Autonomous truck cleared to drive on US roads for the first time
newscientist.com
newscientist.com
These trucks can and will put people out of work. Right now a large majority of the testing of autonomous vehicles is focused on "perfect environments". Yes, Google will say that test driving autonomous cars in San Francisco is far from a perfect environment; however, in a future where people are loosing their livelihoods, ways to hack autonomous vehicles will be explored, and I'm not talking about just hacking a vehicle CAN bus. All of these vehicles rely on technologies like radar, LIDAR, IR vision, etc. None of which are fool-proof and each of which could fall to very specific attacks. I hope that the auto manufacturers are considering this future. After all, we have technologies today like GNU Radio and the HackRF One which have brought what was once the domain of the nation state to the general public.
These days it's sabotage against GMO crops. You'll find Monsanto and other companies will place their GMO test fields in regions of the country known to be sympathetic and/or will keep them under armed guard.
I don't quite like that description either. I think it's more that they want to encourage interdependence within the community. They are not opposed to technology that they believe doesn't threaten that community dynamic.
If the number of truck drivers goes down by 90% in the next 80 years, that would be fine. If it goes down by 90% in the next 8 years, then we'll have a lot of mis-trained ex-truck drivers out there, and that's nothing to scoff at.
Farmers transitioning is a bad analogy. People live longer, require more education, and have less control over their work life if they aren't tech-savy.
I see 4 possibilities. We have an aging work population 65 and older working entry level skilled jobs to pay off college loans they got later in life. Or we have a larger "welfare" state, we have a better "public higher education" program, or we have more millennials taking care of their boomer parents because their parents couldn't find work, and social security hasn't kicked in yet.
You're saying that as if it were a bad thing.
http://www.marketwatch.com/story/no-truck-driver-isnt-the-mo...
Efficiency is celebrating the fact that you can drop your expensive annual costs and the liability of employing humans to conduct your business. People are expensive and demand things like rights and benefits. Machines can be had with a purchase/lease and maintenance costs.
Machines won't negotiate or get a lawyer. That's efficiency.
Meanwhile, we'll strip the earth to build these machines in the name of efficiency while more of the population is made useless and left to rot by an economy that functions without them.
Specifically why is this time different?
It was the most common occupancy in 29 states in 2014. http://www.npr.org/blogs/money/2015/02/05/382664837/map-the-...
Who is the us "efficiency" has helped? At the end of the day, do I benefit from the fact that Company X's truck no longer needs a driver?
Does that benefit really outweigh ruining the livelihoods of millions of people, whose family now can't pay taxes and lost their ability to see doctors or to maintain a roof over their heads?
When do we get to reap the benefits of this automation?
There is a lot of money being made by letting people go and letting the tax payer pay for the societal damage.
Will those millions of drivers be transferred to a new department within the company? Will they get a severance check from the profits of this automation? Or will unemployment and lack of medical care shave 20+ years off their lives so a shipping company can save a dime that society gets to pay for?
Capitalism drives the development of increased efficiency, you cannot have capitalism without that drive, as it creates competitive advantages.
Welfare isn't effective I agree. What do you think about a NIT?
Whenever people CAN exploit (in the name of capitalism, or racial superiority, or any other equally sociopathic concept) they will. Why are we working hard to put our neighbors out of the job in the name of some ideology? It seems a lot like you literally care more about the efficiency of the market than you do giving people a form of sustenance (or income they can barter for sustenance).
Of course it will.
Think of how many in industries related to horses (hay, stalls, cleaning, horseshoes...) were put out of business when the automobile was invented.
Think of the guys that used to drive around and put blocks of ice in your fridge before electricity.
Or countless millions of others that were put out of work when a better/different way was found to do something.
It's fine. Those people will re-train. When there are no more coal plants there will be a lot more people working in the solar/wind business. It's OK, don't panic.
No, people will re-train, but those folks directly put out of work in this instance will not likely turn around and immediately find work as a software engineer maintaining autonomous vehicles. In the long run society retrains, but there is a disruption period that affects real people.
Putting people temporarily out of work is not a reason to not move society forward.
I do not see where I suggested that.
In Denmark the Metro is operated by a computer, but the whole thing is built around this. The next step to me is not a truck, but a "normal" train that still has to interact with the rest of the train network around it. Then we can move on to trams and then trucks and buses.
Besides, operators will probably desire a small crew to watch over their trains anyway for the foreseeable future, regardless of whether a human is in the pilot's seat.
