jim.shamlin.com

Chapter 13: Human and Artificial Intelligence

Psychology is still struggling to define the term "intelligence," and the concept is becoming even hazier as our consideration of what might constitute intelligence is expanding.

Historical definitions of intelligence have generally focused on two factors:

  1. The capacity to learn from experience
  2. The ability to react and adapt to changes in context

Contemporary experts have gone little further. A 1986 survey of intelligence researchers, they embraced the two principles of the historical definition, and added focus on metacognition (understanding and control of one's own thinking processes), and in the present day there is a great deal of focus on cultural definitions, with the notion that "intelligence" means different things in different cultures (Spellman 2000). Given the lack of consensus we are largely stuck with the original definition of the term: learning and context.

It may be more productive to consider the topic from the layman's perspective, from which there are a number of meanings. A person is regarded as intelligent because:

The difference in cultural perspectives on intelligence relates to the manner in which those skills are applied: some cultures value the ability to memorize information, others to understand relationships, others value the ability to avoid difficulties, others the ability to achieve results.

For example, the present trend in the United States is to value the application of knowledge to relationships. "Emotional Intelligence" is the pertains to domain knowledge of emotions, the ability to interpret and influence the emotional state of others in order to achieve positive outcomes in social situations.

There is general disdain for measures of intelligence, generally among those who propose a different metric than is being used. Most people seek a definition of intelligence that proves their own, and in academic circles the definition shifts to favor the politics of the day by emphasizing whatever aspects of personality are supportive of an agenda.

Measures and Structures of Intelligence

Early measures of intelligence considered psychophysical capabilities. The mind was useful as a control center for the body, hence intelligence was expressed in sensory acuity and motor coordination.

Francis Galton (1822-1911) conducted a myriad of experiments of psychophysical capacity, which considered whether a person was able to detect minute differences in the weight of objects, the volume or pitch of sounds, or shades of color. While Galton's tests demonstrated differing levels of ability among individuals in interpreting sensory information, there was no statistical correlation between test scores and other measures of intelligence, such as the grades of student subjects.

Benet and Simon (c 1911) tested intelligence as a means of differentiating people of high intelligence, average intelligence, and low intelligence (to the point of mental retardation) by the facility with which tasks were performed. The outcome of such a task involves not only the ability to control motor function, but also the ability to determine what needs to be done to achieve a desired outcome. This was the original of the notion of "mental age" - in that they measured the facility by which individuals of certain ages performed a task and used the averages as a measure of intelligence.

William Stern (1912) further refined this metric into the Intelligence Quotient, by dividing the mental age as determined by a test with the chronological age of a subject, expressed as a percentage. That is to say that a ten-year-old with the mental capacity of a seven-year-old was said to have an IQ of 70, whereas another with the mental capacity of a twelve-year-old was said to have an IQ of 120.

Stern's quotient proved inadequate largely due to the differences in the rate of development. Given that intelligence develops rapidly during youth and slows during the teenage years and adulthood, the mental capacity at age 8 is significantly higher than that at age 6, whereas the difference between a person aged 28 is not better to the same degree than one aged 26. And especially for adults, a comparison to individuals of advanced age fails: it is not easy to measure the intelligence of a 90-year-old, not meaningful to be compared to one.

Wechsler further refined the measure of intelligence by separating the population into age groups, developing the Wechsler Adult Intelligence Scale (WAIS), the Wechsler Intelligence Scale for Children (WISC), and Wechsler Preschool and Primary Scale of Intelligence (WPPSI), each of which used tests of verbal and problem-solving abilities of subjects.

Current tests of intelligence use a variety of questions that test various capacities:

The nature of any test depends on the capacities that the designer of the test considers to be significant in indicating intelligence, so tests vary greatly. There is also an approach that considers the process rather than the outcome - what a subject attempts in his effort to discover the solution rather than the accuracy of the solution.

To this day, no psychologist seems to have gotten it quite right, and there have been multiple attempts to identify the factor or combination of factors that accurately measure intelligence, such that the scores on a test can be correlated to some meaningful evidence of a person's intelligence.

The author lists a few individuals whose theories have been generally accepted as having some validity in the present day:

Information Processing and Intelligence

Information-processing theorists are primarily interested in the way in which information is mentally manipulated once it has been received, particularly concerning the speed an accuracy of information processing, which is an important factor in the consideration of intelligence.

Process Timing Theories

The "inspection time" is the amount of time required to consider information and make a decision. It is somewhat polluted by the process of reporting a decision - to say a word or to click a button requires a separate process to actuate the facilities of speech and motion.

