Yup. An algorithm should be in much better position to recognize its confidence about what it sees and can happen, and alter the driving speed to reduce the amount of uncertain spots that fall within its stopping range.

That is, unless they use neural network magic black boxes to implement the self-driving part.

>That is, unless they use neural network magic black boxes to implement the self-driving part.

Is this actually done? I thought neural networks were more or less used for the vision system to detect parts of the scene (pedestrians, stop-signs, etc). This is the only area where to me it makes sense to take the blackbox tadeoff right now.

Neural nets are also likely to be used to model the behavior of other road users (i.e. 'how is this car/pedestrian/etc likely to move in the next few seconds'?). No-one seems to be seriously considering an end-to-end neural network (i.e. inputs are sensors, outputs are throttle and steering wheel), but neural nets are pervasive components of the full system.