Topic: limitations of artificial intelligence and cognitive science

topics > computer science > Group: artificial intelligence

relationship between brain and behavior

expert users
frame problem
language and life as a game
limitations of formalism
limitations of robots
reality is a machine
people better than computers
people vs. computers
philosophy of mind
problem of classifying information
problem of other minds
thought is computational


In 1950, Turing predicted intelligent machines in fifty years with a billion words of memory. Fifty-six years later, gigabyte computers are commonplace, but computers are not intelligent. From the beginning, computers computed faster and more accurately than people. Within restricted domains such as backgammon, Scrabble, and chess, they outperform almost all humans. Does this demonstrate that computers are intelligent?

Is intelligence more than computation? Is artificial intelligence nothing more than an artificial flower than fools the casual observer?

We do not understand intelligence, the brain, and mind. Philosphers have addressed these problems. Intelligence involves interpretation, language, emotion, knowledge, exploration, social context, and observers. An expert recognizes known patterns in new situations and adjusts the response accordingly. In language, walking, and daily life, we are all experts. (cbb 6/06)

Subtopic: is AI feasible up

Quote: review of the limits and capabilities of machines; is artificial intelligence feasible? [»whitI11_1988]
Quote: Rabin's proofs only disprove the feasibility of theorem-proving systems and combinatorially complex data correlations; already abandoned by most of AI [»carbJG3_1978]

Subtopic: AI as fake intelligence up

Quote: the thinness of the psychological reality of AI models may be masked by marvelous verisimilitude; e.g., SHRDLU answers appropriately even though the response is canned [»dennDC_1978a]
Quote: even though a program may appear graceful, it looks like a Rube Goldberg machine when executed by hand [»dennDC_1978a]
Quote: AI can only produce zombanimals; it can not produce people with free will, creativity, consciousness, and abstract reasoning [»brinS10_2000]
Quote: artificial intelligence is like intelligence just as artificial flowers are like flowers; from a distance they may look alike [»parnDL12_1985]
Quote: a limited imitation of intelligence will intrude with incomprehensible breakdowns [»winoT_1986]
Quote: a weather simulation is certainly not weather; why do we think that a computer simulation of the mind could think? [»searJR_1984]

Subtopic: artifial intelligence is not intelligence up

Quote: in 40 years, no tangible progress has been made towards realizing machine intelligence with digital computers; may be impossible [»wilkVW8_1992]
Quote: AI concentrates on discrete aspects of intelligence while AI critics concentrate on the continuous aspects [»sowaJF_1984]

Subtopic: behavior/description/neuron firings is not intelligence up

Quote: if same-behavior-ergo-same-mental-phenomena than radios would be conscious; the Turing test is similarly mistaken
Quote: is a book that describes Einstein's brain aware? If so, what if it is never opened up and consulted? [»penrR_1989]
Quote: for shared beliefs, what is the property shared by different neurophysiological squiggles?; problem with token identity theories [»searJR_1992]

Subtopic: AI is too small up

Quote: the problem with AI is the vast gulf between a pilot system and the size of a human cognitive system [»cherC5_1988]
Quote: most AI is done in the blocks world because researchers do recognition, spatial understanding, sensing, etc.; makes AI brittle [»brooRA1_1991]
Quote: the DENDRAL expert system had a quite limited impact on organic chemistry; it was not useful for large difficult problems [»lindRK6_1993]

Subtopic: computers are not intelligent up

Quote: computers have no more intelligence than a wooden pencil
Quote: a computer model of thought is just a model; no one expects to get wet in a pool filled with molecular models [»searJR1_1990]
Quote: a computer program can not give a system a mind [»searJR_1984]
Quote: after watching the natural intelligence of a 6-month old, I decided that making kids smarter was better than making computers smarter; I gave up AI in a flash [»druiA4_1997]

Subtopic: we do not understand intelligence, the brain, mind up

Quote: AI is oversold because nobody knows how to do learning, so nobody knows how to make smart machines [»schaR12_1984]
Quote: if a machine is to simulate a brain, the designer must understand the brain first [»taubM_1961]
Quote: we have no idea what a 'mind' is; even, what does it mean for a physical object to have a 'mind'?; confused [»kripSA_1982]
Quote: Wallace's Paradox is the unlikely evolution of human intelligence from apes [»brinS10_2000]

Subtopic: AI is out-of-date philosophy up

Quote: artificial intelligence models of language are generally equivalent to older philosophical models [»winoT_1986]
Quote: artificial intelligence research follows the rational decision making model [»winoT_1986]
Quote: the defenders of strong AI and functionalism believe that the materialist tradition and science affirms their position [»searJR_1992]
Quote: AI based on old ideas: Descartes: understanding by a symbolic representation; Kant: concepts are rules; Frege: rules can be formalized and manipulated without interpretation [»dreyHL1_1994]

Subtopic: interpretation and discrimination up

Quote: intelligence lies not in the generation of new ideas but in their interpretation and discrimination [»ashbWR_1956]