For example, one experiment presents a computer screen displaying two lines, requiring the subject to press either of two buttons to indicate which is longer. The intelligence of the subject is assessed as two factors: the accuracy is measured as a percentage of correct answers and the reaction time is the number of seconds it takes to complete a sequence.

Some investigators suggest that the reaction time reflects the speed at which information travels among the neurons of the brain, whereas others suggest that it measures the efficiency of the path through which the information travels.

Whatever the case it is generally found that subjects with higher intelligence (determined by a standard IQ test) generally perform faster than subjects of lower intelligence. But because there are multiple processes involved and the time differences are in milliseconds, it cannot be firmly concluded that processing time alone causes the variance: individuals of higher intelligence may have more rapid processes of perception or motor reaction as well as, or instead of, more rapid processing of the information.

In experiments that require a verbal rather than a motor response (to say "left" or "right" to indicate which line is longer), a similar complication arises in that the lexical-access speed may also be a factor in the difference between subjects.

Other experiments (Hunt and Lansman) measure the difference in reaction speeds among subjects who are asked to perform two different tasks at the same time (alternating attention from one to another), and it is generally found that more intelligent individuals have better accuracy and speed in these experiments, though the same complications apply.

Finally, the author refers to a number of researchers who have examined the link between information processing speed and learning, to determine if individuals who show faster speeds learn more or less from their experiences. Thus far there has no been a consistent correlation between the two, suggesting that computation and retention are two separate processes, but further research is necessary in this regard.

Working Memory

Recent work (Kyllonen and Christal) suggest that the capacity of working memory, rather than its speed, may be influential in intelligence. There is a high level of correlation between a subject's intelligence as demonstrated by test scores and the quantity of information they retain in working memory, as demonstrated by their ability to remember sequences of numbers of items, repeat phrases verbatim, recount equations of various length, or retain multiple facts from reading a passage of text.

The author speculates that working memory is "probably not all there is to intelligence" but acknowledges that the more information a person can manage to hold in working memory, the better able they are to recognize complex relationships among multiple factors and derive a solution based on more input rather than faster processing.

Componential Theory and Complex Problem Solving

The term "components" pertains to the mental process that are brought to bear in performing a given task: a person must perceive, remember, analyze, and perform various other component tasks in order to be successful at a larger task.

Componential analysis therefore seeks to dissect a demonstration of intelligence into its component parts, considering the efficiency and effectiveness in terms of each component as well as the manner in which an individual assembles components to effect an outcome (which in itself lends efficiency to mental processes, as using fewer components to accomplish a goal results in greater alacrity and facility).

As an example, a person who is attempting to respond to an analogy (A is to B as C is to which of four options?) must recognize the definition of seven terms. Understanding a single term means being able to recognize the word, match it to memory or parse it against lexical analysis of word parts, then analyze the relationship between A and B, then consider the relationship between C and each of the four options for the answer, then evaluate which conclusion is best suited. When broken into its components, an analogy requires a very complex feat of cognition - and if an analyst fails to dissect the task, his results fail to consider exactly which mental processes it is evaluating.

Biological Bases of Intelligence

We presently recognize that the human brain is the organ in which intelligence resides. While everyday expressions still preserve metaphors about emotions and the heart and dexterity and the hands, we no longer maintain the belief that these organs are the center in which intelligence exists independent of the brain.

Early studies of the brain were "resounding failures" in their inability to observe the organ in action, and theorizing by inference about its function, giving rise to myth and superstition that have obfuscated rather than facilitated scientific progress. The size of the organ, its shape and coloration, and the like have nothing to do with intelligence or the function of the organ.

Aside of the consequences of injury, the physical properties of the brain itself have not been successfully correlated to functions to the degree that knowledge or intelligence can be assessed by physical examination. More recent studies "offer some appealing possibilities."

Studies of electrochemical activity in the brain have shown some correlation to intelligence. Several studies have suggested that the speed at which impulses flow is correlated, but other studies contradict this notion. Others still show greater correlation in men than in women.

Other studies consider the consumption of glucose in the brain, and indicate that more intelligent subjects consume less glucose for the mental activity involved in solving a problem - suggesting that their minds are not faster, but merely more efficient than less intelligent individuals.

Another set of studies considers the flexibility of neural circuitry. That is , in addition to consuming less glucose, more intelligent individuals show a more distinct pattern of glucose consumption in areas of the brain, which underscores the theory of efficiency: when a person struggles with a task, he must literally put more of his brain to work on the problem. Another interpretation is that intelligent subjects are more mentally focused. As such, glucose consumption is likely an effect rather than a cause of intelligence, but suggests that the efficiency of glucose-consuming mechanisms is correlated. But again, there are studies to the contrary.

