic or rule-base programming to solve a problem like a human seems to solves it [parnDL12_1985]

Subtopic: AI as assistant up

Quote: Vellum has a Drafting Assistant to suggest relationships and commands such as 'on', 'intersection', 'tangent'; a use of "animated language" [tognB_1992]

Subtopic: AI as symbol manipulation up

Quote: AI assumes that one can explain the high-level traffic of symbol activations in its own terms, without neural events [hofsDR_1979]
Quote: cognitive science rests on the assumption that cognitive terms can be related independently of other descriptive systems [pylyZW_1986]

Subtopic: AI as string transformation up

Quote: use GAWK for AI because any logic merely defines how strings can be transformed into other strings; GAWK is designed for string transformation [louiRP8_1996]

Subtopic: AI as novice up

Quote: the mental processes of a novice are easily imitated by a computer; outperforms novice because of more rules and context-free elements [hallH_1992]

Subtopic: AI as brute force up

Quote: Scrabble is very fast but not smart; still achieves resounding victories almost all human players [appeAW5_1988]

Subtopic: AI as physical up

Quote: to avoid delusion, should build complete, intelligent systems with real sensing and real action; enhance incrementally [brooRA1_1991]
Quote: physical grounding hypothesis: an intelligent system must be grounded in the physical world with appropriate sensors and actuators [brooRA6_1990]
Quote: artificial knee with automatic knee locking reflex on contact with ground [bekeGA4_1986]

Subtopic: random search up

Quote: software must have internal regularies which can be found by random search [menzT1_2007]
Quote: random search of scheduling solutions more effective than complete search; master or collar variables determine the behavior [menzT1_2007]
Quote: solve large N-queens problems using random search; up to board size 26 [menzT1_2007]

Subtopic: collar variables up

Quote: collar variables are variables that determine the behavior of the rest of the system [menzT1_2007]

Subtopic: machine learning up

Quote: quickly and accurately predict human pose from a single depth image; convert image to body parts; machine learning from test images [shotJ1_2013]
Quote: combine many models for a machine learner; e.g., Netflix winners used over a 100 learners and their combination further improved the results [domiP10_2012]
Quote: all machine learners group nearby examples into the same class; so lots of data is better than a clever learner [domiP10_2012]
Quote: machine learning systems automatically learn programs from data; learning = representation + evaluation + optimization; search among the classifiers for the highest-scoring [domiP10_2012]
Quote: machine learning must generalize beyond the examples of the training set; doing well at training time is easy, do not contaminate your classifier [domiP10_2012]
Quote: overfitting can be subtle; generalization error includes consistent bias and random variance [domiP10_2012]
Quote: a more powerful machine learner is not necessarily better than a less powerful one; naive Bayes may produce better results than a rule learner [domiP10_2012]

Subtopic: learning up

Quote: for autonomous helicopters, learn accurate, trajectory-specific local models from expert demonstrations; can outperform the expert pilot in difficult maneuvers [coatA7_2009]
Quote: the goal of temporal difference methods is to match the learner's current prediction for a pattern with the next prediction at the next time step [tesaG3_1995]
Quote: a program could learn if a higher-level program changed its evaluation functions according to the results of games played [shanCE3_1950]
Quote: if could teach computers, they could digest the world's libraries and achieve superhuman effectiveness; basic goal of AI [schwJ_1987]
Quote: a research program for AI is to hand-code a broad knowledge base, then acquire knowledge through reading, and finally learn by discovery [lenaDB1_1991]
Quote: a program could randomly vary its evaluation function and select those variations that perform best [shanCE3_1950]
Quote: produce a thinking machine by imitating each part and letting it roam the countryside; too slow and too large to be practical [turiAM9_1947]
Quote: Turing created a universal machine by training, on paper, an unorganized machine with rewards and punishments; circular memory of 64 squares [turiAM9_1947]

Subtopic: chess up

Quote: a paper machine for playing chess acts as if it were alive; it is difficult to distinguish from a rather poor chess player [turiAM9_1947]
Quote: want to represent chess as numbers and operations on numbers, and to reduce strategy to a sequence of computer orders [shanCE3_1950]
Quote: computers have applications in matrix operations, differential equations, logical problems including chess and draughts, commercial and industrial subjects, fault diagnosis, programming, and pure mathematics [ferr8_1952]
Quote: play chess by maximizing and minimizing the value of the current position; computed by adding terms for such properties as doubled pawns [shanCE3_1950]
Quote: a chess program needs functions to determine stable positions (no pieces en prise) and potentially useful moves [shanCE3_1950]
Quote: a chess program should play a fairly strong game, at speeds comparable to humans [shanCE3_1950]
Quote: can use statistical variations during the chess opening and can play by the book, like masters [shanCE3_1950]

Subtopic: Go up

Quote: use Monte-Carlo tree search for computerized Go games; select between simulated policies instead of evaluated board positions [gellS3_2012]

Subtopic: remote control up

Quote: capture the flag by security attacks of 7 autonomous machines against each other; a supercode attack exploits others attack code to further ones own goals [hurlG2_2018]
Quote: formal analysis of a plan execution module for NASA's Remote Agent; first complete, AI control of a space-craft [haveK8_2001]

Subtopic: AI not trustworthy up

Quote: AI systems can go awry in unexpected ways; who knows what the machine knew and when it knew it; problem of trust of largely untrusted technology [hurlG2_2018]

Related up

Group: coordination system
Group: relationship between brain and behavior
-
Topic: actor machines
Topic: artificial neuron nets
Topic: evolutionary systems
Topic: limitations of robots
Topic: logic programming
Topic: motion planning for robots
Topic: networks of relays
Topic: problem solving
Group: psychology
Topic: self-regulating systems
Topic: software models of reality
Topic: theorem proving systems

Subtopics up

AI as assistant
AI as brute force
AI as imitating intelligence
AI as novice
AI as physical
AI as problem solving
AI as string transformation
AI as symbol manipulation
AI not trustworthy
chess
collar variables
emergent behavior
game-theoretic solutions
Go
learning
machine learning
random search
remote control

Updated barberCB 2/06
Copyright © 2002-2023 by C.B. Barber
Thesa, Avev, and thid-... are trademarks of C.B. Barber