I believe we're fast approaching the need to re-evaluate how we treat (un)employment and the role of individuals with an imposed obligation towards economic contribution, particularly due to the automation and/or efficiency we gain through technology. Will we always have enough work for everyone to sufficiently provide for themselves without "redistribution of wealth" programs? More importantly, if we force everyone to work to provide for themselves, is the work done actually creating value to society? Is it necessary?
"Daimler’s tractor-trailer, called the Inspiration Truck, relies on cameras and radar to guide itself. It does not use LIDAR, a sensor that is being used by other self-driving vehicle makers such as Google. Like all autonomous vehicles today, the Inspiration Truck has some limitations. It can’t drive on its own during heavy rain, snow or crosswinds above 30 or 40 mph.
"Daimler developed a training system for its drivers that Nevada’s department of motor vehicles approved. It’s fairly simple, teaching them how to activate and deactivate the autonomous driving system. It says it picked Nevada because of how thorough its rules are for autonomous vehicles."
A thorough National Public Radio report[2] also describes this as a one-state, limited program of experiments.
[1] http://www.washingtonpost.com/blogs/innovations/wp/2015/05/0...
[2] http://www.npr.org/blogs/alltechconsidered/2015/05/10/405598...
[1] https://en.wikipedia.org/wiki/Dead-man%27s_vigilance_device
Say currently one driver can drive 8 hours per day. So if you need to drive for 24 hours, it's going to take 3 days. If you had 3 lorries doing that journey, it would take 3 days.
But say you had your semi-autonomous road train. Now you could do the journey in 24 hours. One driver drives for 8 hours, then the next guy, then the next guy. The two lorries behind just tag along.
Imagine 3 x 25 hour journeys, where the destinations didn't quite match. Previously this would take 4 days. This time: there's 24 hours of shared journey and an extra unshared hour of travel at the end of the route. Now your drivers have to split up at that point. The guy who was driving the road train last will have to wait 16 hours, so his journey takes 24+17 = 41 hours; the guy who was driving before him will have to wait 8 hours, so his journey takes 24+9 = 33 hours; the other guy can jump in his lorry and go straight away, so his journey takes 25 hours.
Actually, why aren't there fully autonomous freight trains on railways yet? We have driverless commuter trains (Docklands Light Railway) why not the same for long distance freight? The infrastructure is there - automated signalling devices in cab, it doesn't seem too big a leap. One hurdle might be unions, I recall ASLEF or someone got upset about the DLR when it first appeared?
in Australia there are a few automation projects for mining rail.
Perhaps trains would form spontaneously as automated trucks encountered each other? You can't plan it too tightly because one might be late. So it sounds like every truck will still need a driver ready to be the leader or on their own.
Keep in mind that one reason trucks are so much more popular than rail trains is their flexibility. One by one as they're ready and different start and end points each.
Who is responsible is going to be the big make / break for all self-driving vehicles. I get the feeling this is going to be very messy.
I mean smoothing out low-incidence, high-impact events is pretty much their business, no?
If they can know for a fact that a certain accident rate is guaranteed for a given car, then they will explicitly disclaim liability for accidents within that rate, similar to how all product warranties expire before the far end of the bathtub curve.
Otherwise it's like selling flood insurance on a flood plain, or health insurance to cancer patients: something that makes no economic sense and only exists where the government has forced it to.
"Exceptional case" is still a well-defined rate for them. They calculate risks and adjust insurance premium accordingly. That's pretty much their business.
With sufficiently large amount of insured parties those "exceptional events" are actually fairly common.
And that's what I was referring to. If the insurance company is in the business of calculating incident rates anyway they might as well cooperate with the autonomous car manufacturer to drive down incidence rates to a covered level. Just like car insurances will offer reduced premiums to experienced, accident-free drivers they can offer reduced fees to vehicle manufacturers who statistically cause fewer accidents.
The warranty example is bad because warranties end instead of getting renewed. They'd happily renew the insurance for the cost of a new product minus epsilon.
What about insurance? Insurance works with un-correlated risks. You mowing down a pedestrian doesn't cause everyone around you to do the same. This keeps insurance companies' instantaneous expected losses manageable. Coding errors and recalls are not un-correlated. If your car has an error, so does–likely–every other car in its run. This concentrates losses and makes them difficult to insure.
There are solutions. But systems will need to be built to allocate risk.