There is also the suggestion that traditional measurements of intelligence fail to consider the process by which solutions are derived. That is, they consider only if the subject got the answer right, but fail to consider the method by which the individual chose to solve it. In measuring physical tasks, it is generally reckoned that a person who expends less time and effort in solving a problem has demonstrated greater proficiency than another individual who arrives at the same solution through greater effort.

Neither do these tests give much consideration of an individual's propensity to learn through experience - though this may be implicit in asking a series of questions of the same type (the subject becomes better at solving analogies after having solved ten of them), the lack of feedback during the test (the individual does not learn, after each response, whether he was correct or incorrect) and the failure to consider process (tests are timed, but score is not augmented for finishing faster) contradict the suggestion that it should be so.

Ultimately, intelligence is difficult to assess because we cannot separate the mind from the body, nor define intelligence except in the concept of the performance of a specific task, nor observe the mental processes involved in problem solving. As such our basic understanding of the phenomenon remains vague and uncertain.

Alternative Approaches

The theory of contextualizm maintains that intelligence can only be understood in a real-life context. That is, a person's intelligence is demonstrated in the efficiency and effectiveness of their actions, and their actions are performed to achieve outcomes in a given environment.

It is not uncommon to speak of an individual demonstrating great intelligence in some situations and painfully little intelligence in others - e.g., the fellow who's a brilliant accountant but hopeless at a cocktail party. Whether this person is generally regarded as intelligent depends on the context in which is was observed.

(EN: This implies, but stops short of stating, that intelligence assessment is also skewed by environment, suggesting that a person who rates as intelligent "in the lab" or "in a test-taking situation" may be incompetent in real-life situations. I'm a bit skeptical, as it's often rendered as an excuse for those who perform poorly on assessments yet still wish to consider themselves to be brilliant. Moreover, it seems along the lines of suggesting a person can be considered strong when lifting weights but is weak when moving furniture. My sense is that intelligence, like physical strength, is a capacity that does not change, though a person's inclination to apply it varies by situation.)

Cultural Context

In the present day, the fashion is to perceive intelligence in the context of culture. It is certainly defined in this context, as the qualities that are valued in a given culture are those that are measured and assessed, such that a person who applies their intelligence to certain activities is regarded as valuable by one culture and foolish or counterproductive in another.

Culture itself is a vague consensus upon the "attitudes, values, beliefs, and behaviors shared by a group of people." This notion may be applied in various ways: it may be used analytically in describing the behavior of a group in the present or in the past, or it may be used prescriptively to suggest how a person ought to think and act to be accepted into a culture, or even to suggest that the values of an entre group of people must be changed.

Cultural intelligence is considered to be important to the understanding of human evolution: cultures that value and encourage certain qualities achieve greater outcomes than those that value different qualities. Or more directly, a culture that makes intelligent choices encourage productive behavior and thrive whereas a culture that chooses unwisely encourages unproductive or counterproductive behavior and ultimately fail - and cease to exist.

People in different cultures may have very different ideas about what it means to be intelligent. A few examples are given:

Perhaps the only universal criterion for intelligence is that there is a sense of purpose in undertaking any action. Things that are done randomly and for no particular reason are universally regarded as being unintelligent - but whether things that are done for a reason are regarded as intelligent depends on whether the assessor understand the reason and deems it to be worthwhile.

One study (Sarason 1979) drives home the impact of cultural values on the assessment of intelligence. Tests of IQ were administered to recent immigrants to the US over a period of time. The initial test conducted when they first arrived demonstrated significantly lower scores than later tests, after the same individuals had become better acculturated to the American way of life.

Clearly, these individuals did not get smarter, but merely learned the attitudes and behaviors valued by American culture. It is therefore reasonable to conclude that measures of intelligence are biased to favor the qualities that are favored by a culture. It has been further supposed that the low results of minority groups within a nation are also likely to reflect lower intelligence because of a difference in values rather than a difference in mental capacity.

Naturally, this has led to some political uproar and the demand for a method of assessing intelligence that is not biased to the qualities valued by a specific culture. But because the very nature of intelligence is in the context of culture, there has been little success at defining a method of study that would not reflect cultural bias.

Not only is intelligence specific to a given culture, but there is also evidence that it is related to experience. An accountant will fare better at mathematical assessments and a carpet merchant would fare better at recognizing shapes and patterns because their routine activities require and help to develop specific mental skills.

