>Further, most AVs explicitly disengage when they encounter situations beyond their understanding the bigger problem here is that the machine may not realize i

>Further, most AVs explicitly disengage when they encounter situations beyond their understanding the bigger problem here is that the machine may not realize it isn't fully understanding the situation; I think that's the more common situation : a computer situational model doesn't match reality, but the machine has the confidence in perception to proceed, producing dangerous effects.