Jeff Shrager writes a comprehensive guide to Herbert Simon's 1961 Heuristic Compiler, a program that uses General Problem Solver (GPS) means-end analysis to generate IPL-V code. After being dormant for roughly 65 years, the archival card deck was transcribed and successfully run on a modern Common Lisp interpreter. The system is composed of three parts: a State Description Compiler that derives code from before-and-after memory states, a Functional Description Compiler that modifies existing code based on imperative phrases, and a General Compiler executive that manages the search process. Notably, the program reproduces the exact machine code printed in Simon's 1963 paper, such as the "INSERT AT END OF VALUE LIST" and "SET SIGNAL MINUS" examples, without any modern language rewrites.
- The guide highlights that the compiler treats routine generation as a problem to be solved via operator search rather than a direct translation task.
- A debugging print statement left in routine U113 happens to output the exact stage-by-stage compilation sequence described in the 1963 paper.
- The original 1961 deck fails to compile the J3 routine; a specific card must be modified to erase a redundant description before the state compiler can execute successfully.
- The guide highlights that the compiler treats routine generation as a problem to be solved via operator search rather than a direct translation task.
- A debugging print statement left in routine U113 happens to output the exact stage-by-stage compilation sequence described in the 1963 paper.
- The original 1961 deck fails to compile the J3 routine; a specific card must be modified to erase a redundant description before the state compiler can execute successfully.
This guide offers an accessible overview of SHRDLU, a landmark AI program from the late 1960s and early 1970s developed by Terry Winograd. It explains how the system processes English conversations about colored blocks by integrating parsing, semantic interpretation, and logical reasoning into a single coherent architecture.
Key points include:
- The historical significance of SHRDLU in AI research at MIT
- A breakdown of its modular subsystems from raw input to natural language response
- How the PROGRAMMAR metalanguage allows grammar rules to act as executable programs
- The function of the Micro-Planner engine for theorem proving within Lisp
- The integration of syntactic, semantic, and geometric world knowledge into a unified system
Key points include:
- The historical significance of SHRDLU in AI research at MIT
- A breakdown of its modular subsystems from raw input to natural language response
- How the PROGRAMMAR metalanguage allows grammar rules to act as executable programs
- The function of the Micro-Planner engine for theorem proving within Lisp
- The integration of syntactic, semantic, and geometric world knowledge into a unified system
This article discusses the history of AI, the split between neural networks and symbolic AI, and the recent vindication of neurosymbolic AI through the advancements of models like o3 and Grok 4. It argues that combining the strengths of both approaches is crucial for achieving true AI and highlights the resistance to neurosymbolic AI from some leaders in the deep learning field.
Artificial intelligence has always been around us, with Timothy J. O’Malley’s 1985 book on AI projects for the Commodore 64 being one example. This book explores various AI techniques used in games and other applications, highlighting the foundational methods that have been around since the 1950s and 1960s. It also discusses modern advancements in AI and the ongoing debate about progress and definitions in the field.