There are also linguistic barriers - most obvious in that a person who is taking a test in a language they do not understand very well will fare poorly because of their inability to understand the questions, and as such will give the wrong responses to questions they could easily have answered in their native language. The language itself may be a barrier, as a person whose native language does not include terms for certain objects or relationships will have a poor understanding of them, even if he learns the terms and concepts in a different language.

Finally, there is also the matter of esteem. A person identifies himself as a member of a group or profession, and in doing so comes to place greater emphasis on certain things. One experiment (Ceci 1985) found that the accuracy of responses to test questions reflected gender identity: if a mathematical word-problem is put in the context of a sporting activity, boys respond correctly much more often than if the very same problem is framed in the context of baking cupcakes. It is reckoned because the incidental details of the problem as stated are regarded as gender-inappropriate, such that the subject's disdain for the context causes them to refrain from applying their intelligence to solving the problem.

Gardner: Multiple Intelligences

Howard Gardner (1983) proposed a theory of multiple intelligences, in which intelligence is not considered to be a single quality, but eight:

  1. Linguistic Intelligence - The ability to understand and utilize spoken and written language
  2. Logical Intelligence - The ability to solve logical reasoning and mathematical problems
  3. Spatial Intelligence - The ability to travel through an environment or arrange things within an environment
  4. Musical Intelligence - The ability to understand and create sound
  5. Kinesthetic Intelligence - The ability to control one's own body to demonstrate dexterity and grace
  6. Interpersonal Intelligence - The ability to understand the motivations and behavior of other people
  7. Intrapersonal Intelligence - The ability to understand one's own motivations and behavior
  8. Naturalist Intelligence - Understanding patterns in nature

At a glance, Gardner's theory seems to be factorial, similar to the way that a test of general intelligence poses different kinds of questions - but his view was not that these are eight dimensions of a single capacity but eight different things to be considered independently and which do not total up to anything.

There are a broad array of areas in which "intelligence" might be measured, and numerous pseudoscientists and self-help authors have been very industrious in defining various forms of "intelligence" to sell books, seminars, and consulting services (emotional intelligence, organizational intelligence, marketing intelligence, dog-walking intelligence, etc.), for to Gardner there are eight qualities that distinguishes an intelligence from a subject domain:

  1. Biology - An intelligence must be isolated to a discrete area of the brain, such that physical damage impairs or destroys behavior derived from that intelligence
  2. Extremes - There must me models who demonstrate extraordinary ability or deficiency in a kind of intelligent behavior
  3. Core Operations - An intelligence consists of component abilities (e.g., musical intelligence is expressed in the ability to detect the pitch or "note" of a sound)
  4. Linear Development - The development from novice to master must be linear, in that proficiency is not instantaneous but incremental
  5. Evolution - An intelligence must be linked to adaptations to environmental change that necessitated traits to be developed
  6. Cognitive - An intelligence muse be linked to specific actions that are learned and consciously practiced, though they may become seemingly effortless given time and experience.
  7. Measurable - An intelligence must lend itself to being gauged and assessed by psychometric evaluation
  8. Encodable - An intelligence must be either expressed by a system of symbols or contextualized within a "culturally devised area"

Gardner's view of the human mind is modular, a model that is not uncommon but not universally espoused: a given intelligence is a form of module that is composed of components and a component may be comprised of other components. Also, a given component may be used in multiple modules.

The author concedes that "evidence for the existence of these separate intelligences has yet to be produced" and Gardner's work has fallen out of fashion with the professional community, chiefly due to its popularity with quacks and pseudoscientists. (EN: in researching more about Gardner's work, one of the chief criticism is that it has been used in pandering to narcissism - toying with the definition of intelligence so that everyone can feel that they are intelligent, or even superior to others, in their proficiency at something, however trivial.)

Sternberg: Triarchic Theory

(EN: The author of the book is Robert Sternberg, so he's presenting his own ideas here with aspirations of eponymism. So be aware.)

A separate approach to intelligence agrees that there are indeed various aspects of intelligence, but that they work together in an interactive and mutually supportive method. He calls his theory triarchic because it considers three "points" by which different expressions of intelligence interact:

  1. Intelligence Relates to the Internal World. Said another way, intelligence resides within the mind of an individual, and requires cognitive processes that interpret the identity and significance of concepts. It develops mental representations of ideas and models outcomes.
  2. Intelligence Relates to Experience. Intelligence is built upon information gathered from experience, as opposed to being intuitive or instinctive. While applying mental models and thinking through a solution are possible, even these models must be derived and transported from actual experience.
  3. Intelligence Relates to the External World. Intelligence is expressed in a subject's actions in the external world. If a process of thought leads to no results-oriented action in the external world, then it is not considered as intelligence, merely "thinking."

