General Problem Solver

General Problem Solver. It was an early computer program created in the late 1950s to simulate human problem solving. Developed by pioneers Herbert A. Simon, Allen Newell, and J. C. Shaw at the RAND Corporation and Carnegie Mellon University, it was a foundational model in the field of artificial intelligence. The program aimed to demonstrate that a single, general-purpose reasoning system could tackle a wide variety of logical puzzles and tasks.

Overview

The development of the General Problem Solver was a direct outcome of research into cognitive psychology and symbolic computation. It was implemented primarily on the JOHNNIAC computer at the RAND Corporation, with significant theoretical work conducted at the Carnegie Institute of Technology. The project was part of the Logic Theorist lineage, seeking to create a more universal machine intelligence. Its creation was contemporaneous with other early AI milestones like the Dartmouth Conference and work on the Geometry Theorem Prover.

Architecture and operation

The system's architecture was based on manipulating symbols within a defined state space to find a path from an initial state to a desired goal state. It operated using a universal problem solver engine that interpreted separately defined task environment specifications. The core procedure relied on a means-ends analysis algorithm, which recursively identified differences between the current state and the goal. This search was executed using a combination of heuristic methods and backtracking within the computer's memory, influenced by concepts from information processing theory.

Key concepts and mechanisms

The central innovation was the formalization of means-ends analysis as a computational process, a concept that became fundamental to automated planning. It introduced the key notion of operators (programming) that could transform one state into another, applied based on a difference table. The program also utilized subgoaling extensively, breaking down complex problems into simpler components. These mechanisms were framed within the broader physical symbol system hypothesis, which argued that such symbolic manipulation was sufficient for general intelligent action, a theory later debated by proponents of connectionism.

Historical significance and influence

The General Problem Solver had a profound impact on multiple disciplines, cementing the reputations of Allen Newell and Herbert A. Simon in both computer science and cognitive science. It directly inspired subsequent planning systems like STRIPS, developed at SRI International for use with Shakey the Robot. The work provided a concrete model for the information processing psychology movement, challenging behaviorism and influencing theorists like Ulric Neisser. Its principles can be traced to modern algorithms in automated theorem proving and operations research.

Limitations and criticism

Critics, including later AI researchers like Hubert Dreyfus and John Searle, argued that the program's approach was too narrow and brittle, failing to capture the contextual understanding of human thought. It struggled with problems requiring vast combinatorial search or real-world common sense knowledge, limitations highlighted by the subsequent challenges of the AI winter. The program's reliance on well-defined problems in toy domains, such as the Tower of Hanoi, contrasted sharply with the messy realities addressed by fields like robotics and natural language processing. These shortcomings helped motivate alternative paradigms, including expert systems and machine learning.

Category:Artificial intelligence Category:History of artificial intelligence Category:Problem solving