The New Federal Safety Guidelines For Self-Driving Cars Are Too Vague... And States Are Already Making Them Mandatory
from the this-probably-isn't-such-a-good-idea dept
The National Highway Traffic Safety Administration earned plaudits from across the tech sphere for its recently released safety guidelines for self-driving cars.With the NHTSA looking to offer guidance to this emerging industry, the agency issued a set of rules that largely just asks manufacturers to report on how they were following the guidelines. The 15-point checklist is vague in quite a few details, but that isn't necessarily a tremendous problem so long as the standards remain voluntary, which they purport to be. To many, this approach struck a good overall balance between oversight and flexibility.
Regulatory ambiguity can, however, turn out to be a real nightmare with standards that are mandatory. Vague rules can leave even the best-intentioned firms at a loss as to how to proceed. Given how much of a premium consumer confidence will be in a market as revolutionary and potentially transformative as autonomous vehicles, it's crucial that manufacturers comply with whatever standards the federal government promulgates.
That's why it's essential to pay close attention to an underappreciated part of the NHTSA guidelines -- the opportunities they afford federal regulators to coordinate with the states on oversight that, in practice, will be anything but voluntary. Indeed, the early signs from the first of what will be many proposed state rules to follow in the wake of the NHTSA guidelines suggests that compulsory standards are exactly what we're going to get.
First up are proposed rules from the California Department of Motor Vehicles, recently revised in response to the NHTSA guidelines. The revised draft of California's model regulations is far more permissive than the original version the agency promulgated late last year, a set of changes that were celebrated by various observers, even me.
But delve closely into the updated DMV proposal, and you'll find a requirement that manufacturers obtain a state permit certifying that any and all vehicle tests are conducted in accordance with the NHTSA’s "Vehicle Performance Guidance for Automated Vehicles." Thus, in the nation's largest testing jurisdiction, the NHTSA standards already are set to be made mandatory.
This is not to say the federal government doesn't have a rule to play in oversight of self-driving cars. The feds are better situated to oversee the development of safety standards, and the door should be open to refine those standards. But coordinating with the states to turn those standards into a set of de facto binding obligations smacks of underground rule-making.
The California DMV might be complicit in this collusion, but it can't be faulted for deferring to federal authority. Were the NHTSA’s safety standards clearer -- an undertaking that presents risks and problems of its own -- California’s approach wouldn't actually be a problem. The fact that the federal guidelines are so vague in so many of the details means that we can't really know either that manufacturers will be able to comply with California's rules or that the state will be able to enforce them.
For now, state regulators should use their discretion to be as liberal as possible about what sorts of vehicle testing comports with the NHTSA safety guidelines. Over the longer term, what we need is for states like California to communicate to the NHTSA that it's up to them to make absolutely clear what does and does not count as compliance.
It's broadly understood how overly restrictive regulations can dampen innovation, but regulatory ambiguity can be just as bad. For regulators, the clock is ticking. It's up to both the NHTSA and state agencies like the California DMV to bring the clarity this new market needs.
Thank you for reading this Techdirt post. With so many things competing for everyone’s attention these days, we really appreciate you giving us your time. We work hard every day to put quality content out there for our community.
Techdirt is one of the few remaining truly independent media outlets. We do not have a giant corporation behind us, and we rely heavily on our community to support us, in an age when advertisers are increasingly uninterested in sponsoring small, independent sites — especially a site like ours that is unwilling to pull punches in its reporting and analysis.
While other websites have resorted to paywalls, registration requirements, and increasingly annoying/intrusive advertising, we have always kept Techdirt open and available to anyone. But in order to continue doing so, we need your support. We offer a variety of ways for our readers to support us, from direct donations to special subscriptions and cool merchandise — and every little bit helps. Thank you.
–The Techdirt Team
Filed Under: autonomous vehicles, nhtsa, regulations, self-driving cars, voluntary
Reader Comments
Subscribe: RSS
View by: Time | Thread
http://www.usatoday.com/story/money/cars/2016/10/05/nhtsa-traffic-deaths-self-driving-cars-vehicles/ 91597722/
Look the idea is great. I got a few words on why it will never become a reality...
Porsche, Ferrari, Motorcycles. I also wonder if these autonomous cars will obey the speed limit. You really think Americans are just going to sit there doing 55? Good luck with that. In addition no one mentions snow/ice. It's a pipe dream.
Driver assist will become popular but totally autonomous? Not likely in our lifetime.
[ link to this | view in chronology ]
Re:
The story you linked to says "no traffic deaths" which is a very different thing. And 30 years in the future.
> In addition no one mentions snow/ice.
Self-driving cars just have to beat humans, who aren't great at dealing with adverse weather.