In terms of application, the author's model does not consider specific domains or activities to represent areas of intelligence, as does Gardner. That is, there is no "interpersonal intelligence" but instead various mental capacities that enable a person to gracefully interact with others based on other abilities: to use linguistic skills to listen and speak, to interpret and express ourselves nonverbally, to predict outcomes and identify motivation, etc.

A person who is intelligent has an awareness of their capacities and abilities in considering and undertaking actions that produce beneficial results. In practical terms, we find ways to make the best of our strengths, to make due in spite of weaknesses, and improve upon both.

The author refers to his own work in coaching students to develop the component abilities that contribute to academic success as a means of improving their intelligence (or more aptly, improving the way in which they apply their existing intelligence). This consisted of examining the skills of those who previously excelled to identify the component skills.

(EN: My sense of this is that the author's model doesn't work very well as a model of intelligence, but could be a highly useful methodology for a teacher or coach to plan a course of skills-improvement instruction and exercises.)

Effective, Ineffective, and Questionable Strategies

Human intelligence is highly malleable, as it is shaped and evolves from experience - and while it can be said that there is a genetic basis of intelligence, it is more obviously influenced by experience in application.

In essence, the learning process is trial-and-error. A person attempts to accomplish a goal by implementing a given strategy for its achievement, and then witnesses after taking action (or sometimes during the course of taking action) that a given strategy was effective or ineffective. So in essence, learning is a process of discovering outcomes and preferring effective strategies to ineffective ones.

The same process applies to thinking as it does to action. There are strategies for assessing the data received from sensory inputs, strategies for the way in which we interpret and predict, and so forth. So when we evaluate an action, we do not merely consider the actions we take, but the mental process by which we arrived at the decision to take those actions. This is how intelligence is developed.

The author speaks of the "head start" program, which consisted not merely of getting children into the educational system at a younger age, but in teaching them effective strategies for learning, particularly in the performance of certain foundational skills such as reading and mathematics. Program participants demonstrate a dramatic and statistically significant improvement in their academic performance not as a result of information gained, but strategies learned.

Another approach to helping intelligence along is to provide a stimulating environment in the home. This is reckoned to be a significant contributor to the difference in academic performance of children of wealthy families and those of poor families, the latter of which are presumed to provide children with a more meager an functional environment.

(EN: I don't think that's necessarily true. Poor people do not necessarily have fewer "things" but merely cheaper ones and a wealthy home is often quite uncluttered and austere. And it's very often remarked by successful people who rise from poor families that having fewer things in the home environment required them to be creative in seeking ways to entertain themselves. So while I don't entirely dismiss the author's argument, I don't have the sense that it is sufficient cause of intelligence.)

The author then lists a few other correlations to higher intelligence in children:

The author cautions that these studies are correlational, and should not be taken as causational. A further caution is that the children examined are considered to be too young, as it is believed that a child's IQ is not indicative as their intelligence as an adult prior to a certain age and that the correlation is less evident in children over the age of six.

While these data suggest that the correlation between wealth and correlation may be politically motivated, there is nonetheless a statistical correlation across societies in disparate cultures. It is likely wealth is not the cause of intelligence, but correlates with the factors that do.

Genetics has also been largely discounted as a cause of intelligence there is some correlation to heredity: while intelligence is may not be definitively proven to be passed from parent to child in the genes, a parent provides for their child an environment and rearing similar to their own upbringing, replicating the factors that contributed to their own development.

There is historical record of intelligence increasing over time in developed populations, which can also be attributed to the increased household wealth and improvement in educational systems an methods - though it is also noted that a "developing" economy corresponds to cultural changes. Per the earlier point on culture, people fare better on intelligence tests designed for developed nations the improvement in results may be a factor of changes in culture that bring a developing nature closer to the culture of the developed nations, by whose standards the criteria for intelligence are set.

The author considers that the factors of intelligence seem to be vaguely defined, but the cognitive skills that contribute to intelligence are universal: the ability to observe, to learn, to reason, and to solve problems are common to various perspectives of intelligence, regardless of the environment or the specific outcomes to which these skills are applied.

Development of Intelligence in Adults

Studies of intelligence indicate that intelligence increases rapidly during childhood years, levels off in adolescence, and then declines in adulthood.

Several theories draw a distinction between "crystallized" intelligence that remains fixed throughout adulthood (until the onset of old age) and "fluid" intelligence that can increase or decrease over shorter periods of time (years or decades). (EN: The description here is very scant, and evidence of this is not presented. I searched out other sources, and found them to be not much better in describing and substantiating the theory. Moreover, what I saw was largely related to memory and skills that are preserved for different durations, so it may be a misinterpretation or misapplication of the concept to apply it to intelligence.)

