Right now, all legislation is based on the fact that a driver in a vehicle is a human.
But when software takes over all driving tasks from humans, a self-driving car will need to get a drivers license.
We should therefore aim to gain insight into how (self-learning) software can exhibit safe and predictable driving behavior on the public roads.
And if an accident occurs, can we then track back how the software made its decisions to - for example - break, evade or accelerate?
With traditional 'if-then' software, this is doable. For AI however, this is the question.
This is something that we, at RDW, do not have a clear vision on.
And this brings a related question for the audience: is AI capable of saving lives in traffic?"