That is assuming that errors are the main cause of insurable incidents and not simply situations for which the software could not compensate beyond stepping on the brakes in the hopes of minimizing impact energy.
Traffic is chaotic and not fully predictable, so there's always some baseline risk that may exceed the risk of non-catastrophic bugs.
It works out OK for everyone; for insurers, a software error is the liability of a large firm with deep pockets, from whom up-front costs can be recovered, as opposed to irresponsible or inadequate human drivers whose personal assets may not cover the loss. From the software vendors' point of view, actuarial data provides very clear targets to shoot for, as well as a lot of data about the financial cost of failures. It's OK to take cost factors into account from an ethical standpoint as long as they're used to maximize safety by efficiently allocating resources, rather than to minimize production cost.
Insurance responds to the rules of the game. If the rules of the game are that damages from autonomous vehicles are capped, or that multi-vehicle collisions automatically share costs, then the cost of the insurance premium will adjust to that reality.
It takes a leap of thinking to get away from the current obsession with precisely allocating fault in vehicle accidents, but it's entirely possible to do.
Out of all the issues, insurance is the easiest one to solve.
Unlike with human drivers that may not stop if they believe they are in danger of getting carjacked I assume an autonomous system will be required to avoid hitting a human at all cost.
- The driving algorithm will need to be capable of causing a 911 call to be made for various reasons, including that
- When there is no human driver, a carjacking is not actually that bad. There is insurance for this kind of problem, and it will be up to insurance companies to figure out if it's even worth attempting to mitigate.
So the robbers wear masks and shoot / paint over / jam the video equipment. They can do this at a distance because there is no risk of hitting a human.
> - The driving algorithm will need to be capable of causing a 911 call to be made for various reasons, including that
So the robbers choose a section of road far away from the nearest police and otherwise conducive at making them difficult to follow.
> - There is insurance for this kind of problem
If it gets to be too frequent of a problem, the insurance will become expensive enough that it will be easier just to keep a human around on more dangerous routes.
Doing any of that would create a very distinct MO. You want to avoid that if you want to avoid being caught.
Also, have you seen how small cameras are these days? Maybe you could load up a crop-duster with paint and cover the whole truck, but there is no way you could hope to successfully shoot out every camera an insurance company could stick on a truck.
> "So the robbers choose a section of road far away from the nearest police and otherwise conducive at making them difficult to follow."
A problem not made any worse by driverless vehicles. If anything, you can program a truck to always follow approved routes while with human drivers you have to account for insubordination.
Crime is all about risk management. Sloppy purchases could leave a trail, yes, but video evidence can also leave a trail. One trail is easier to cover than the other: lots of people buy paint and guns, not many rob trucks.
> have you seen how small cameras are these days?
This isn't hide-and-seek. The game is not over for the perp if a shitty miniature camera buried in the truck's frame catches a few blurry frames of a masked man. I was responding to the "just do machine learning on LIDAR data" proposal that got thrown out elsewhere in the thread. Equipment that can reliably identify people who control the physical situation and don't want to be identified is probably expensive and vulnerable. Making it relatively invulnerable makes it even more expensive.
> A problem not made any worse by driverless vehicles
Yes, precisely. They never were a deterrent and "just call 911" was/is not going to stop robbers. The deterrent is sitting in the cab.
The sort of MO that I am talking about is "Always shoots out the cameras after stopping the truck, using a rifle" (or whatever the specific details happen to be).
The reason you do not want to establish this sort of MO is that it allows the police/FBI to connect different incidents, different truck hijackings, and figure out that they were all perpetrated by the same gang of criminals.
Once they do this, they can begin analysing where and when you are likely to strike in the future. They can use information such as "This gang was out committing crime on April 15th, June 3rd and July 18th" to narrow down their list of suspects, eliminating people who might be matches for one date/location but not for the others.
> "This isn't hide-and-seek."
When you are talking about shooting out cameras, it really is. I mean, ignoring the hollywood silliness of it, chances are you are going to be in that cameras field of view before you shoot it, and any footage that camera already shot will not be destroyed by the bullet smashing the lense. The LIDAR is really not going to be the primary source of information about the thieves anyway, the cameras will be.
> "They never were a deterrent and "just call 911" was/is not going to stop robbers."
Calling 911 is about catching robbers, not deterring them. The sooner the police arrive on the scene, the more likely the criminals will be caught.
