Heuristic techniques do not yield systems that one can trust
David L. Parnas, December 1985 (Communications of the ACM). Parnas wrote this after resigning from the Strategic Defense Initiative's battle management computing panel. In this section he weighs AI and expert systems as a fix for the SDI software problem. Published December 1985.
Software Aspects of Strategic Defense Systems
More from David L. Parnas · AI and safety · AI and security · Before ChatGPT
Who else?
What it refers to: Heuristic or rule-based programming (AI-2), in which rules of thumb gathered from human experts are encoded into a program, considered for SDI battle-management software.
Other voices
- “You can't trust code that you did not totally create yourself” Ken Thompson,
- “The AI is not interpretable” Christina Delimitrou,
- “These systems are brittle” Radford et al.,
- “The fact that we don't know what it does is the feature” Kelsey Hightower,
- “They can perform more ethically than human soldiers” Ronald C. Arkin,
- “Programs wanting things, knowing things, expecting things, believing things” Edsger W. Dijkstra,










