Escaping the Quicksand: A Call to Arms (arxiv.org)

A new call to action titled "Escaping the Quicksand" highlights a critical issue in the AI/ML community—addressing the burgeoning technical debt that accumulates from decades of software development practices. As AI-enabled engineering accelerates coding processes, it simultaneously exacerbates the risks tied to quickly growing vulnerabilities within these systems. The paper stresses the need for a paradigm shift away from traditional testing and debugging methods, advocating for a more integrated approach combining testing, specification, and mathematical proof of correctness. The authors propose that developers begin co-developing executable specifications alongside existing code to clarify design and enhance testing efficiency. They also emphasize the importance of creating robust semantics infrastructure to support this new framework, which would enable diverse testing methods, from property-based testing to symbolic execution. By calling upon the community to build these tools and infrastructure, they aim to establish a stronger foundation for future AI and human development, significantly improving the reliability and safety of computing systems.