[ link to this | view in chronology ]
Re:
If you've paid any attention to how rapidly innovation has happened in this area over the past decade, you'd realize how ridiculous that sounds. We're practically at fully autonomous vehicles today and there will be more and more on the road basically every day. I'd bet that within 10 years, greater than 20% of the vehicles on the road will be autonomous (and I consider that a conservative guess). That's well within our lifetimes.
[ link to this | view in chronology ]
Re: Re:
I predict that the biggest stumbling block to getting autonomous cars on the road will be the price. Based on nothing more than my own intuition, I can easily see the first wave of autonomous cars costing well over $100,000. At such a price, the average middle class family isn't likely to buy one. In fact, the majority of people won't be able to afford one. Even after a few years when the price comes down, it will probably never fall below $50,000-60,000, which is still out the range of most people. Sure, there's the used car market, but how many people are going to want an outdated self-driving car when the newer ones are likely to have more features and be safer? I mean, how many people today want an original iPhone?
[ link to this | view in chronology ]
Re: Re: Re:
Nah. It'll be much cheaper. You already have comma.ai getting ready to launch its autonomous add-on for $1000 and $24/month: http://www.theverge.com/2016/9/14/12912668/comma-ai-george-hotz-self-driving-car-price
That's not fully autonomous, but it's pretty good, and George is definitely planning to go much more autonomous and keeping things cheap. It's not going to be nearly that expensive.
[ link to this | view in chronology ]
Re: Re: Re:
http://www.fool.com/investing/2016/08/04/how-much-do-driverless-cars-cost.aspx
[ link to this | view in chronology ]
Re: Re: Re: Re:
Since when has actual cost had any bearing on sale price?
[ link to this | view in chronology ]
Re: Re: Re:
Fully autonomous cars would be everywhere and less expensive than traditional cars if not for a couple of things:
1) Non-autonomous vehicles. The biggest challenge for engineers is trying to predict the stupid humans that don't follow traffic rules. Nearly every accident on the road involving an autonomous vehicle has been because a human did something dumb or illegal.
2) Allowing humans to drive the autonomous vehicles. Cars would be far less expensive and much safer if they did not need the traditional controls that we are currently requiring. Think of how much safer a car with no windows would be? Placing humans in safer seating positions would save lives. Removal of the control systems that we have would save weight, space, and money.
[ link to this | view in chronology ]
Regulations
The more government gets involved in self driving cars the more I begin to believe that I'll never ever get to buy one....
[ link to this | view in chronology ]
Re: Regulations
Time and time again, setting higher standards for safety, for fuel efficiency, and for low emissions has spurred innovation in developing ways to safer, more efficient, less polluting cars less expensive and more available to the mass market.
[ link to this | view in chronology ]
Re: Re: Regulations
Hope that the innovation continues.
[ link to this | view in chronology ]
Re: Regulations
[ link to this | view in chronology ]
Population control?
how to say this...
I dont care what you say there are 2 REAL problems..
#1. Who is responsible for this CAR.. Im not driving it. AND IF I CANT FIX IT, Im not going to INSURE IT..
#2. I dont care HOW careful you are, accidents happen. There are to many conditions, from NATURE to roads.. Even IF' you have the ability to take control of the car, while you are sitting back relaxing and something happens...you are not Fast enough to change the outcome..
I can see the computer taking the Cautious speed and protections to SLOW down and control the car on an ice covered road, and the passenger getting upset and taking control..
EXPERIENCE is the greatest thing is car driving.. YOU WILL not hve the experience needed to control your car or Judge your OWN capabilities..
[ link to this | view in chronology ]
that's very interesting what you say. i can see a selfycar driving a family of four right into a tornado. car ain't scared of no tornado.
[ link to this | view in chronology ]
Re:
[ link to this | view in chronology ]
yeah, i wanna see the algorithm that decides what direction to drive to get away from a tornado.
can the car pass between that wall and the trunk of that tree to avoid a runaway truck? i want to see how they plan to make that happen.
[ link to this | view in chronology ]
This makes it easier to understand beyond of foggy dreams (being propagated to be real, right soon ... they aren't, for now it's just projections, or but slow motion automatons, so algorithms may for sure remain in track with what happens).
Take the term of the social part of experience and add to it the common way to arrange things of reality, as it is common for the last decennies (and beyond), this is by exploitation.
Assumed, technical realization is perfect, so there is no stand-alone thing as self-driving cars, on a straight self-understood base. Instead, to get this right: what we have here, is exploitation of a way to share the common ressource, that is, the streets, the know how, how to share this ressource in terms of social behaviour. Expectations about, how to generalize this behaviour of participants in traffic is frozen into algorithms. Now, this is first of all kind of exploitation of behaviour of people, when looking at the cultural episode in history as of right now.
