AI Aims to Solve Software Bugs Using Mathematical Methods
Silicon Valley-based startups are combining artificial intelligence with mathematical verification methods to enhance software security and completely eliminate code bugs against cyberattacks.
Researchers and technology startups are aiming to resolve security vulnerabilities in computer code by adapting AI techniques developed for solving complex mathematical problems to software verification processes.
Startups like Palo Alto-based &Truth are attempting to prove that software operates according to its design intent by using the Lean programming language, developed by Microsoft researcher Leonardo de Moura. This method aims to detect and clear bugs in code in advance to address the cybersecurity crisis.
Organizations such as Axiom, Logical Intelligence, OpenAI, and Anthropic are also continuing their work on similar methods. However, experts point out that applying mathematical rules presents challenges due to the complexity of real-world code.
Assessment: Antithesis CEO Will Wilson states that it is difficult to predict all possible behaviors in situations where software interacts with the real world. &Truth co-founder Chris Hsu defines software security as one of the most critical tests of our time.
Sources
- nytimes.com · Link