Richard Murray: "Can We Really Use Machine Learning in Safety Critical Systems?"

[@jbay088]

"If you're saying I want to spend a certain time in transportation systems then maybe hours is the right argument"... The audience all laughed, but this is a legitimate point, not a joke; many trip planning decisions are made with hours as a metric. Given a fixed destination, miles is of course the right metric, but very often the destination is not fixed. If the family is deciding where to go camping this weekend, you'll probably pick somewhere within a one to three hour drive, and if you can travel faster, that mostly opens up the opportunity for longer-distance trips, rather than shorter travel times. If you're planning where to work or where to live, then you'll again draw a radius based on commute time, rather than distance. Next time you open Google Maps to check for directions, make a note of which number you pay attention to -- the trip distance, or the trip time?

[@weisongwen3042]

nice presentqtion！

[@nonelost1]

What happens if both the machine and the human are “on the ball“ Great! What happens if the machine fails but the human saves the day? Good! And what happens if the human fails but the machine saves the day? Good! But what happens if both the human and the machine “drop the ball?” Bad! Unless this happens significantly less often than seven deaths per 10 billion passenger miles, don’t even!