Don't expect such exploitation to translate to being fine for all instances.
Social behaviour stems from the never-ending discurse between humans, propably non-verbial in this case. When frozen into algorithms, this aims at kinds of dictatorship, as the most vivid part is missing, the interaction for further progress from behaviour today into behaviour tomorrow.
This is a slow process, but an essential one. Exploitation here means to abolish real progress. There is no self for robots. Instead, people have to obey absolutely, think as well of once foolish parts in algorithms are out in the wild.
Taking the terms of social behaviour and of exploitation into account here. So, how does this fit into huge structures for standardization processes and similar cultivations, when trying to balance needs and hopes?
You cannot change the underlying basics here, which is linked into historical categores, and exclusively not into mathematical ones.
Fascination is on the driver's side here (and paired with greed, this gets automatted within itself easily), having grown to a self-understood demand, this makes people blind.
Expectation of the whole geography reduced to being nothing but of only virtual existence, and each car is just another living room, the landscape being a movie for to be consumed.
Yes, the plot is very inviting for to forget about the basics.
[ link to this | view in chronology ]
two cars approach from behind and one pulls in front of you, only to begin slowing. your car dutifully slows as well. the car behind crowds in close. every attempt by your car to pass the car ahead is thwarted by that driver and you ultimately come to a complete stop on the dark highway. doors open ahead and behind and men approach your car. they aren't wearing hoods. they won't need them.
one of the men pushes a button on a tablet and your transmission goes into park. we all know what happens when your car is put into park: the helpful doors unlock themselves for you.
[ link to this | view in chronology ]
Re:
Thankfully with today's human-driven cars no-one is ever run off the road, boxed in and forced into an accident for insurance claims, boxed in on the street and robbed or shot. This never happens today and the vague spoooky idea that evil spoooky people will do things to you is the best reason to halt progress.
[ link to this | view in chronology ]
The cars need a lot more security features before I would risk my life getting into 1.
[ link to this | view in chronology ]
Re:
Yet you're fine getting into a car driven by a human who has many more problems and is much more likely to put your life at risk? At this point, self-driving cars have driven millions of miles with a much lower accident rate than human drivers. In fact, most of the accidents have been caused by other human drivers. The only accident that Google has announced was its own cars' fault was a minor brush with a bus, the kind that happens many times a day with human drivers.
So I'm always a bit confused by these claims. Do self-driving cars need to get better? Sure. But they're incredibly good today. Almost certainly better and safer than human drivers. So what's the fear?
[ link to this | view in chronology ]
Re: Re:
Weekly/monthly actually take to a garage for a security update if not connected to net?
Drive-by hack attack?
HP-like delayed release update?
Sony-like remove feature update?
I'm just guessing, not the op.
[ link to this | view in chronology ]
re: sony feature removal
I also don't want Bing to tell me where I have to buy something I searched for last week.
[ link to this | view in chronology ]
self-driving vehicles
The issue is that the radar units that all the current truck manufacturers use crashes at least once a day or sees a overpass or blowing trash across the road as a solid object and locks the brakes up to avoid hitting it. Or the software for the unit locks up and then cruise and any speed over 55 is locked out till the unit reboots which means sitting parked for 2 hours till the onboard computers shut down and can be forced to restart. I'm all for the eventual goal of completely self driven vehicles however until the software and the programmers and engineers all get everything working properly, there are always going to be issues that are going to cause accidents and/or property damage.
[ link to this | view in chronology ]
Re: self-driving vehicles
It would be unfortunate for someone to die because of a dumb software error, but people die to day for doing lots of dumb things while driving, or because other people do dumb things while driving.
There will always be accidents and property damage. The goal isn't to eliminate them. The goal is to reduce them, drastically.
[ link to this | view in chronology ]
The states generally get around this by making compliance with the tougher state standards a requirement for registration, sale, or transfer of a vehicle. So while you could drive a non-California emissions legal car in the state, you couldn't plate it, sell it, or buy it (except for off road use only).
What the state of California is doing here is pretty simple: They are pointing to a standard being set by the feds and saying "we agree". Nothing much more than that. The rules are vague because it's an evolving situation, not a hard and fast situation. As jupiterkansas points out, that flexibility in the rules gives space for innovations and alternate approaches. Can you imagine for a moment the rules for autonomous cars being set to the standards of the first examples a few years back, and limited to only that technology and approach? It would be a disaster. The space in the rules and guidelines are about allow the technology to flourish.
It's much better to set a series of objective tests as the baseline goal, and see how different companies get to those goals.
[ link to this | view in chronology ]