On a more granular level, it is maintained that memory skills vary with age: with short-term memory skills better in younger subjects and long-term ones better in older subjects.

Some researchers question the assertion that intelligence declines with age independent of a specific identifiable cause. It is clear that some conditions that affect the physiology of the brain develop at an advanced age and progress over time (e.g. Alzheimer's and Parkinson's Diseases), and the methods of studies of intelligence fail to separate individuals with these conditions from the general population.

Another counterpoint to the assertion of the decline of intelligence is the cultural aspect of intelligence. Age is a factor of culture, as much as nationality, and a 25-year-old and a 50-year-old have different values and different experiences even if they reside in the same geographic area. So it follows that using the instrument and analytical methodology that form the basis for assessing intelligence at age 25 is as awkward to a 50-year-old is as prone to bias as using an assessment designed for Chinese villagers on American urbanites.

In spite of the controversy and counterarguments. there does seem to be some general agreement on the basic pattern that intelligence, like physical strength, increases rapidly during the early developmental years, levels off during maturity, and declines during old age. In general, the fluctuation of intelligence during adulthood reflects the environment and activities of the individual. That is, those involved in cognitively demanding activities develop and maintain a higher level of mental "strength" than do those whose daily routine presents little opportunity for mental "exercise."

There is also some argument to be made about mental capacities. A person who is knowledgeable and experienced must consider any new information in the context of a much larger amount of information that is already in memory, which results in a slower reaction time to any new stimulus - which can be misinterpreted as a slower processing time. That is to say, a person who is knowledgeable may be processing at a very fast rate but considering a much larger body of information before responding, whereas a less experienced person has little in store to think about and can respond more quickly - and as a result the alacrity of a poorly considered response may be mistaken for a sharper intellect.

Based on various research, the author suggests three basic principles for the development of intelligence during adulthood:

  1. Certain abilities decline or cease to increase as rapidly during adulthood, but the increase in other abilities is counteractive.
  2. The environment and activities of an adult are highly influential in the development or atrophy of their mental abilities.
  3. The increased amount of knowledge and expertise that adults increases the amount of information they must process, which negatively affects response speed by improves effectiveness.

It is particularly noted that when success requires on accuracy rather than on speed, adults outperform youth for their ability to do careful work based on experience - much in the way that a novice can bang out amateurish work at a fast pace and a master takes a much longer amount of time to produce.

There is also the concept of wisdom, separate from intelligence, which increases with age. A wise person realizes that there is more to a problem than is initially presented, and thinks through scenarios that test variables that are not expressed in the problem statement. Whether this additional effort is wasted or worthwhile depends largely on whether the additional information the subject considers leads him to a more successful outcome.

The concept of wisdom may be culturally derived, as a "successful" outcome depends on the values of assessment. Other factors that influence the perception of wisdom include reasoning ability, shrewdness, breadth of knowledge, judgment, use of information, and keen insight.

This concept of wisdom is not a separate ability from intelligence, but accounts instead for certain practices that the present definition of intelligence fails to account - in essence, the difference is academic whereas the practice is the same: the quality of being "smart" subsumes both intelligence and wisdom, and academics are focused on the former as if the two were separable qualities.

Artificial Intelligence

In the present age of fascination with computer technology, there is an inordinate amount of attention being paid to the phenomenon of "artificial intelligence" - and the misinformed belief that computers are capable of thinking.

(EN: putting it in the context of the present age is interesting, especially when you consider how foolish we presently regard the utterly bone-headed ruminations of seemingly intelligent individuals in the era where religion held sway and much effort was put into describing the way in which supernatural forces were capable of influencing events. It seems likely that the present day fascination with electronics will seem just as silly in the future when the focus has moved to another topic.)

By its nature, any computer program is designed to solve a problem. When a problem is narrowly defined (identify which number is largest or add two numbers together and report the sum) the program is not regarded as intelligent. As the program gains complexity (subtract the smallest number in a set from the largest number in the set and report the difference) it becomes difficult for human beings replicate the process of logic that has been performed. And when relationships among a large set of data are evaluated in a myriad of ways (consider the return and risk of various investment options along with the resources and risk tolerance of an individual to provide him with an investment plan for retirement) if is difficult for human beings to conceptualize how the programming works - and they regard it as being intelligent (applying an independent process of thought rather than merely following instructions) simply because they do not understand how it does what it does.

