Road to autonomous vehicles update: take (parts of) California
November 1, 2018 1:44 PM Subscribe
You might have been taken aback by headlines like "Waymo can test fully driverless cars on California roads" (Engadget), and while it is big news, the official California DMV notice is limited in scope, in part that this permit is limited to Santa Clara County, and requires remote engineer monitoring. The CA DMV also notes that "60 manufacturers are currently permitted to test autonomous* vehicles in California with a safety driver." But autonomous vehicles aren't just coming to Silicon Valley, they're rolling out around the world, as summarized by Synced Review (via Medium), who keep up with self-driving news from around the world, and also pulled from KPMG’s AV readiness index (summary; full PDF report).
For more details of US states, the Insurance Institute for Highway Safety's recently summarized that there are nine other states that have permitted autonomous vehicle deployment, plus D.C., another states that permit testing, and seven states allow vehicles to operate without a driver, either for testing or deployment.
* Note that there are five levels of automation, where level zero is completely manual operations, and only at level five is the system fully automation in all scenarios. Level 2, or L2, means that at least one driver assistance system of "both steering and acceleration/ deceleration using information about the driving environment" is automated. L3 requires human interaction in moments of safety concern, which is tricky to re-engage the driver suddenly. So many companies are looking ahead to L4, where the vehicle operates on its own within the "operational design domain (ODD)" of the vehicle—meaning it does not cover every driving scenario.
Autonomous features have been creeping into vehicles for a while. Back in 2016, Car and Driver compared "brainless driving" features in four luxury cars, which covered cruising, braking, and steering, and tested those features in city, on highways, and in a rural area, and parking (jump to the summary of results). Similar features are far more common in 2018, and there are even more advanced or active safety system features that are optional or standard across a variety of cars and models, as recently summarized by Consumer Reports.
Also of note in the topic of gradual roll-outs, from Ars Technica: Waymo began commercial service recently, and hardly anyone noticed, which was by design, and even an option because Waymo is backed by Google's wealth, where Uber and other companies need investors, and to get investors one needs to promote their successes. Ars Technica reports how startups are building driverless cars without Google’s billions, and notes "Driverless shuttles may be the first step toward fully self-driving cars."
A few videos with additional information on different autonomous systems in development:
For more details of US states, the Insurance Institute for Highway Safety's recently summarized that there are nine other states that have permitted autonomous vehicle deployment, plus D.C., another states that permit testing, and seven states allow vehicles to operate without a driver, either for testing or deployment.
* Note that there are five levels of automation, where level zero is completely manual operations, and only at level five is the system fully automation in all scenarios. Level 2, or L2, means that at least one driver assistance system of "both steering and acceleration/ deceleration using information about the driving environment" is automated. L3 requires human interaction in moments of safety concern, which is tricky to re-engage the driver suddenly. So many companies are looking ahead to L4, where the vehicle operates on its own within the "operational design domain (ODD)" of the vehicle—meaning it does not cover every driving scenario.
Autonomous features have been creeping into vehicles for a while. Back in 2016, Car and Driver compared "brainless driving" features in four luxury cars, which covered cruising, braking, and steering, and tested those features in city, on highways, and in a rural area, and parking (jump to the summary of results). Similar features are far more common in 2018, and there are even more advanced or active safety system features that are optional or standard across a variety of cars and models, as recently summarized by Consumer Reports.
Also of note in the topic of gradual roll-outs, from Ars Technica: Waymo began commercial service recently, and hardly anyone noticed, which was by design, and even an option because Waymo is backed by Google's wealth, where Uber and other companies need investors, and to get investors one needs to promote their successes. Ars Technica reports how startups are building driverless cars without Google’s billions, and notes "Driverless shuttles may be the first step toward fully self-driving cars."
A few videos with additional information on different autonomous systems in development:
- How Self-Driving Cars Work, aTEDxWilmingtonSalon presentation by David Silver, who leads the Self-Driving Car Team at Udacity, after working for Ford's autonomous program
- That's all views of well-marked, clear roads, so here's a short video promoting Ford’s Industry first autonomous vehicle tests in snow, posted in Jan. 2016
- And that video is from driving through a well-marked, fake city, so here's a more recent presentation of MIT's CSAIL-supported program, navigating unmapped, unstriped rural roads, from the partnership with Toyota on preparing self-driving cars for country roads
Previous "Road to Autonomous Vehicle" posts:
- The long, winding road to fully automated cars (Sep 7, 2017)
- The winding road to fully automated cars passes more milestones (Dec 27, 2017)
And yes, Propellerheads' "Take California" came to mind while writing this, why do you ask?
posted by filthy light thief at 1:47 PM on November 1, 2018 [1 favorite]
- The long, winding road to fully automated cars (Sep 7, 2017)
- The winding road to fully automated cars passes more milestones (Dec 27, 2017)
And yes, Propellerheads' "Take California" came to mind while writing this, why do you ask?
posted by filthy light thief at 1:47 PM on November 1, 2018 [1 favorite]
From the Embark link...
"the driver will regularly go many hours down the road without actually being involved, and when they are involved, it’s usually just for a few seconds"
That's not how driving works.
Not seeing how this ends well when safety drivers, being people, tending to be lazy, will overrely on the imperfect automation and create accidents as a result.
posted by kokaku at 1:51 PM on November 1, 2018
"the driver will regularly go many hours down the road without actually being involved, and when they are involved, it’s usually just for a few seconds"
That's not how driving works.
Not seeing how this ends well when safety drivers, being people, tending to be lazy, will overrely on the imperfect automation and create accidents as a result.
posted by kokaku at 1:51 PM on November 1, 2018
The National Motor Vehicle Crash Causation Survey estimated that driver error was a critical reason of 94% of collisions between 2005 and 2007. That's more than 2 million crashes in 2.5 years in one country alone. We need to get this right, the sooner the better.
posted by mosst at 1:55 PM on November 1, 2018 [9 favorites]
posted by mosst at 1:55 PM on November 1, 2018 [9 favorites]
"That's not how driving works. "
It kind of is sometimes, I've gotten in the car, blanked out, and been arriving at my destination many times in my life. Long car trips are a nightmare for me, in part because there's nowhere more sleep inducing than a car on a long, steady, highway journey. The bigger part though is that it's hypnotic and boring and much of the time I'm just holding still unless something weird happens.
