Allen Newell

• 1. Introduction

  ›

  Allen Newell was born on the 19th of March 1927, and by the time he died on the 19th of July 1992, he had helped build the intellectual foundations of an entirely new science. What that science would eventually become is something most people now take for granted. But in the early 1950s, no one could have predicted that a physicist from Stanford, restless with pure mathematics, would end up co-creating the very first artificial intelligence programs.

  Newell spent his career split between two institutions: the RAND Corporation in Santa Monica and Carnegie Mellon University, where he held appointments in computer science, business, and psychology simultaneously. That unusual combination was not an accident. Newell genuinely believed that understanding the mind and building intelligent machines were the same project, approached from different angles.

  He and his longtime collaborator Herbert A. Simon received the ACM Turing Award in 1975, the highest honor in computing. But the awards only hint at the questions worth asking: How did a man trained in physics come to define the field of artificial intelligence? What ideas did he contribute that still shape how computers reason today? And what was the grand unified theory he spent the last years of his life trying to complete?

• 2. From Physics to Information

  ›

  Newell finished his physics degree at Stanford in 1949 and moved to Princeton to study mathematics as a graduate student. He stayed only a year. His early exposure to game theory, then a largely unknown field, combined with his mathematics work, convinced him that pure theory alone was not what he wanted. He preferred a blend of experiment and theory, and Princeton's graduate mathematics program did not offer that mix.

  In 1950, he left for the RAND Corporation in Santa Monica, where he joined a group studying logistics problems for the Air Force. Working alongside Joseph Kruskal, he produced two theoretical papers on organization theory, which trained his attention on how groups make decisions under pressure.

  His dissatisfaction with small-scale laboratory experiments on decision making pushed him toward something more realistic. Joining RAND colleagues John Kennedy, Bob Chapman, and Bill Biel at an Air Force Early Warning Station, he studied how flight crews actually handled information and made choices in real operational conditions. The Air Force provided funding in 1952 to build a simulator for studying cockpit interactions. Out of those flight-crew studies came a conviction that would shape everything he did afterward: information processing, not motivation or emotion, is the central activity inside any organization, human or machine.

• 3. The Birth of a Field

  ›

  In September 1954, Newell attended a seminar where Oliver Selfridge demonstrated a running computer program that learned to recognize letters and other patterns. That demonstration changed the direction of his life. Within a couple of months he was writing a 1955 paper titled The Chess Machine: An Example of Dealing with a Complex Task by Adaptation, which laid out a design for a chess-playing program that would operate in a humanoid fashion.

  The paper caught the attention of Herbert A. Simon, an economist who would later win the Nobel Prize. Simon and programmer J. C. Shaw joined Newell, and together the three built the Logic Theorist, which is widely regarded as the first true artificial intelligence program. Newell's specific contributions were the technical architecture that made it possible: list processing, which became the most important programming paradigm in AI; the application of means-ends analysis to general reasoning; and the use of heuristics to constrain the search for solutions.

  The three men presented the Logic Theorist at the Dartmouth conference of 1956, an informal gathering of researchers curious about machine intelligence. That conference is now widely considered the birth of artificial intelligence as a formal discipline. The people who attended went on to lead AI research for the following two decades, and Newell stood among the most prominent of them.

• 4. The General Problem Solver

  ›

  After Dartmouth, Newell and Simon founded an artificial intelligence laboratory at Carnegie Mellon University and began a collaboration that would last for decades. Their next landmark program appeared in 1957: the General Problem Solver, a system that embodied means-ends analysis in a more rigorous and general form than the Logic Theorist had managed.

  Means-ends analysis is a reasoning strategy: identify the gap between where you are and where you want to be, then select an action that reduces that gap. The General Problem Solver made this strategy explicit and mechanical, demonstrating that a single underlying method could handle a wide variety of reasoning tasks.

  Out of this work grew a bolder philosophical claim: the physical symbol systems hypothesis. Newell and Simon argued that any intelligent behavior, in a human or a machine, could ultimately be reduced to the manipulation of symbols according to rules. This hypothesis was controversial from the moment they proposed it. Critics argued it missed essential features of biological intelligence. Supporters saw it as the most precise scientific account of mind ever offered. Either way, it set the terms of debate in cognitive science and AI for a generation, and the debate has never fully resolved.

• 5. Soar and the Unified Theory

  ›

  Newell's final and most ambitious project was a cognitive architecture he called Soar. The goal was to build a single computational framework that could account for all aspects of human cognition, from perception and memory to problem solving and learning, within one unified system.

  His unified theory of cognition was published in 1990, two years before his death. Newell regarded Soar not as a finished answer but as a direction. Some of his last correspondence focused on how Soar could be improved, and he was still working toward that objective when he died on the 19th of July 1992.

  The field he initiated, cognitive architectures, remains active today in both artificial intelligence research and computational cognitive science. Soar itself is still maintained and studied, and researchers continue to extend it in the directions Newell pointed toward. The ACM-AAAI Allen Newell Award was later named in his honor, as was the Award for Research Excellence of the Carnegie Mellon School of Computer Science. His final honor arrived in the same year he died: the U.S. National Medal of Science, awarded in 1992.