Individuals who are knowledgeable about cognition and technology recognize that computers do not think - they merely execute their programming, which is to say that they follow processes that implement the intelligence of the human being who wrote the programs. In effect, they are mindlessly following their instructions.

What makes computers interesting to researchers is that they can isolate logic. A computer follows instructions flawlessly - unlike a human being who is prone to bring his own ideas to the mix and alter the way in which he acts in pursuit of an objective. Also, a computer program documents the course of logic, so it can be clearly diagnosed where the programming went wrong when a desired outcome is not achieved by examining the input, functions, and output.

The Turing Test

Given that our understanding of intelligence in humans is imprecise, there can be no precise method of assessing the intelligence of computers. The best that has been offered so far is Alan Turing's principle of AI, which loosely suggests that a computer can be considered intelligent if it can trick a person into thinking they are interacting with another human being. (EN: In the Turing test, it is implicit that AI cannot be judged to be superior to human intelligence, merely its equal.)

The classic form of the Turing test involves an interrogator who asks questions, a human being who tenders answers, a computer that tenders answers, and an intermediary between the interrogator and the subjects such that the interrogator is not aware of their identity. At some point, the interrogator identifies which respondent is human. The scoring of the test depends on accuracy (the interrogator correctly identifies the respondent) and the number of questions that must be asked before the distinction can be made accurately.

In other instances, the interpreter is dispensed with - the interrogator uses a computer to pose questions and view answers, unaware of whether the answers he sees were generated by a program or entered by a human being at another terminal.

Other instances depart from Turing's design by comparing the output of a computer to that of a human being by means of statistical analysis to determine, based on various factors, how similar machine responses are to an aggregated profile of human responses.

In some instances, a Turing-like test is used to assess the effectiveness of a computer, seeking to determine whether it is better than a human being. The chess-playing "Deep Blue" program was assessed by its ability to defeat human opponents, eventually defeating a chess grandmaster (Gary Kasparov in 1997) in a match.

However, this does not assess the intelligence of a computer, merely its facility in accomplishing a task. Deep Blue functioned entirely by brute force, assessing the outcome of every possible move far more quickly than a could a human being - but the criteria for assessing whether a move was "good" was derived from human intelligence, such that the computer was not solving problems, merely executing instructions.

Early Programs

Written in 1957, the Logic Theorist program was intended to discover proofs for theories in elementary symbolic logic. The same year, the General Problem Solver was written to track the differences between a present scenario and the desired outcome.

Such programs utilized a flowchart approach, in which a course was determined by a linear process of analysis that presented simple yes/no examinations that would redirect the path to one of many conclusions. More sophisticated models would assemble data into a register that would be analyzed at once to determine an outcome, which had the benefit of considering combinations of items rather than individual items.

The author refers to Eliza and Parry, early programs used in psychology, both of which were chatterbots. Eliza was based on the model of Rogerian psychotherapy, which elicited information by asking open-ended questions, then questions related to a response that the operator had given whereas Parry attempted to model the mental processes of a paranoid schizophrenic. Both were said to be rather good at maintaining a conversation and performed well on Turing tests, to the point that only 48% of human practitioners who reviewed transcripts suggested that the logic was artificial.

(EN: I worked with chatterbots personally in the 1990s, and examined Eliza specifically - it performed very well based on simple logic in a limited topic, and was quite remarkable in that regard. The problem is that the intent of the program is to keep a conversation going, not tot take it anywhere, so the program mostly responds to the last sentence a user entered or, when that is not fruitful, reaches further back into the conversation to find something to say. Ultimately, it is playing with language to have the appearance of understanding - that is, to make a statement or ask a question that seems relevant because it uses the same words as the speaker. To respond to a statement that "I feel X" with "Why do you feel X?" does not require understanding X or even recognizing what X is, but merely recognizing that X is a noun and that a question including that noun will be taken as relevant. It's also notable that psychoanalysis is a structured conversation in which the analyst is interrogating a patient and speaks in a contrived manner to elicit information while providing none - so the very nature of human interaction in these situations is highly unnatural, making it a fairly easy target for computer simulation.)

Expert Systems

The concept of an "expert system" is one that can take on more of a cognitive task with the goal of being able to perform a highly complex operation end to end.

Consider that computers are used to perform parts of a task. That is, given a word problem such as "if you have four dollars and nails cost 75 cents a dozen, how many nails can you buy" a human solver recognizes that this sentence presents a math problem and uses a calculator for them to find the answer (you can buy 60 nails and will have 25 cents left over). An expert system attempts to take on the human task of reading the word problem and figuring out which calculations must be done, which is currently handled by the human solver - that is, it recognizes you must divide one sum by the other, multiply the result by the number of items per package to determine the number of items, and then multiply the remainder by the original number to determine the change.