"Not seeing how this ends well when safety drivers, being people, tending to be lazy, will overrely on the imperfect automation and create accidents as a result.""
As opposed to right now where people, being people, tend to be shitty at everything if there's another person at all involved and create accidents constantly as a result?
posted by GoblinHoney at 2:02 PM on November 1, 2018 [9 favorites]
It kind of is sometimes, I've gotten in the car, blanked out, and been arriving at my destination many times in my life. Long car trips are a nightmare for me, in part because there's nowhere more sleep inducing than a car on a long, steady, highway journey. The bigger part though is that it's hypnotic and boring and much of the time I'm just holding still unless something weird happens.
"Not seeing how this ends well when safety drivers, being people, tending to be lazy, will overrely on the imperfect automation and create accidents as a result.""
As opposed to right now where people, being people, tend to be shitty at everything if there's another person at all involved and create accidents constantly as a result?
posted by GoblinHoney at 2:02 PM on November 1, 2018 [9 favorites]
So basically they're going to put their buggy software into production and blame users (other drivers and pedestrians) for their mistakes, false assumptions, and oversights?
This is literally how most people drive already, let's be honest.
I want this to have tons and tons more regulation and testing, but I was a commuting cyclist and pedestrian for years before I started driving in my late 20s. I have been nearly hit on multiple occasions and at one point literally driven off the road over a curb by people not paying attention, and on more than one of those occasions got shouted at or flipped off by the person driving. So.
posted by Sequence at 2:02 PM on November 1, 2018 [6 favorites]
This is literally how most people drive already, let's be honest.
I want this to have tons and tons more regulation and testing, but I was a commuting cyclist and pedestrian for years before I started driving in my late 20s. I have been nearly hit on multiple occasions and at one point literally driven off the road over a curb by people not paying attention, and on more than one of those occasions got shouted at or flipped off by the person driving. So.
posted by Sequence at 2:02 PM on November 1, 2018 [6 favorites]
Does anyone really WANT driverless cars? Maybe its generational, but I enjoy driving. And where else is that time going to go? To people's phones no doubt. Do we really need this?
posted by BigBrooklyn at 2:10 PM on November 1, 2018 [1 favorite]
posted by BigBrooklyn at 2:10 PM on November 1, 2018 [1 favorite]
BigBrooklyn: "Does anyone really WANT driverless cars? Maybe its generational, but I enjoy driving. And where else is that time going to go? To people's phones no doubt. Do we really need this?"
I don't know what your generation is but I'm in my fifties and have hated driving for a long time. Podcasts make it a little better but it's just dead time that could be better used.
posted by octothorpe at 2:15 PM on November 1, 2018 [7 favorites]
I don't know what your generation is but I'm in my fifties and have hated driving for a long time. Podcasts make it a little better but it's just dead time that could be better used.
posted by octothorpe at 2:15 PM on November 1, 2018 [7 favorites]
I wonder how long before running out in front of driverless cars becomes something that kids dare each other to do. They have to stop, right? It seems like it would be really easy to manipulate autonomous vehicles into doing things if you understood how they avoided obstacles.
posted by oulipian at 2:17 PM on November 1, 2018
posted by oulipian at 2:17 PM on November 1, 2018
I'm not sure what generation you are BigBrooklyn, but yes, I do want driverless cars. I'm nearly 50 and most of the driving I do is purposeful and while I don't mind it there are times, a lot of them, when I'd rather not have to do it. Six hour drives at night to see family? I'll pass. Any drive, literally anything, in stop and go traffic? Jesus, take the wheel.
The thing is, it really has to be 100% hands free or it's no go. A system that requires me to pay attention so that I can take over at any moment, but leaves me doing nothing 99% of the time is probably not a good thing. If I'm just sitting there doing nothing, I'm likely going to tune out. That's exactly what you don't want.
When I drive long distances I don't even like to use cruise control, because I find that having to control the car's speed means that I'm paying attention to what I'm doing at least some of the time.
posted by It's Never Lurgi at 2:20 PM on November 1, 2018 [7 favorites]
The thing is, it really has to be 100% hands free or it's no go. A system that requires me to pay attention so that I can take over at any moment, but leaves me doing nothing 99% of the time is probably not a good thing. If I'm just sitting there doing nothing, I'm likely going to tune out. That's exactly what you don't want.
When I drive long distances I don't even like to use cruise control, because I find that having to control the car's speed means that I'm paying attention to what I'm doing at least some of the time.
posted by It's Never Lurgi at 2:20 PM on November 1, 2018 [7 favorites]
oulipian, I predict that "Trolley Probleming" will be the hot dangerous teen craze of futureyear 20XX! That's when two or more kids run out in front of a self-driving car from different angles to see which one it decides to run over.
posted by prize bull octorok at 2:23 PM on November 1, 2018 [15 favorites]
posted by prize bull octorok at 2:23 PM on November 1, 2018 [15 favorites]
I want driverless cars for two reasons. One is I dislike driving. The other is, driverless cars will have to be at least as good as the average driver (if not better) to be mass marketed, but they should be that good *all the time*. They don't stay up too late and wake up groggy. They don't talk on their phone while driving. They don't get drunk. Eliminating drunk drivers alone will cut the traffic fatality rate in the USA.
posted by fings at 2:26 PM on November 1, 2018 [5 favorites]
posted by fings at 2:26 PM on November 1, 2018 [5 favorites]
Thinking about how driverless will have to manage risks, I can imagine several scenarios for which I have not yet been able to imagine a solution. The projects should therefore all be shutdown.
posted by skewed at 2:30 PM on November 1, 2018 [2 favorites]
posted by skewed at 2:30 PM on November 1, 2018 [2 favorites]
Haha, brilliant. A future King Solomon might use “Trolley Probleming” to decide who’s the mom.
