Slashdot Mirror
Tesla Will Have Self-driving Cars In Just Two Years, Elon Musk Boldly Declares (fortune.com)
An anonymous reader writes: In a new interview with Fortune, outspoken Tesla CEO Elon Musk said the electric automaker is just two years away from developing fully autonomous vehicles that can operate ably and safely in any type of environment. While Musk has long championed an automotive age filled with self-driving cars, this is the most optimistic timeline for their deployment we've seen Musk make yet. In fact, Musk in 2014 said the requisite technology to manufacture a self-driving car was still about five to six years away. "I think we have all the pieces," Musk said, "and it's just about refining those pieces, putting them in place, and making sure they work across a huge number of environments—and then we're done. It's a much easier problem than people think it is."
As I recall, Google has been saying for while that they'd have something ready by 2017. On the one hand it seems like it should be surprising if Tesla manages to make it to market at the same time, since they got a much later start. On the other hand, it probably shouldn't surprise us if multiple companies get there at about the same time, since it's less about the cleverness in building the system than it is about having all of the fundamental technological pieces to do it. In particular, I think deep learning neural networks are the core technology that will make effective fully-autonomous cars feasible (plus the sensors, but those have been available for years), and to a large degree the whole world got access to that theory and practice at about the same time.
What is surprising to me is that we haven't heard more from the likes of Freightliner. IMO, that is the first really major market for self-driving vehicles, and those don't even need to be fully autonomous. If tractor-trailer rigs can just drive themselves on the freeway, freight companies can immediately get rid of 90% of their drivers and massively reduce their costs, by having a human drive the truck to the freeway then hop out and let it travel to the destination city, where another human will drive it through town to its destination. Plus, given the price of a semi tractor, adding $50K or even $100K for an automation suite is a relatively small incremental cost, while it's a rather large chunk of change for a passenger vehicle.
(Disclaimer: I work for Google, but not on self-driving vehicles and I don't know any more about the status of Google's system than what is in the public press.)
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
The google cars are driving themselves already now. One could easily just replicate the technology they use and get those things onto the streets, right now. But it would be extremely risky, because google drives them in a fairly controlled environment, and the number of accidents that will happen will multiply by a large count. The question is whether one should start throwing a technology onto the markets when its still incomplete and not polished, or whether one should wait some years before that is possible.
When you launch rockets, you have fairly moderate risk connected to it. Yes, money can burn, but unless you have manned missions, no human will get harmed. Most rocket launches don't have humans on board. Cars on the other hand drive so that they can transport humans. Many cars also drive to transport cargo, but even if they drove without human oversight, they would still be on roads populated by cars with humans. So the risk connected is far higher for cars. Also, with rockets, the astronauts chose themselves if they want to become astronauts, and live with the risk of dying in a rocket accident. But with cars, you can't chose if a self driving car is with you on the street.
We should do what the AC in the rocket launch story suggested: wait until the first service pack is out. We shouldn't throw an immature technology on the market.
Also, one has to talk about software updates for self driving cars. Almost every hardware stops getting software updates by their manufacturer at some point. You can't have cars with EOLed firmware driving on the streets. Nobody should make money by selling security improvements that just mean to flip a switch.
So when will they be able to handle round-abouts and unmarked lanes?
I don't know about roundabouts, but Tesla Autopilot, can already handle unmarked lanes. Tesla collects GPS data while you are driving. So if you drive down a street with unmarked lanes, it just follows the same path as other Teslas that have driven down the same street. My street has no lane markings, and my wife's Tesla can drive it on Autopilot with no problems. So this is a solved problem, and the solution is already in use by consumers.