Getting a computer to perform calculations is a very simple matter - it's what they were designed to do. Getting computers to analyze input to determine what needs to be done has been an insurmountable task because input tends to be unstructured - so the result thus far is that the input must be massaged so that the computer can understand it, ensuring that the question is phrased in a certain way and contains sufficient information for the computer to process it.

(EN: My sense is this is a significant source of job dissatisfaction in the present age: computers are installed to help people - but instead the people must still work to massage the input to the system so that the computer can do its part. The irony is that the people are serving the needs of the machine rather than the other way around. And often, it's a very bad trade, because in order to feed the computer, the people have to do more work and the nature of the work they do is more tedious.)

The author provides examples of a few expert systems in the medical field that are used to perform basic diagnosis, when diagnosis of a condition can be done from quantitative data (such as the levels of certain molecules in a blood sample). That is to say, rather than simply reporting the numbers to a physician who interprets them, the analysis program interprets the numbers to arrive at a diagnosis.

The Intelligence of Intelligent Programs

The lack of a clear definition of "intelligence" complicates the attempt to prove a claim that computers are or are not intelligent. There are also other factors germane to the nature of computers that complicate this evaluation.

One factor is the nature of serial computing: a computer process follows a single thread of execution, considering one command at a time, to arrive at a conclusion whereas the human mind performs a multitude of operations simultaneously. The author asserts that technology has progressed to the point where a computer can have multiple threads of execution and multiple computers can work together across a network, so the serial processing limitation is a moot argument.

(EN: I would agree that the technology is now capable of parallel processing to a theoretically infinite degree, so it is possible in theory for an infinite network of computers to run an infinite number of commands at the same time - but this ability is not exploited by the software to the same level of complexity of the human mind, and the level of complexity of the human mind remains entirely unknown. I expect that in practice, because any given simulation has a finite amount of hardware resources and the programmers may have failed to leverage it adequately, that intelligent programs are still more limited than the human mind in processing multiple simultaneous threads of execution.)

Another factor that is cited is the absence of intuition: human minds are not limited in the way in which they attack a problem, and the solution often will "pop" into someone's mind as if by divine inspiration. However, it is more reasonable to characterize the "eureka" factor as unconscious through rather than absence of thought. There is a process by which a person knows or discovers something, even though they may not be aware of what that reason is. That is, we may not immediately know why we are in a given mood, but on reflective examination we can identify the stimulus and connections that evoked that response.

In essence, the argument that computers lack intuition is an accusation that the programmers failed to consider all the various algorithms that might be brought to bear in the task of analysis - much in the way that the human mind might activate a seemingly unrelated thought process to solve a problem. And again, in theory, it is possible for the designers of a system to develop an "intuitive" program that accounts for all possible algorithms and evaluates not only the problem, but its own problem-solving process, and activates algorithms that are not obviously connected to the situation.

If any human process of thought can be discovered and described, it is possible to write a computer simulation that follows the very same course of logic. Hence if every possible process of thought is documented, the computer can model the human mind in every possible way. The problem again is not a limitation of technology, but sufficiency of programming instruction.

A third factor is that there is a difference between simulating actual intelligence and merely simulating the appearance of intelligence. Consider the chatterbots from earlier in the chapter: I is fairly simple to ask a seemingly intelligent question or to make a statement that seems relevant merely by toying with words without understanding their meaning. That is to say that a chatterbot acts in the manner of a human being can fake their way through a conversation on a topic they do not comprehend, which is clever but not particularly intelligent.

This is likely true of humans as well, as we are very good at pretending to be intelligent when we have lost the plot, so much so that another person may mistake us for being more intelligent than we actually are - so ultimately when we assess a person as being intelligent it is an assessment of their behavior rather than the actual intelligence that drives it. The chief different between humans and computers in this regard is that computer programming is subject to much closer inspection, so it's easier to see where a program is cheating or faking intelligence.

Even so, this third factor is difficult to dispute or support. Given that we have not defined "intelligence" to a sufficient degree, making the further difference between actual and feigned intelligence is even more nebulous.

(EN: It's likely not so difficult after all, if you recall that intelligence is assessed not by the effort, but by the outcome. Intelligence includes the ability to achieve positive outcomes - and this considered a program that is capable of conversing like a human being is not intelligent if its contribution to the conversation fails to produce an outcome, which is in fact the common shortcoming of chatterbots. They are good at talking, but aren't actually applying intelligence.)

JIM.SHAMLIN.COM
DISCLAIMERS · CONTACT