Otherwise, we might use it to settle arguments (like old timey duels), decide who gets the girl (or boy), ration scarce resources (like food), the possibilities are endless!
posted by notyou at 2:32 PM on November 1, 2018 [1 favorite]
Otherwise, we might use it to settle arguments (like old timey duels), decide who gets the girl (or boy), ration scarce resources (like food), the possibilities are endless!
posted by notyou at 2:32 PM on November 1, 2018 [1 favorite]
At the moment, I'd bet that I'll ride in an autonomous car, before I ride on my city's latest public transport project (Sydney Metro, construction recently started, completion pegged for 2024 but not including the inevitable time and cost blowout)
posted by other barry at 2:32 PM on November 1, 2018 [2 favorites]
posted by other barry at 2:32 PM on November 1, 2018 [2 favorites]
mosst: The National Motor Vehicle Crash Causation Survey estimated that driver error was a critical reason of 94% of collisions between 2005 and 2007. That's more than 2 million crashes in 2.5 years in one country alone. We need to get this right, the sooner the better.
And I can't imagine that things have gotten better since 2007, with the increased presence (and casual acceptance for the use of) mobile devices in the hands of drivers.
To which I keep thinking "how far do autonomous vehicles have to come before they're simply less dangerous, or simply less fatal, than people? For the media's love of reporting on misery, car crashes are like gun violence -- not reported unless it's serious, so you don't really see how wide-spread, and how mundane, it is. Vehicle crashes, like gun injuries, aren't the result of dangerous individuals or groups, but the accidental misuse of dangerous "instruments" by people who aren't being careful enough.
posted by filthy light thief at 2:34 PM on November 1, 2018 [3 favorites]
And I can't imagine that things have gotten better since 2007, with the increased presence (and casual acceptance for the use of) mobile devices in the hands of drivers.
To which I keep thinking "how far do autonomous vehicles have to come before they're simply less dangerous, or simply less fatal, than people? For the media's love of reporting on misery, car crashes are like gun violence -- not reported unless it's serious, so you don't really see how wide-spread, and how mundane, it is. Vehicle crashes, like gun injuries, aren't the result of dangerous individuals or groups, but the accidental misuse of dangerous "instruments" by people who aren't being careful enough.
posted by filthy light thief at 2:34 PM on November 1, 2018 [3 favorites]
Does anyone really WANT driverless cars?
I would like to empower seniors, the visually impaired and a whole range of people with disabilities, people without cars who are under-served by public transit and taxis/ridesharing, and I would like drunk or tired drivers to stop killing people.
But other than that, I also like driving and I'll keep doing it myself most of the time.
posted by allegedly at 2:35 PM on November 1, 2018 [6 favorites]
I would like to empower seniors, the visually impaired and a whole range of people with disabilities, people without cars who are under-served by public transit and taxis/ridesharing, and I would like drunk or tired drivers to stop killing people.
But other than that, I also like driving and I'll keep doing it myself most of the time.
posted by allegedly at 2:35 PM on November 1, 2018 [6 favorites]
skewed: Thinking about how driverless will have to manage risks, I can imagine several scenarios for which I have not yet been able to imagine a solution. The projects should therefore all be shutdown.
But there are currently people in similar situations, because they are already driving those cars. It's too late to make cars go away*, so why would difficult scenarios be worse off if handled by "robots"?
* Though maybe with more people using for-hire, on-demand ride services, the idea that private vehicles and ride-hailing is a "movement tax" that the individual carries without thought, so why don't we spend more money on transit and increase safety, air quality and congestion!
Also, I am very much a fan of not driving myself places. Transit is great, and I use it daily, but it's of limited service outside of my commute. Maybe microtransit (but with increased density? Free rides for 6 months?) + autonomous vehicles = fewer single-rider vehicles?
posted by filthy light thief at 2:41 PM on November 1, 2018 [1 favorite]
But there are currently people in similar situations, because they are already driving those cars. It's too late to make cars go away*, so why would difficult scenarios be worse off if handled by "robots"?
* Though maybe with more people using for-hire, on-demand ride services, the idea that private vehicles and ride-hailing is a "movement tax" that the individual carries without thought, so why don't we spend more money on transit and increase safety, air quality and congestion!
Also, I am very much a fan of not driving myself places. Transit is great, and I use it daily, but it's of limited service outside of my commute. Maybe microtransit (but with increased density? Free rides for 6 months?) + autonomous vehicles = fewer single-rider vehicles?
posted by filthy light thief at 2:41 PM on November 1, 2018 [1 favorite]
I'm sorry, my comment was facetious, and in a dumb way.
posted by skewed at 2:45 PM on November 1, 2018
posted by skewed at 2:45 PM on November 1, 2018
Assuming BigBrooklyn is not pulling our leg with his comment, here is my response.
While there are plenty of people who enjoy driving there are many more who don’t or don’t enjoy the driving that they must do most of the time, (Bumper to bumper commute into Boston anyone?) There are many other people who cannot drive for reasons of physical disability including millions of elederly past their driving years. The rural elderly, for whom food and particularly medical care are often long distances away will benefit especially. There are many younger people who have no interest in learning to drive, which they consider a dangerous and expensive activity.
Yes, there are many, many people who welcome the advent of driverless cars.
posted by haiku warrior at 2:50 PM on November 1, 2018 [7 favorites]
While there are plenty of people who enjoy driving there are many more who don’t or don’t enjoy the driving that they must do most of the time, (Bumper to bumper commute into Boston anyone?) There are many other people who cannot drive for reasons of physical disability including millions of elederly past their driving years. The rural elderly, for whom food and particularly medical care are often long distances away will benefit especially. There are many younger people who have no interest in learning to drive, which they consider a dangerous and expensive activity.
