The Coming Need for Formal Specification (benjamincongdon.me)

A recent discussion in the AI community highlights an essential shift: as AI increasingly generates code, the role of engineers may evolve from writing implementation code to focusing on tests and formal specifications. While many initially believed that AI would struggle with producing accurate code, advancements in AI's understanding of code patterns have allowed for significant strides in automatically generating unit tests. However, a more complex challenge remains in verifying holistic system behavior and ensuring component interactions are correctly designed, a task traditionally reliant on deep understanding and expertise. The article advocates for a rising importance of formal verification, which allows developers to mathematically prove that software behaves as intended. As the cost of generating code decreases, the same is anticipated for formal verification, making it more accessible. With this shift, the author envisions a future where AI not only assists in code generation but also aids in creating formal models and proofs, expediting the verification process. However, to make this vision feasible, there is a pressing need to educate more developers in formal methods, as current expertise in this niche area is scarce. The call to action suggests incorporating formal verification into computer science curricula to prepare a new generation of engineers for these evolving demands.