4 ms·
Uber’s autonomous cars drove 20,354 miles and had to be taken over at every mile
- babesh 10y agoHow generically do these bugs get fixed? How do these bugs get fixed?
- convolvatron 10y agojust last week I saw an uber 'self driving car' in front of me get radically confused about the presence of a bike line to its right. i think it may have been trying to take a turn. turn signal on, hew over into the bike lane. stop abruptly before it hit the curb. try to back up. slew to the left back into the car lane and stop again. turn signal off. repeat. as a software person, I can't understand why any corporate lawyer would let these massive liability exposures run around a city with bikes and pedestrians and other erratic drivers.
- PhantomGremlin 10y agoI can't understand why any corporate lawyer would let these massive liability exposures run around A 'lawyer' isn't something stamped out by a cookie cutter. Just like humans, they're all different. Uber can simply acquire their lawyers straight out of Hollywood villain central casting: Job interviewer: What should we do about all these pesky laws that regulate taxi services? Prospective Uber lawyer: err ... move fast and break things? Job interviewer: Congratulations, you're just the type of employee we're looking for! Until very recently that's worked out pretty well for them.
- true_religion 10y agoLawyers can only advise clients, they cannot force action.
- sokoloff 10y agoThey averaged more than one intervention per mile, but did not "have to be taken over at every mile". I worked on a Daimler-Benz autonomous vehicle back in 1991 and our bus could drive many highway miles at a stretch without intervention. What Uber is doing is many, many times harder than what we were doing, but I have to imagine that they had many clean, fully autonomous miles (and many others that had multiple interventions per mile to offset).
- adrianmonk 10y agoNot only that, but if I were in charge of picking routes for autonomous test vehicles to drive, I'd pick scenarios that the cars have trouble with. Except for a small amount of regression testing, not much point in sending it out to deal with scenarios I know it can already handle. I'd want to gather data on whether my software improvements have given it the ability to handle scenarios it previously failed on. Point being, it's probably not a random sample of possible routes. It's probably skewed. So the statistics probably aren't very meaningful.
- Justsignedup 10y agoHer "one year ago" piece on tipping culture has literally the same 2 ending paragraphs. I am not going to give her praises for journalism, but I do agree that her anecdotes are true, and have experienced similar things. At least she didn't have to worry about getting canceled on when the driver noticed you're not white.