Yes, there are many, many people who welcome the advent of driverless cars.
posted by haiku warrior at 2:50 PM on November 1, 2018 [7 favorites]
Fellow MeFite allegedly beat me to my major points.
posted by haiku warrior at 2:53 PM on November 1, 2018
posted by haiku warrior at 2:53 PM on November 1, 2018
I drive, but I also hate driving, and am eager for safe, reliable driverless cars.
posted by salt grass at 3:09 PM on November 1, 2018 [2 favorites]
posted by salt grass at 3:09 PM on November 1, 2018 [2 favorites]
"A system that requires me to pay attention so that I can take over at any moment, but leaves me doing nothing 99% of the time is probably not a good thing."I guess you've never driven in morning traffic, where people do all kinds of things except pay attention.
posted by Marky at 3:23 PM on November 1, 2018
Automated safety features could save lives. We lose far, far too many people to automobile carnage, and those are just the direct deaths. There are also the many people living with injuries caused by automobiles, and all the people with reduced mobility options because it's too unsafe to travel outside of a car, or too inconvenient to travel the longer distances required because cities and towns have been designed around automobiles, not people.
We shouldn't use the potential existence of automated cars as an excuse to ignore all the much-less-exciting technology that already exists for saving lives, like designing safer streets. It's a lot harder to lull yourself into a fugue state when you're not driving on wide, straight, fast streets with very few things to pay attention to outside your car.
We could even require that cars have fewer things to pay attention to inside them. How the fuck is it okay to have enormous bright touchscreens built into automobile dashboards? Why do the agencies responsible for regulating automobile manufacturers allow that? That's aside from all the things drivers bring into the cars with them that are more interesting than driving, and we could do more about that if it weren't just accepted as something everyone does.
We could fund transit better, and encourage redesign of existing transit systems for greater efficiency, making it more useful for all trips, not just commutes. We could use some street space just for transit, so that it's not so often slower than a private automobile would be. Bus-only lanes don't require any more investment than paint, and they're a huge improvement for transit speed and reliability.
We could change our land use priorities, so that the distances between our homes and other destinations aren't so punishingly far. Think about how much closer together things could be if we weren't more concerned with storing people's cars while they're not being used than we are with making sure places are easily accessible to people without a car. Retail businesses are often required to build and maintain enough car parking so that no driver need fear not getting a space on the busiest shopping days of the year. Black Friday's in three weeks: keep an eye on parking lots that day, and note how much space goes unused even then.
Automated cars would be great. Even the limited tech that exists now has potential, but even the adaptive cruise control systems on the market mostly fail to recognize and respond to stopped vehicles ahead . (Also from IIHS, a review of an aftermarket warning system looks a lot more promising right now. It can't attempt to control the car for you, it just helps alert you to problems, helping you be more attentive.
There are plenty of problems automated cars could solve, but if we don't bother solving the problems we can solve right now because of handwaving about future tech, the carnage will continue for a lot longer than it will if we take all our options for stopping it seriously.
posted by asperity at 3:32 PM on November 1, 2018 [11 favorites]
We shouldn't use the potential existence of automated cars as an excuse to ignore all the much-less-exciting technology that already exists for saving lives, like designing safer streets. It's a lot harder to lull yourself into a fugue state when you're not driving on wide, straight, fast streets with very few things to pay attention to outside your car.
We could even require that cars have fewer things to pay attention to inside them. How the fuck is it okay to have enormous bright touchscreens built into automobile dashboards? Why do the agencies responsible for regulating automobile manufacturers allow that? That's aside from all the things drivers bring into the cars with them that are more interesting than driving, and we could do more about that if it weren't just accepted as something everyone does.
We could fund transit better, and encourage redesign of existing transit systems for greater efficiency, making it more useful for all trips, not just commutes. We could use some street space just for transit, so that it's not so often slower than a private automobile would be. Bus-only lanes don't require any more investment than paint, and they're a huge improvement for transit speed and reliability.
We could change our land use priorities, so that the distances between our homes and other destinations aren't so punishingly far. Think about how much closer together things could be if we weren't more concerned with storing people's cars while they're not being used than we are with making sure places are easily accessible to people without a car. Retail businesses are often required to build and maintain enough car parking so that no driver need fear not getting a space on the busiest shopping days of the year. Black Friday's in three weeks: keep an eye on parking lots that day, and note how much space goes unused even then.
Automated cars would be great. Even the limited tech that exists now has potential, but even the adaptive cruise control systems on the market mostly fail to recognize and respond to stopped vehicles ahead . (Also from IIHS, a review of an aftermarket warning system looks a lot more promising right now. It can't attempt to control the car for you, it just helps alert you to problems, helping you be more attentive.
There are plenty of problems automated cars could solve, but if we don't bother solving the problems we can solve right now because of handwaving about future tech, the carnage will continue for a lot longer than it will if we take all our options for stopping it seriously.
posted by asperity at 3:32 PM on November 1, 2018 [11 favorites]
As a confirmed pedestrian, yeah I absolutely do want driverless cars because *literally* this morning a driver (not car fault) obviously texting blew through a red light towards me and my 3yo, and then someone else got aggressive with my SO on the way home because he sticks to the 20mph speed limit. And this is a daily thing, not remotely unusual.
Oh, and my aunt who practically brought me up can barely remember my name but still drives on a daily basis and it terrifies me that she might kill someone but the DVLA won't do jack shit.
posted by threetwentytwo at 3:35 PM on November 1, 2018 [7 favorites]
Oh, and my aunt who practically brought me up can barely remember my name but still drives on a daily basis and it terrifies me that she might kill someone but the DVLA won't do jack shit.
posted by threetwentytwo at 3:35 PM on November 1, 2018 [7 favorites]
There's an angle on the autonomous vehicle story that I've basically never seen reported anywhere other than an AI blog. (again more recently)
California DMV requires AV test licensees to put out a report every year on the performance; amongst other things, how far test cars are driving autonomously and how many "disengagements" there are. A disengagement is when the autonomous control system is, well, disengaged and the safety driver takes over. This could be because the driver thinks it's about to do something dangerous, because the system doesn't know what to do, or because of some sort of technical failure, like a computer rebooting.
