4 ms·

> I'm pretty sure the self-driving cars thing was bullshit the whole time. A great way to string along investors. I somewhat agree that self-driving cars isn't

> I'm pretty sure the self-driving cars thing was bullshit the whole time. A great way to string along investors. I somewhat agree that self-driving cars isn't an obvious solution for Uber when you think about it, because it basically makes the entire business of Uber a stepping stone towards a completely different company. That is to say the majority of everything Uber has built would no longer be relevant. Of course, I suspect Travis always knew this and really it was actually Uber itself that was always bullshit (or at least by the time the company became a unicorn, nobody really knew anything when they were still a glorified limo company) and self-driving cars always the real endgame. The problem was it didn't go according to plan. Considering the relationship Travis and Anthony Levandowky established and the whole history there (Uber legitimately tried to defend Anthony from the Waymo lawsuit until it was blindingly obvious they could no longer do so), I think Travis was dead serious about self-driving cars for a long time. The problem is Travis is a great entrepreneur and businessman, but robotics is an unforgiving field. It's easy to sell an idea. "Wouldn't self-driving cars be great?" And honestly 5 years ago I felt like it might be possible, and many other seasoned roboticists legitimately got on-board (so not just for the money, although some did and I don't blame them because holy shit I have seen some of those offers and they were insane). But at this point, and this is very much an open debate, I believe you need to solve general AI before you can have a meaningful self-driving car product. Everything above applies for L4 autonomy, anything that requires driver attention (such as Tesla's Autopilot) or extremely limited use-cases (such as GM's SuperCruise, which is 100% independent from GM's self-driving car division called Cruise) is a completely different discussion.

jacques_chester 7y ago
I briefly bought into the self-driving hype too, but I had a useful reference case to help me come down to earth: self-driving trains. Rio Tinto (the world's second-largest mining company) has sunk a lot of money into developing self-driving trains. Trains. That have right of way. Their own trains, that they own, on their own rail, that they own, liberally covered with communications towers and sensors (many of which my father put up during his time there). It's about as close to the simplest version of self-driving there is. And they weren't hurting for smarts: they raided quite a few universities for even vaguely relevant talent. They did it, though. Level 4 autonomous trains. It cost them a billion dollars and six years just for one crack at the software. Before that at least a decade of work for who knows how much money. But: trains. That's the functional state of the art worldwide. Trains.
- foobiekr 7y ago
  I would love some links to this case. That is a wonderful counter to the endless reams of self-driving car propaganda that shows up in my personal life on a frequent basis.
  - jacques_chester 7y ago
    Recently saying that they had successfully tested it: https://www.railwaygazette.com/australasia/rio-tinto-completes-autohaul-autonomous-train-project/47822.article https://www.railwaygazette.com/australasia/rio-tinto-complet... Earlier announcement (2014) that they were planning to do so: https://www.railjournal.com/signalling/rio-tinto-gets-set-for-automated-heavy-haul-tests/ https://www.railjournal.com/signalling/rio-tinto-gets-set-fo... Note that the original contract with Ansaldo for development was approximately half what it wound up being. That they will admit to. And I suspect leaves out a lot of the work around the lines which can be used for autonomous and non-autonomous purposes. But then it goes back further, to this article from 2008: https://www.railwaygazette.com/news/rio-tinto-to-go-driverless/33047.article https://www.railwaygazette.com/news/rio-tinto-to-go-driverle... Predicting that autonomous trains would be done in 5 years. That is, they expected to be finished by 2013 ... when the next contract was signed. I won't be surprised if it's announced a few more times in the coming years. Mine automation is a big deal, Rio Tinto have spent heavily on it. Not just trains, they also have spent a lot on developing autonomous mine trucks. The pit is covered with a lot of beacons to help the trucks position themselves and they use regular scans of the site to update the map. I'm not sure how far they got with those. BHP-Billiton seems not to have gone as far, but they have spent a lot of money in remote operations: having the control centres for multiple mines consolidated into special-purpose offices in Perth etc. Cameras everywhere, das blinkenlights, the whole shebang.

jacques_chester 7y ago

I briefly bought into the self-driving hype too, but I had a useful reference case to help me come down to earth: self-driving trains. Rio Tinto (the world's second-largest mining company) has sunk a lot of money into developing self-driving trains. Trains. That have right of way. Their own trains, that they own, on their own rail, that they own, liberally covered with communications towers and sensors (many of which my father put up during his time there). It's about as close to the simplest version of self-driving there is. And they weren't hurting for smarts: they raided quite a few universities for even vaguely relevant talent. They did it, though. Level 4 autonomous trains. It cost them a billion dollars and six years just for one crack at the software. Before that at least a decade of work for who knows how much money. But: trains. That's the functional state of the art worldwide. Trains.

foobiekr 7y ago
I would love some links to this case. That is a wonderful counter to the endless reams of self-driving car propaganda that shows up in my personal life on a frequent basis.
- jacques_chester 7y ago
  Recently saying that they had successfully tested it: https://www.railwaygazette.com/australasia/rio-tinto-completes-autohaul-autonomous-train-project/47822.article https://www.railwaygazette.com/australasia/rio-tinto-complet... Earlier announcement (2014) that they were planning to do so: https://www.railjournal.com/signalling/rio-tinto-gets-set-for-automated-heavy-haul-tests/ https://www.railjournal.com/signalling/rio-tinto-gets-set-fo... Note that the original contract with Ansaldo for development was approximately half what it wound up being. That they will admit to. And I suspect leaves out a lot of the work around the lines which can be used for autonomous and non-autonomous purposes. But then it goes back further, to this article from 2008: https://www.railwaygazette.com/news/rio-tinto-to-go-driverless/33047.article https://www.railwaygazette.com/news/rio-tinto-to-go-driverle... Predicting that autonomous trains would be done in 5 years. That is, they expected to be finished by 2013 ... when the next contract was signed. I won't be surprised if it's announced a few more times in the coming years. Mine automation is a big deal, Rio Tinto have spent heavily on it. Not just trains, they also have spent a lot on developing autonomous mine trucks. The pit is covered with a lot of beacons to help the trucks position themselves and they use regular scans of the site to update the map. I'm not sure how far they got with those. BHP-Billiton seems not to have gone as far, but they have spent a lot of money in remote operations: having the control centres for multiple mines consolidated into special-purpose offices in Perth etc. Cameras everywhere, das blinkenlights, the whole shebang.