The complexity of deep learning algorithms has exceeded the point where humans are able to comprehend "why" an algorithm arrived at a particular decision. The implications are many, ranging from doomsday predictions to more immediate concerns such as the ability to depend on deep learning.
The deployment of tens of thousands of self-driving vehicles will require a degree of transparency into data and algorithms not seen before. Algorithms will need to be verified as well as held accountable.
The question is, who or what agency or group will ensure the accountability of self-driving vehicles?