The first six months that Waymo (back then Google Auto) tested their vehicles on California roads, Sep 2014 to Feb 2015, they had a disengagement every 700 miles or so. That's the equivalent of over 1,400 disengagements per million miles. One year later, their disengagement rate had improved to a disengagement about every 5000 miles, which is the equivalent of around 200 disengagements per million miles. Seven times better in the first year!
The most recent six months available is through November 2017, so that's almost two more years for improvement. Waymo's disengagement rate is no longer one every 5000 miles, it's... actually, it's still about one every 5000 miles. A really good November pushed them up to one every 6000 miles. With all of the resources of Google at their back, the smartest engineers in the world were not able to improve substantially in two years, and they're still around 200 disengagements per million miles. (And Waymo is streets ahead of their competitors.) I'm eagerly anticipating the 2018 reports to see whether there has been actual progress again, or whether the technology is plateaued.
To compare 200 disengagements per million miles, there's a subset of drivers that have approximately 10 collisions per million miles -- drivers with a blood alcohol content of 0.08. The superset of all drivers, sober, sleepy, stoned, distracted, young, old and otherwise, has a little less than 4 collisions per million miles. I haven't had time to calculate what the rate for a baseline driver -- a 25 to 70 year old who is not drunk, tired or texting but it's obviously much less, and that's clearly the standard we should be setting.
In AI circles, there's a thing that happens on the regular since the early 1970s called "AI Winter" where a bunch of progress is made, and ambitious predictions are made, but then it turns out that it's just not possible to get there, and the field quiets down for a while. I wouldn't be surprised if we're there with current AI techniques -- turns out our current machine learning approaches solve some problems, but just can't get there on others.
I'd like to think that AV development will either soon break this barrier so it's not stuck 20 times worse than a drunk driver, or that the technology is shelved until it does. But if human life was prioritized over moneyed entrenched interests, we wouldn't have the ceaseless death toll we currently have on the roads.
And I'd like to note that there's a shit ton of stuff that can be done immediately (lower speed limits! stronger enforcement! properly funding transit!) and on the way to any autonomous vehicle future (building cities that aren't car dependent! improving existing streets for all the non-car users!) that will save lives whether we get perfect autonomous drivers or just shitty humans.
posted by Homeboy Trouble at 3:52 PM on November 1, 2018 [25 favorites]
California DMV requires AV test licensees to put out a report every year on the performance; amongst other things, how far test cars are driving autonomously and how many "disengagements" there are. A disengagement is when the autonomous control system is, well, disengaged and the safety driver takes over. This could be because the driver thinks it's about to do something dangerous, because the system doesn't know what to do, or because of some sort of technical failure, like a computer rebooting.
The first six months that Waymo (back then Google Auto) tested their vehicles on California roads, Sep 2014 to Feb 2015, they had a disengagement every 700 miles or so. That's the equivalent of over 1,400 disengagements per million miles. One year later, their disengagement rate had improved to a disengagement about every 5000 miles, which is the equivalent of around 200 disengagements per million miles. Seven times better in the first year!
The most recent six months available is through November 2017, so that's almost two more years for improvement. Waymo's disengagement rate is no longer one every 5000 miles, it's... actually, it's still about one every 5000 miles. A really good November pushed them up to one every 6000 miles. With all of the resources of Google at their back, the smartest engineers in the world were not able to improve substantially in two years, and they're still around 200 disengagements per million miles. (And Waymo is streets ahead of their competitors.) I'm eagerly anticipating the 2018 reports to see whether there has been actual progress again, or whether the technology is plateaued.
To compare 200 disengagements per million miles, there's a subset of drivers that have approximately 10 collisions per million miles -- drivers with a blood alcohol content of 0.08. The superset of all drivers, sober, sleepy, stoned, distracted, young, old and otherwise, has a little less than 4 collisions per million miles. I haven't had time to calculate what the rate for a baseline driver -- a 25 to 70 year old who is not drunk, tired or texting but it's obviously much less, and that's clearly the standard we should be setting.
In AI circles, there's a thing that happens on the regular since the early 1970s called "AI Winter" where a bunch of progress is made, and ambitious predictions are made, but then it turns out that it's just not possible to get there, and the field quiets down for a while. I wouldn't be surprised if we're there with current AI techniques -- turns out our current machine learning approaches solve some problems, but just can't get there on others.
I'd like to think that AV development will either soon break this barrier so it's not stuck 20 times worse than a drunk driver, or that the technology is shelved until it does. But if human life was prioritized over moneyed entrenched interests, we wouldn't have the ceaseless death toll we currently have on the roads.
And I'd like to note that there's a shit ton of stuff that can be done immediately (lower speed limits! stronger enforcement! properly funding transit!) and on the way to any autonomous vehicle future (building cities that aren't car dependent! improving existing streets for all the non-car users!) that will save lives whether we get perfect autonomous drivers or just shitty humans.
posted by Homeboy Trouble at 3:52 PM on November 1, 2018 [25 favorites]
i love driving, and will continue until i'm no longer able to (currently 50 - plenty of driving years left).
what i'm most looking forward to is making my morning commute, and then sending my car out as a lyft driver all day, and then have it come back for me at quittin time. this is an ideal scenario because it lightens the load on congested parking, while also cutting down on the actual number of cars on the planet. people who like driving can keep driving, and those that are happy to never drive will never need to.
posted by rude.boy at 3:56 PM on November 1, 2018 [3 favorites]
what i'm most looking forward to is making my morning commute, and then sending my car out as a lyft driver all day, and then have it come back for me at quittin time. this is an ideal scenario because it lightens the load on congested parking, while also cutting down on the actual number of cars on the planet. people who like driving can keep driving, and those that are happy to never drive will never need to.
posted by rude.boy at 3:56 PM on November 1, 2018 [3 favorites]
That's when two or more kids run out in front of a self-driving car from different angles to see which one it decides to run over.