Subtopic: computation is observer-relative up

Quote: most computational theories of mind assume a homunculus; even if reduce to 0/1's still need a homunculus to interpret them [»searJR_1992]
Quote: cognitive science can never be a natural science since computation is observer-relative
Quote: psychology must posit internal representations, but something is a representation only for or to someone; i.e., a homunculus that leads to an infinite regress [»dennDC_1978a]
Quote: consciousness and intentionality are intrinsic and ineliminable, while computation is observer relative; contrary to the dominate position [»searJR_1992]
Quote: we think organization is the answer to intelligence from unintelligent components; but intelligence has third-person criteria that makes the question incoherent [»searJR_1992]

Subtopic: coordination up

Quote: unlike AI, perhaps nature is graceful all the way down

Subtopic: computers are not emotional up

Quote: an artifact can not be a genuine emotional brain [»thriMW_1983]

Subtopic: task vs. emotion up

Quote: the design of a user agent should be guided by its task and not its social dimension
Quote: Microsoft Office Assistant has a social dimension that encourages an adverse reaction [»schaH1_2001]

Subtopic: novice vs. expert up

Quote: a computer is at best a powerful beginner, competent in artificial microworlds, yet incompetent in the real world of human expertise
Quote: humans are expert performers for much of everyday life; inaccessible to computers [»hallH_1992]
Quote: Socrates demonstrated that experts could not articulate the principles and rules underlying their expertise [»dreyHL1_1994]
Quote: DENDRAL's successors generated possible solutions to a problem; they generated too many candidates for difficult problems, while simple problems could be done by hand
Quote: expert systems are brittle, i.e., they do well up to some limit, then they collapse into complete incompetence [»lenaDB1_1991]

Subtopic: exploration vs. problem solving up

Quote: problem-solving requires exploration instead of just search [»hewiC4_1985]
Quote: problem solving requires exploration: no initial state, pre-defined operations, or goal state
Quote: intelligence is the ability to figure out a procedure rather than the ability to carry one out [»kugeP9_1979]
Quote: the advanced beginner can easily learn situational elements by example; this is a severe limitation on computer intelligence [»hallH_1992]

Subtopic: language depends on context up

Quote: stock market displays allow dealers to refer to stock in terms of its current price [»jiroM12_1993]
Quote: the information for a deal follows no set pattern, both parties supply information; difficult for voice recognition [»jiroM12_1993]
Quote: people distinguish classes via short descriptions; using language specific to those classes; brevity is achieved without lose of information [»bongM_1967]
Quote: the antinomies of formal logic arise because classifications must be modified for unforeseen objects [»poinH_1908, OK]

Subtopic: knowledge vs. reason up

Quote: the knowledge principle: the power of AI depends on the program's knowledge of its task domain instead of the program's reasoning processes [»feigEA5_1996]
Quote: early AI work did not appreciate the sheer amount and variety of specialized knowledge underlying intelligent behavior [»lindRK6_1993]
Quote: general AI methods failed when the specialized knowledge needed for real-world problems swamped their heuristic methods and data encodings

Subtopic: skepticism about knowledge representation up

Quote: cognitive science and universal grammars treat the existence of patterns as evidence for mental representations and unconscious rules; unfounded [»searJR_1992]
Quote: cognitive science rests on the assumption that cognitive terms can be related independently of other descriptive systems [»pylyZW_1986]
Quote: skeptical about knowledge representation even though it is fundamental to cognitive science, linguistics and artificial intelligence [»winoT_1986]
Quote: AI researchers write their programs for other AI researchers; incredibly complex data structures [»coddEF_1990]

Subtopic: user agent up

Quote: Microsoft Office Assistant poorly rated for troubleshooting and for learning new applications [»schaH1_2001]
Quote: anthropomorphic agents useful as a motivator or coach while learning; also entertainment [»schaH1_2001]

Subtopic: elaborate behavior from limited capabilities up

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

Subtopic: AI better than experts up

Quote: TD-Gammon is a self-training neural network for backgammon that outperforms other programs and, sometimes, human experts [»tesaG3_1995]
Quote: TD-Gammon is successful because of the randomness of backgammon and a fairly smooth outcome function [»tesaG3_1995]
Quote: even with a random initial network, TD-Gammon would terminate in, at most, several thousand moves

Related Topics up

Group: psychology   (9 topics, 307 quotes)
Group: relationship between brain and behavior   (9 topics, 332 quotes)

Topic: consciousness (58 items)
Topic: expert users (25 items)
Topic: frame problem (13 items)
Topic: language and life as a game (30 items)
Topic: limitations of formalism (93 items)
Topic: limitations of robots (6 items)
Topic: reality is a machine (48 items)
Topic: people better than computers (35 items)
Topic: people vs. computers (55 items)
Topic: philosophy of mind (78 items)
Topic: problem of classifying information (42 items)
Topic: problem of other minds (11 items)
Topic: thought is computational
(60 items)

Updated barberCB 3/06
Copyright © 2002-2008 by C. Bradford Barber. All rights reserved.
Thesa is a trademark of C. Bradford Barber.