How fast does the average truck hijacking happen? How fast does it happen when the truck that is being robbed has been immobilized? Maybe you can get the trailer off and onto another truck, but that takes time and could be hindered any number of ways. Immobilizing the truck isn't a good idea when there is a human driver that could be harmed, but with a self-driving truck there is no reason why it wouldn't be done.
On the contrary, payouts for a driver taken hostage, injured, or killed on company business would be immense. Insurance companies would probably be quite happy about crimes taking place without humans getting hurt anymore.
Having a human present is a significant deterrent against robbing in the first place and against liberal use of firearms because if things go south with a human present then the charge is murder. A robotic truck presents no such deterrent. Lower the risk of knocking off a truck, and more trucks will get knocked off.
I'm not saying it can't be solved, I'm just saying that the solution won't be "buy insurance."
I expect that insurance companies will eat the cost of hijackings, but push for more effective enforcement of existing laws, leveraging the surveillance capabilities of autonomous vehicles.
You also can't threaten a computer to stop it from immediately calling the police. A human driver might fear for his life and delay contacting the police until the encounter is over, but the computer driver could immediately call the police the moment it thinks it is being hijacked.
As others said, the truck might call 911 before it has come to a stop. Also, 911 operators might get a remote override switch that allows them to instruct the truck to take some evasive action.
I do expect additional laws in the area of autonomous cars, but not to prevent robberies. If autonomous cars become really, really, good and ubiquitous, there is no reason for pedestrians to watch out for cars when crossing a street. That will lead to some funny, but disruptive (for the flow of road traffic) videos on YouTube (yes, you can play Frogger on a 8-line highway with your eyes closed and survive. The crocs aren't robots yet, though)
I wouldn't be so sure about that. I don't know about the US, but here in Germany the autobahn is supposed to be off-limits for people (and vehicles with a design speed below 60km/h for that matter). This is obviously to allow drivers to drive with the assumption that they don't have to be ever-vigilant about cyclists or children crossing the street.
Autonomous vehicles will likely operate on the same assumptions to behave more optimally given the constrained environment, i.e. drive in a more aggressive manner than they would do in a city.
I.e. strolling at moderate pace over a highway would probably have about the same risk as hiding ducked behind a parked car to jump in front of a bus in the city. Both are situations that no sane person should engage in and both are considered acceptable risks not worth defending against.
I'd be surprised if autonomous vehicles are allowed to drive faster than manual vehicles any time soon.
There are significant efficiency gains that can be obtained on highways and for road convoys.
And I expect that what constitutes an acceptable risk will go down and down, as long as technology allows for it. It's already happening al the time. Seat belts are compulsory, air bags are compulsory, ABS is compulsory, etc.
But yes, the frog get thing is not something I would advise to do. On the other hands certainly, in cities, it will be much, much safer for pedestrians to step out into traffic without paying much attention to that traffic. That will affect how pedestrians behave.
Well, I do think that the risk of pedestrians on the autobahn isn't something we optimize for.
In the city? Yes.
Reducing the risks of occupants? Yes.
Driving defensively on the autobahn because someone might walk into the traffic? No.
The potential optimizations for the restricted environment that will be made there will be for increased traffic flow, more flexible lane switching or things like that. Maybe even saving fuel by driving in the tailwind of another vehicle.
The laws of physics will still impose lower limits on stopping distances, and pedestrians who willfully cause abrupt stoppages resulting in juries or property damage will have the book thrown at them - by that stage most vehicles are likely to have multiple cameras built in that will help identify the perpetrators of such acts.
Also, as pointed out elsewhere, vehicles will probably call 911 automatically as soon as the probability of an accident spikes.
Unless a whole group of people starts crossing the road all at the same time and holding hands (in which case, well, you can't stop people from committing mass suicide I guess), even a very rapidly moving car should be able to swerve out of its way to avoid a pedestrian walking across a motorway, so long as it detects said pedestrian at a sufficient distance.
If you assume that the self-driving cars talk to each other, then each has access to the sensors of the cars far ahead in front, so even in busy traffic they should be able to detect a pedestrian standing by the side of the motorway and start adjusting their flow accordingly, well ahead of any drunken stumbling by said pedestrian.
Unlike human drivers, cars can adjust their trajectories at high speed with minimal risk of accident.
... assuming perfect weather and perfect street surface conditions.