I thought that was already called "Russian Roulette".
I'm another person who actively enjoys driving and I'll continue to do so as long as I can. But I dread getting to the age when I can no do so safely, so I'm also hoping that by the time I reach that point a workable alternative will be available. While I'm absolutely wary of the "buggy software" aspect of the current state of driverless cars, I'm not sure how best to get that process going.
posted by Greg_Ace at 4:22 PM on November 1, 2018 [1 favorite]
I thought that was already called "Russian Roulette".
I'm another person who actively enjoys driving and I'll continue to do so as long as I can. But I dread getting to the age when I can no do so safely, so I'm also hoping that by the time I reach that point a workable alternative will be available. While I'm absolutely wary of the "buggy software" aspect of the current state of driverless cars, I'm not sure how best to get that process going.
posted by Greg_Ace at 4:22 PM on November 1, 2018 [1 favorite]
Thanks for all your responses, I'm not pulling legs, twas was an honest question, and for the record, I'm 50/solid GenX through and through.
OK I'll give you the "I have a long commute and I could be doing other things" and not assume everyone is like me, but I'd far prefer we invest in better rail systems. Driverless cars seems too much like doubling down on the wrong thing. That way you could put your phone or laptop away and talk to an honest to god person if you so desired (remember that?).
posted by BigBrooklyn at 4:23 PM on November 1, 2018 [1 favorite]
OK I'll give you the "I have a long commute and I could be doing other things" and not assume everyone is like me, but I'd far prefer we invest in better rail systems. Driverless cars seems too much like doubling down on the wrong thing. That way you could put your phone or laptop away and talk to an honest to god person if you so desired (remember that?).
posted by BigBrooklyn at 4:23 PM on November 1, 2018 [1 favorite]
I'd far prefer we invest in better rail systems
The investment is coming from a different location. Google isn't taking money away from public transit and spending it on driverless cars instead. Your city/county can continue to fund public transit as much as it wants.
posted by It's Never Lurgi at 4:37 PM on November 1, 2018 [4 favorites]
The investment is coming from a different location. Google isn't taking money away from public transit and spending it on driverless cars instead. Your city/county can continue to fund public transit as much as it wants.
posted by It's Never Lurgi at 4:37 PM on November 1, 2018 [4 favorites]
Driverless cars seems too much like doubling down on the wrong thing.
No, its literally that drivers (all sorts of drivers! You might be one of them and not even know!) are often terrifyingly bad at it but want to continue driving regardless of the impact on other people.
Proper public transport can solve a hell of a lot of issues, but it can't solve the really shitty driver who wants to keep driving.
posted by threetwentytwo at 4:39 PM on November 1, 2018 [2 favorites]
No, its literally that drivers (all sorts of drivers! You might be one of them and not even know!) are often terrifyingly bad at it but want to continue driving regardless of the impact on other people.
Proper public transport can solve a hell of a lot of issues, but it can't solve the really shitty driver who wants to keep driving.
posted by threetwentytwo at 4:39 PM on November 1, 2018 [2 favorites]
Driving is practically a Dunning-Kruger test case: do you think you are a good driver, Y/N?
...and ~40% of the people answering that question will be wrong.
posted by aramaic at 4:43 PM on November 1, 2018 [1 favorite]
...and ~40% of the people answering that question will be wrong.
posted by aramaic at 4:43 PM on November 1, 2018 [1 favorite]
Even if they can only tackle highway driving that would still make both my commute and my job a heck of a lot more pleasant, so I'm all for this stuff. Shame that it'll probably hit luxury SUVs about fifteen years before it trickles down to compact cargo vans like the one I drive.
I'm also just not convinced that we aren't massively underestimating the difficulty of this problem set, but that's a separate issue.
posted by Anticipation Of A New Lover's Arrival, The at 4:54 PM on November 1, 2018
I'm also just not convinced that we aren't massively underestimating the difficulty of this problem set, but that's a separate issue.
posted by Anticipation Of A New Lover's Arrival, The at 4:54 PM on November 1, 2018
I mean, try taking a 32-foot extension ladder onto the subway, sometime. See where that gets you.
posted by Anticipation Of A New Lover's Arrival, The at 4:56 PM on November 1, 2018 [4 favorites]
posted by Anticipation Of A New Lover's Arrival, The at 4:56 PM on November 1, 2018 [4 favorites]
My big concern is that the big corporations with lots of money will find ways to blame everything but themselves when things go wrong (as we've seen many times with manufacturers and mechanical/design problems). And they'll push to make the roads better for cars rather than the cars better for the roads (the 21st century version of jaywalking laws that make it harder to be a pedestrian or having automobile laws applied to cyclists).
posted by kokaku at 4:57 PM on November 1, 2018 [2 favorites]
posted by kokaku at 4:57 PM on November 1, 2018 [2 favorites]
The investment is coming from a different location. Google isn't taking money away from public transit and spending it on driverless cars instead. Your city/county can continue to fund public transit as much as it wants.
Or, doesn't want?
I want to say something about people needing to get behind efforts that are in their own best interests, or the need for tech companies to work on actual better societal solutions (a value Google should have). I'll just leave it to your imagination, because I need to go make dinner, and that's about as far as it will get anyway.
posted by BigBrooklyn at 5:29 PM on November 1, 2018
Or, doesn't want?
I want to say something about people needing to get behind efforts that are in their own best interests, or the need for tech companies to work on actual better societal solutions (a value Google should have). I'll just leave it to your imagination, because I need to go make dinner, and that's about as far as it will get anyway.
posted by BigBrooklyn at 5:29 PM on November 1, 2018
Is it actually true that drivers aren't able to reliably identify whether they are better or worse (that is, safer or less safe)? It seems like they should be able to.
posted by value of information at 7:03 PM on November 1, 2018
posted by value of information at 7:03 PM on November 1, 2018
Even if they work, will it really solve anything? You could have cars running around with nobody in them to pick up the kids. You could have cars circling the block for hours looking for a parking space, with nobody in them.
