The Problem of Reliability and Robustness

Machines t we've had for the last hundred years have been deterministic. As we start to build in probablistic things, we start to have machines whose behaviour is perhaps not as certain as we would we would like. How will we go about determining what reliability or robustness means in this ai world? So if you think about lao and aeroplane work, nowadays, there's actually provable, verifiable software for certain parts of the kernels. I just like we know that about usb drivers. We don't have formal verification to show that cars will do as they're instructed. The regulations on cars are pretty loose. And we mostly rely on it seems to do what