Showing posts with label human cognition. Show all posts
Showing posts with label human cognition. Show all posts

Artificial Intelligence - Who Was Allen Newell?


Allen Newell (1927–1992) was an American writer who lived from 1927 to 1992.

 Allen In the late 1950s and early 1960s, Newell collaborated with Herbert Simon to develop the earliest models of human cognition.

The Logic Theory Machine depicted how logical rules might be used in a proof, the General Problem Solver modeled how basic problem solving could be done, and an early chess software mimicked how to play chess (the Newell-Shaw-Simon chess program).

Newell and Simon demonstrated for the first time in these models how computers can modify symbols and how these manipulations may be used to describe, produce, and explain intelligent behavior.

Newell began his career at Stanford University as a physics student.

He joined to the RAND Corporation to work on complex system models after a year of graduate studies in mathematics at Princeton.

He met and was inspired by Oliver Selfridge while at RAND, who led him to modeling cognition.

He also met Herbert Simon, who would go on to receive the Nobel Prize in Economics for his work on economic decision-making processes, particularly satisficing.

Simon persuaded Newell to attend Carnegie Institute of Technology (now Carnegie Mellon University).

For the most of his academic career, Newell worked with Simon.

Newell's main goal was to simulate the human mind's operations using computer models in order to better comprehend it.

Newell earned his PhD at Carnegie Mellon, where he worked with Simon.

He began his academic career as a tenured and chaired professor.

He was a founding member of the Department of Computer Science (today known as the school), where he held his major position.

With Simon, Newell examined the mind, especially problem solving, as part of his major line of study.

Their book Human Problem Solving, published in 1972, outlined their idea of intelligence and included examples from arithmetic problems and chess.

To assess what resources are being utilized in cognition, they employed a lot of verbal talk-aloud proto cols, which are more accurate than think-aloud or retrospective protocols.

Ericsson and Simon eventually documented the science of verbal protocol data in more detail.

In his final lecture ("Desires and Diversions"), he stated that if you're going to be distracted, you should make the most of it.

He accomplished this via remarkable accomplishments in the areas of his diversions, as well as the use of some of them in his final effort.

One of the early hypertext systems, ZOG, was one of these diversions.

Newell also collaborated with Digital Equipment Corporation (DEC) founder Gordon Bell on a textbook on computer architectures and worked on voice recognition systems with CMU colleague Raj Reddy.

Working with Stuart Card and Thomas Moran at Xerox PARC to develop ideas of how people interact with computers was maybe the longest-running and most fruitful diversion.

The Psychology of Human-Computer Interaction documents these theories (1983).

Their study resulted in the Keystroke Level Model and GOMS, two models for representing human behavior, as well as the Model Human Processor, a simplified description of the mechanics of cognition in this domain.

Some of the first work in human-computer interface was done here (HCI).

Their strategy advocated for first knowing the user and the task, then employing technology to assist the user in completing the job.

In his farewell talk, Newell also said that scientists should have a last endeavor that would outlive them.

Newell's last goal was to advocate for unified theories of cognition (UTCs) and to develop Soar, a proposed UTC and example.

His idea imagined what it would be like to have a theory that combined all of psychology's restrictions, facts, and theories into a single unified outcome that could be implemented by a computer program.

Soar continues to be a successful continuing project, despite the fact that it is not yet completed.

While Soar has yet fully unify psychology, it has made significant progress in describing problem solving, learning, and their interactions, as well as how to create autonomous, reactive entities in huge simulations.

He looked into how learning could be modeled as part of his final project (with Paul Rosenbloom).

Later, this project was merged with Soar.

Learning, according to Newell and Rosenbloom, follows a power law of practice, in which the time to complete a task is proportional to the practice (trial) number raised to a small negative power (e.g., Time trial -).

This holds true across a broad variety of activities.

Their explanation was that when tasks were completed in a hierarchical order, what was learnt at the lowest level had the greatest impact on reaction time, but as learning progressed up the hierarchy, it was less often employed and saved less time, thus learning slowed but did not cease.

Newell delivered the William James Lectures at Harvard in 1987.

He detailed what it would take to develop a unified theory in psychology in these lectures.

These lectures were taped and are accessible in CMU's library.

He gave them again the following autumn and turned them into a book (1990).

Soar's representation of cognition is based on searching through issue spaces.

It takes the form of a manufacturing system (using IF-THEN rules).

It makes an effort to use an operator.

Soar recurses with an impasse to solve the issue if it doesn't have one or can't apply it.

As a result, knowledge is represented as operator parts and issue spaces, as well as how to overcome impasses.

As a result, the architecture is how these choices and information may be organized.

Soar models have been employed in a range of cognitive science and AI applications, including military simulations, and systems with up to one million rules have been constructed.

Kathleen Carley, a social scientist at CMU, and Newell discussed how to use these cognitive models to simulate social agents.

Work on Soar continues, notably at the University of Michigan under the direction of John Laird, with a concentration on intelligent agents presently.

In 1975, the ACM A. M. Turing Award was given to Newell and Simon for their contributions to artificial intelligence, psychology of human cognition, and list processing.

Their work is credited with making significant contributions to computer science as an empirical investigation.

Newell has also been inducted into the National Academies of Sciences and Engineering.

He was awarded the National Medal of Science in 1992.

Newell was instrumental in establishing a productive and supportive research group, department, and institution.

His son said at his memorial service that he was not only a great scientist, but also a great father.

His weaknesses were that he was very intelligent, that he worked really hard, and that he had the same opinion of you.

~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram

You may also want to read more about Artificial Intelligence here.

See also: 

Dartmouth AI Conference; General Problem Solver; Simon, Herbert A.

References & Further Reading:

Newell, Allen. 1990. Unified Theories of Cognition. Cambridge, MA: Harvard University Press.

Newell, Allen. 1993. Desires and Diversions. Carnegie Mellon University, School of Computer Science. Stanford, CA: University Video Communications.

Simon, Herbert A. 1998. “Allen Newell: 1927–1992.” IEEE Annals of the History of Computing 20, no. 2: 63–76.

What Is Artificial General Intelligence?

Artificial General Intelligence (AGI) is defined as the software representation of generalized human cognitive capacities that enables the ...