The integration of AI into large scale engineering projects, such as self-driving cars, raises the issue of the need for robustness to imperfection in these systems. There is also a desire for algorithms and systems to be provably correct and safe, according to NIST and computer scientist Don Knuth.