Show HN: Asimov's 4 Laws for AI Contexts

Category: ai-ml

Tags: ai-safety, llm, prompt-engineering, agent-alignment, markdown

Score: 4.8/10 (Innovation: 5, Technical: 3, Documentation: 7, Utility: 4)

This project translates Asimov's Four Laws into a formal framework called Expected Harm Calculus (EHC-4), using mathematical loss functions and lexicographic ordering to guide LLM safety and decision-making. It provides a pure-Markdown specification, system prompts, and evaluation runbooks for integrating safety constraints into autonomous AI agents, emphasizing prompt-injection resistance and harm minimization.

Target audience: AI researchers, AI safety engineers, LLM application developers

Repository: https://github.com/davidsonff/asimov4laws · MIT · 3 stars

View on Hacker News