Extensive discussion a few days ago: <a href="https://news.ycombinator.com/item?id=15741019" rel="nofollow">https://news.ycombinator.com/item?id=15741019</a>
I'm not in Pittsburgh or any of the places where people are taking rides in Uber's autonomous test vehicles, but from a few YouTube videos it appears they can't get more than a few blocks in urban traffic without needing human intervention. They've been at it for about 2.5 years now, not including the work that was inherited from the CMU robotics department.<p>I think if a company can program and train an autonomous vehicle to reliably go 5 or 10 miles in city traffic, then it's an indicator they've got the genius level problems solved. Safety validation is a whole other ball game, labour intensive but not super difficult.<p>A few companies have passed that benchmark. There are others who, in spite of being ambitious and well-capitalized, and having had a few years to incubate, appear to be painting by numbers, and don't seem to know what to do when the instruction manual runs out of instructions.<p>Talent is the big X factor in this race and not everybody's got it. Do you think Uber will be ready to receive another 24,000 Volvo's on top of the ~200 they've got somewhere in the 2019/2020 timeframe?<p>I'm not betting on it.