一项名为“可执行考古学”（Executable Archaeology）的研究通过运行原始代码，成功复现并分析了两个历史悠久的计算机程序理论模型：赫伯特·西蒙（Herbert Simon）于1961年提出的启发式编译器（Heuristic Compiler），以及麻省理工学院程序员学徒项目（MIT Programmer's Apprentice project）在1984年开发的KBEmacs 1。该研究旨在利用现代手段重新激活这些早期计算系统，以探究其背后的编程逻辑与知识调用机制 1。
在具体复现结果方面，研究人员验证了Heuristic Compiler的输出与Simon在1963年论文中展示的编译代码一致，而KBEmacs则重现了其1984年12月自检过程中的SQRT序列 1。值得注意的是，此次详细的复现工作大量借助了Anthropic公司开发的Claude Code（Opus 5.5版本）作为辅助编程工具 1。尽管这两个系统诞生年代不同且初衷各异，但分析显示它们在适应现有例程和调用编程知识方面存在显著共性 1。相比之下，当代强大的AI编码助手往往缺乏这种基于模式匹配和显式、可执行的编程理论基础 1。
Researchers have successfully resurrected two early artificial intelligence programming systems, Herbert Simon’s 1961 Heuristic Compiler and the MIT Programmer's Apprentice project’s KBEmacs from 1984, using a methodology termed "Executable Archaeology" 1. This approach involves running original code to study historical computational systems 1. Specifically, the reproduction of the Heuristic Compiler matched the compiled code found in Simon’s 1963 paper, while the revived KBEmacs successfully replicated the SQRT sequence from its December 1984 self-test 1.
The extensive replication work was largely facilitated by modern AI tools, specifically Anthropic’s Claude Code (Opus 5.5), which assisted in reconstructing the detailed execution processes 1. The comparative analysis revealed that despite originating in different eras with distinct objectives, both historical systems shared common foundations in pattern matching and knowledge invocation for adapting existing routines 1. These findings suggest that contemporary powerful AI coding assistants often lack the explicit, executable theoretical foundation for programming that characterized these earlier models 1.
评论
还没有评论，欢迎留下第一条。