I have lived in urban and suburban areas and even have experience in very rural areas. The suburban area I most recently lived in was ruined by Waze when it started to redirect freeway traffic onto residential streets. Ironically, this town once had a trolley and was on a rail stop that no longer exists.
The transit we need will never come from personal or shared-ride vehicles, self-driving or not. It will come from better planning and better infrastructure.
posted by sjswitzer at 7:49 PM on November 1, 2018 [2 favorites]
I have lived in urban and suburban areas and even have experience in very rural areas. The suburban area I most recently lived in was ruined by Waze when it started to redirect freeway traffic onto residential streets. Ironically, this town once had a trolley and was on a rail stop that no longer exists.
The transit we need will never come from personal or shared-ride vehicles, self-driving or not. It will come from better planning and better infrastructure.
posted by sjswitzer at 7:49 PM on November 1, 2018 [2 favorites]
mean, try taking a 32-foot extension ladder onto the subway, sometime. See where that gets you.
Someone’s obviously never lived in New York
posted by Automocar at 8:16 PM on November 1, 2018 [1 favorite]
Someone’s obviously never lived in New York
posted by Automocar at 8:16 PM on November 1, 2018 [1 favorite]
"I have been nearly hit on multiple occasions and at one point literally driven off the road - Sequence ".
Me too, many times - worse I've had three drivers deliberately change sides of the road and drive at me, forcing me off the road (in NZ, US and Scotland - always rural areas). I was going to say but that won't happen with self-driving cars, but then I thought hacking - just how temping would it be to play a real game of GTA by hacking a car? How difficult would it be? Just a matter of time I would think.
posted by unearthed at 8:34 PM on November 1, 2018
Me too, many times - worse I've had three drivers deliberately change sides of the road and drive at me, forcing me off the road (in NZ, US and Scotland - always rural areas). I was going to say but that won't happen with self-driving cars, but then I thought hacking - just how temping would it be to play a real game of GTA by hacking a car? How difficult would it be? Just a matter of time I would think.
posted by unearthed at 8:34 PM on November 1, 2018
Re making roads better for the SD cars rather than the people
- a different case maybe but in NZ robotic cow milking is becoming very common. An issue is where a number of cows do not handle the transition at all well - rather than tuning the robots (or reconsidering the whole idea) farmers just cull the 'recalcitrant ' animals. There's no discussion on what the effects are of removing this genepool. What about people who don't handle driverless cars? The visually impaired, aged, handicapped, wandering drunks...
posted by unearthed at 8:42 PM on November 1, 2018
- a different case maybe but in NZ robotic cow milking is becoming very common. An issue is where a number of cows do not handle the transition at all well - rather than tuning the robots (or reconsidering the whole idea) farmers just cull the 'recalcitrant ' animals. There's no discussion on what the effects are of removing this genepool. What about people who don't handle driverless cars? The visually impaired, aged, handicapped, wandering drunks...
posted by unearthed at 8:42 PM on November 1, 2018
That's all views of well-marked, clear roads, so here's a short video promoting Ford’s Industry first autonomous vehicle tests in snow, posted in Jan. 2016
This is going to sound all no true scotsman but geez that is just barely snow; I mean you can still see the lines on the road thru the snow cover. You would have thought for a promotional video they could have cobbled together some footage of an actual snow covered road.
posted by Mitheral at 10:29 PM on November 1, 2018
This is going to sound all no true scotsman but geez that is just barely snow; I mean you can still see the lines on the road thru the snow cover. You would have thought for a promotional video they could have cobbled together some footage of an actual snow covered road.
posted by Mitheral at 10:29 PM on November 1, 2018
In Mountain View, CA the youth have already figured out that fake-charging into a road will cause Waymo cars to emergency brake. I've seen it several times.
The Waymo cars are very polite around bicyclists, which is nice.
posted by pdoege at 11:14 AM on November 2, 2018
The Waymo cars are very polite around bicyclists, which is nice.
posted by pdoege at 11:14 AM on November 2, 2018
On one hand, yes, they're putting these systems on the road, and they're buggy and prone to problems.
On the other hand, closed testing really doesn't get you very far AT ALL in solving all kinds of critical problems in autonomous vehicle technology, and if you don't road-test in real environments, then you'll never get to a point where the tech is solidly performing to expectations and ready for market.
So that's a Catch-22 to work through. Personally I don't have big concerns with encountering autonomous vehicles around the Bay Area. My experiences with them have been that they're far more reliable, cautious, alert, and behaved than most human drivers. And marijuana recently got legalized around here, so there are more unfocused, loopy drivers these days than ever before. But yeah, the tech is still primitive in lots of areas - alarmingly so, maybe, to those with some real insight into the industry.
Does anyone really WANT driverless cars? Maybe its generational, but I enjoy driving. And where else is that time going to go? To people's phones no doubt. Do we really need this?
The thing I'm cautiously looking forward to is autonomous mobility on demand (AMoD) services, as someone who lives in a dense urban area and owns a car but doesn't drive it a whole lot. I can see this replacing both car ownership and the current Lyft/Uber model for me.
It is estimated that Mobility on Demand (e.g. Uber, Lyft) costs ~$2 / mile, car ownership ~$0.8 / mile, and AMoD will be ~$0.3 / mile.
Part of this is because you're able to optimize and micromanage not just routes, but the movement and placement of entire fleets of vehicles. Of course, this translates to a lot of human drivers being out of jobs. But Lyft/Uber are badly exploiting their human drivers anyway, and the "gig economy" is effectively deeply unsustainable from both business and regulatory angles.
posted by naju at 2:23 PM on November 2, 2018
On the other hand, closed testing really doesn't get you very far AT ALL in solving all kinds of critical problems in autonomous vehicle technology, and if you don't road-test in real environments, then you'll never get to a point where the tech is solidly performing to expectations and ready for market.
