Universal Programmability - How AI Can Help

W. Tichy,Mathias Landhäußer,Sven J. Körner
DOI: https://doi.org/10.5445/IR/1000037684
Abstract:Everyone should be able to program. Programming in informal, but precise natural language would enable anyone to program and help eliminate the world-wide software backlog. Highly trained software engineers would still be needed for complex and demanding applications, but not for routine programming tasks. Programming in natural language is a monumental challenge and will require AI and software researchers to join forces. Early results, however, appear promising. Combining natural language understanding and ontological reasoning helps remove defects from requirements statements, transforms requirements into UML models, and might even enable script-like programming in specific, narrow domains. An important precondition for rapid progress in this area are benchmarks that help compare different approaches and stimulate competition among researchers.
What problem does this paper attempt to address?