The implication is that if some of our roads gradually become devoid of humans, robbing automated vehicles will become a "thing". It doesn't actually sound that far fetched, really. It would presumably be safer for the crooks than robbing a real truck today for three reasons: 1) they wouldn't have to threaten a human (thus lowering the stakes if they get caught), 2) the trucks likely would not be built to shoot back (think of the potential liability for the trucking company if it were), 3) the trucking companies would likely not put much energy into pursuing the criminals, so long as the losses were small in the grand scheme of things.
It's similar to the notion of people shooting down delivery drones and stealing the packages. People will do it, and it will likely be tolerated to some small degree. Even today we already have people stealing packages off doorsteps, although there is occasionally vigilante justice in those cases:
http://noevalleysf.blogspot.com/2012/11/crime-beat-suspected...
Everyone is always a few seconds away from having access to terrible crimes.
The mere introduction of new forms of crime does not necessarily lead to an increase in criminality. Nor does reactive policies to specific criminal narratives necessarily reduce it.
In other words, it's going to happen, just like every other crime already does. Existing infrastructures of human society can easily facilitate this new form of misconduct.
I believe accessibility may increase the frequency someone attempts a crime but not the number of people who will engage in it.
Think about grocery stores ... many leave items outside the store. You're expected to inconveniently walk in and pay for it.
Yet do you see people rampantly ripping off those potted plants or cantaloupes? Although that crime is highly accessible it doesn't happen.
Now if we were to leave cigarettes out there, of course it would disappear. But the question here is whether there would be more thieves or just more thievery from the same individuals?[1]
The theory is that certain people have relationship dynamics to society which in their mind, justifies these actions. This theory supports looting grocery stores during some catastrophic natural disaster. Under this theory, if you can change the dynamics by which someone identifies with society, you'd see fewer issues.
But this is mere speculation. Is there any literature on this?
----
1. In the lens of this theory, one may claim looting cigarettes is part of their terrible nicotine addiction. This unfortunate narrative constitutes an (unjustifiable) reason to steal the item. The idea "I'm a screw up so this is the role I play in society" is a very Zimbardo-prison way of looking at things - but I think it's unfortunately true.
so can an autonomous vehicle. in fact it has to be able to do that to avoid certain kinds of accidents.
The only one of those that is more true of a human driver is the first one.
According to mainstream economics, jobs are not an externality (although they are during a recession). The total amount of money to go around after an efficiency increasing change, will be greater. With taxation, this means that in theory everyone can be made better off. In practice, the truckers will be worse off, but you could at least arrange taxes so that the percentage of people earning less than X decreased, for all X.
Then the only argument left is that the change is somehow unfair to truckers. But that viewpoint unfairly privileges the status quo.
One thing that is true is that the adjustment period will be hard.
That’s no way to go about this. Obviously, not increasing efficiency would be bone-headed, but it is centrally important to create a socially acceptable transition for those (and any people) who are affected by this. That is currently not happening and hasn’t been happening in the past.
The luddites were right. The changes industrialization brought really did suck for them and for them it really didn’t get any better ever, until they died. From that point of view smashing the machines is entirely rational, even if those same machines created tremendous, unimaginable wealth. Obviously I’m not advocating smashing machines here, but, you know, we do have to find workable solutions for the luddites.
I mean, I’m more than happy to consider solutions to this problem that tackle the problem of unemployment as a whole, not just this small part we just considered. I think finding a general solution would definitely be even better and thinking about this problem (e.g. basic income) definitely often tends to consider unemployment as a whole.
My position is that even with the current welfare system, technologies that increase productivity, but make some people unemployed, are still good from a utilitarian perspective. The trickle down effect (including through the existing welfare system), which benefits the entire poor population, makes up for the negative effect on the newly unemployed.
EDIT: as always, I have no problem with downvotes for my other opinions, but it's really sad to see downvotes for economic orthodoxy. If anyone reading this downvoted me, please educate yourself on mainstream economic thinking. You don't want to be remembered as an anti-vaccination advocate.
Taxi driving is one of the jobs many immigrants end up doing in a new town. If we automate the taxi, what do we do with all the taxi drivers?
http://www.npr.org/blogs/money/2015/02/05/382664837/map-the-...
Autonomous with a requirement for a human driver doesn't sound like real progress. Title might be a little misleading.
With the trucks, you have the same person overseeing his own job, except that it'll be done safer. 1 person = 1 job (safer).
Your analogy would make more sense if the trucks were remotely supervised and 1 person could oversee several trucks via remote connection.
How long until they get there? It does not look like a huge change from the person on a wheel.
Automatic cars wouldn't be allowed at the roads in one step.