So that's a Catch-22 to work through. Personally I don't have big concerns with encountering autonomous vehicles around the Bay Area. My experiences with them have been that they're far more reliable, cautious, alert, and behaved than most human drivers. And marijuana recently got legalized around here, so there are more unfocused, loopy drivers these days than ever before. But yeah, the tech is still primitive in lots of areas - alarmingly so, maybe, to those with some real insight into the industry.
Does anyone really WANT driverless cars? Maybe its generational, but I enjoy driving. And where else is that time going to go? To people's phones no doubt. Do we really need this?
The thing I'm cautiously looking forward to is autonomous mobility on demand (AMoD) services, as someone who lives in a dense urban area and owns a car but doesn't drive it a whole lot. I can see this replacing both car ownership and the current Lyft/Uber model for me.
It is estimated that Mobility on Demand (e.g. Uber, Lyft) costs ~$2 / mile, car ownership ~$0.8 / mile, and AMoD will be ~$0.3 / mile.
Part of this is because you're able to optimize and micromanage not just routes, but the movement and placement of entire fleets of vehicles. Of course, this translates to a lot of human drivers being out of jobs. But Lyft/Uber are badly exploiting their human drivers anyway, and the "gig economy" is effectively deeply unsustainable from both business and regulatory angles.
posted by naju at 2:23 PM on November 2, 2018
OK Automocar, but even if you're serious (which… have you seen a 32-footer? I get that New Yorkers are ingenious, but…) I have five more ladders in and on my van, and that's not even talking about any of the rest of my gear. My point is just that some of us are on the road because we have to be, and spending five hours a day driving (especially in traffic!) sucks, and if I could just kind of chill and maybe get some paperwork done rather than having to constantly make sure that I didn't accidentally kill someone with my company's rolling billboard, it would have a significant positive impact on my overall quality-of-life.
posted by Anticipation Of A New Lover's Arrival, The at 7:29 PM on November 2, 2018
posted by Anticipation Of A New Lover's Arrival, The at 7:29 PM on November 2, 2018
"Unbundling The Autonomous Vehicle" Who's Who, What's What in the Underlying Technology
also btw!
Elon Musk: The Recode interview
also btw!
Elon Musk: The Recode interview
Which one of them, do you think, is the furthest ahead or closest to you all?and fwiw :P
Self-driving, maybe Google, Waymo? I don’t think anyone is close to Tesla in terms of achieving a general solution for working on —
Overall solution.
Yeah. Yeah. You can definitely make things work like in one particular city or something like that by special-casing it, but in order to work, you know, all around the world in all these different countries where there’s, like, different road signs, different traffic behavior, there’s like every weird corner case you can imagine. You really have to have a generalized solution. And best to my knowledge, no one has a good generalized solution except ... and I think no one is likely to achieve a generalized solution to self-driving before Tesla. I could be surprised, but...
So none of the car companies. None of the car companies.
No.
Do you ever look and go, “Okay, that’s interesting what they’re doing there.”
The other car companies ... I don’t wanna sound overconfident, but I would be very surprised if any of the car companies exceeded Tesla in self-driving, in getting to full self-driving.
You know, I think we’ll get to full self-driving next year. As a generalized solution, I think. But that’s a ... Like, we’re on track to do that next year. So I don’t know. I don’t think anyone else is on track to do it next year.
What I want to get at is why you’re doing that. It’s not a trivial … Why do you think you want to push yourself that hard?posted by kliuless at 8:22 PM on November 2, 2018
Well, the other option would have been, Tesla dies.
Right.
Yeah. Tesla cannot die. Tesla is incredibly important for the future of sustainable transport and energy generation. The fundamental purpose, the fundamental good that Tesla provides is accelerating the advent of sustainable transport and energy production.
Which I think most people credit you for doing. Pushing everyone else into it at the same time, correct?
Yes. The success of Tesla is, by far, the biggest forcing function for the other carmakers to get into into electric cars. They’ve said so.
No, there’s no question. I was just having a discussion with someone the other day, and I said, “He has pushed everybody into this, really dramatically. There wouldn’t have been this much investment. There wouldn’t have been this.”
Yes. It’s very important for the future of the world. It’s very important for all life on Earth. This supersedes political parties, race, creed, religion, it doesn’t matter. If we do not solve the environment, we’re all damned.
And this way via sustainable transportation.
Yes. It sort of blows my mind, all these social justice warriors driving around in diesel cars. It’s outrageous.
You’re doing this to yourself because you think that the world depends ... Not the fate of the world. You’re not a cartoon character.
No, I think the electrification of transport, and there’s also an important part of Tesla which is solar and stationary batteries, because you need to generate electricity in a standard, sustainable way with solar and then store it at night when the sun goes down with batteries, and then use that energy from the sun to power cars. Without Tesla, this would still happen. There would still be a transition to sustainable energy, but it would take much longer. History will judge this, obviously, but I would say on the order of 10 years, maybe 20 years.
So, pushing it forward by that much.
Yes. I think it’s probably fair to say that Tesla has advanced sustainable energy by at least five years, conservatively, and maybe closer to 10, and then if we continue to make progress, we might advance it by 20 years. This could be all the difference in the world.
Uber Wants To Resume Self-Driving Car Tests On Public Roads
posted by octothorpe at 4:26 AM on November 3, 2018
posted by octothorpe at 4:26 AM on November 3, 2018
Driverless cars to become brothels of the future, replacing pay-by-the-hour hotels
posted by homunculus at 2:17 PM on November 23, 2018
posted by homunculus at 2:17 PM on November 23, 2018
« Older Flasher Music Video | “life is suffering” Newer »
This thread has been archived and is closed to new comments
posted by kokaku at 1:47 PM on November 1, 2018 [5 favorites]