Senators Want to Sneak Safety Exemptions for Self-Driving Cars Into Law

AV crash

A group of senators led by South Dakota Republican John Thune wants to let companies rush self-driving cars to market before any federal safety standards related to autonomous systems have been drafted.

A coalition of 65 consumer advocacy and street safety organizations has warned against the bill known as AV START, which would preempt state and local safety regulations of self-driving cars without spelling out any federal safety rules (although it would allow U.S. DOT to draft some). In addition, the bill would exempt AVs from many safety standards that apply to all other cars. Each manufacturer would get an allotment of 100,000 vehicles to sell for use on public streets within three years.

In a letter sent to Senators yesterday [PDF], the coalition — which includes the American Public Health Association, America Walks, and the League of American Bicyclists — demands the addition of public safety standards before the legislation is enacted. But Thune, the Commerce Committee chair who represents one of the most rural, least-populated states in the nation, wants to include the AV START language as a rider to the Federal Aviation Administration reauthorization bill, which is viewed as a must-pass. The maneuver would prevent debate and an up-or-down vote on AV START as stand-alone legislation.

The rush to pass AV START before safety standards are enacted is proceeding despite the killing of Elaine Herzberg by a self-driving Uber car in Tempe, Arizona, this spring. A National Transportation Safety Board report revealed that Uber had programmed the car not to brake in certain situations, and that the backup driver was watching TV on her phone at the time of the crash.

Jason Levine, executive director of the Center for Auto Safety, a consumer advocacy group, says the coalition doesn’t oppose self-driving cars but wants to ensure safety protections are in place before laws legalize sales to the public.

“The promise is that these are going to be safer,” he said. “There’s no requirement that these be even as safe as what we have right now. That’s really a problem. “

Even Keith Crain (yes, that Keith Crain), the long-time publisher of Automotive News, thinks lawmakers are shirking their duty to protect the public:

Autonomous vehicles may provide the opportunity to save even more lives and prevent more injuries. But they must also adhere to the same strict standards that exist today.

It is bad enough that automakers are testing these vehicles on public roads. To even think about manufacturing these vehicles for the public without meeting today’s standards is simply irresponsible.

The coalition has proposed nine changes to the AV START legislation, including a mandate that data recorded in crashes be disclosed, and a requirement that AVs pass a “vision test” showing they can process visual information about their surroundings as well as human drivers must demonstrate to be licensed.

41 thoughts on Senators Want to Sneak Safety Exemptions for Self-Driving Cars Into Law

  1. Bar them from urban areas until they are safe. If those in the suburbs and rural areas want unsafe self-driven cars on the road, then fine for them.

    Frankly, this calls for letter from those in urban areas to his constituents, asking why then (through him) are doing this to us?

  2. The industry itself will demand regulation because AVs will live or die on safety. The AV indiatry can’t allow low-quality software providers into the market just like the airlines can’t. Consumers won’t make the distinction and neither will accidents. So aviation type levels of safety regulation are coming – are indeed inevitable. That’s one of the huge impediments to producing self-driving vehicles at an affordable price point. There are literally millions of lines of SW to evaluate according to Design Assurance Level and perhaps re-write. There is a whole federal regulation structure that has to be stood up. All of that will be hugely expensive.

    And I haven’t even mentioned maintenance. Driving a pristine car with pristine sensors on pristine Arizona streets is one thing. It certainly isn’t a good validation of actual vehicle usage.

    Rushing these vehicles to market is a good way to kill the whole industry. Not that I am greatly troubled by that. I will never get in an AV.

  3. The claim in this story that the driver was watching TV on her phone is not substantiated by the link in to the NTSB report. Sloppy reporting

  4. Remember the Apple Newton? Visionary product, but didn’t work well. It wasn’t until the Palm Pilot came out that the PDA worked well.

    Robot cars are in the Apple Newton stage: visionary, but full of bugs. The difference is that the Apple Newton didn’t kill anyone.

  5. If you’ve ever ridden a train to go from one terminal of an airport to another terminal of that same airport, you’ve likely already been in an automated vehicle (AV). Very few airports operate trains with actual humans at the front controlling them – almost all airports use computers and a remote control room staffed with a few humans to operate their train systems. The difference of course is that AV is undoubtedly safe when it’s *guided* (by rails or other firm barriers that limit the pathway the AV can take) but these automated cars are attempting to operate everywhere without any restrictions on the paths they can take.

  6. People said the same about automated elevators. They couldn’t fathom getting into an elevator without a human conductor there to make sure nobody was caught in the closing doors. Now, virtually all of us step onto an elevator without thinking at all about how algorithms are determining whether those doors can close without catching someone or whether the car can start moving if the doors aren’t fully closed. It’s not the algorithms that are the problem. It’s the rush to put all this out there without proper testing and failsafes built in. Automated elevators receive far more safety inspection and regulation than AVs do in most states.

  7. I’m not enthusiastic about continuing to relegate life and death decisions to humans who are not only poorly trained, but in many cases utterly incapable of driving an automobile regardless of training due to deficiencies in reflexes, spatial perception, cognitive abilities, and attitude.

    The problem here isn’t so much that some humans (IMO < 10%) are capable of driving very safely given enough training. Rather, it's the idea of universal driving where we put the other 90% in charge of heavy machinery they can't safely operate. That's really why we need AVs. Unfortunately, most of the country isn't set up to get around with any mode other than automobiles. But if everyone uses automobiles, you have the problem of most of them being driven by incompetent drivers, with the resultant 40,000 deaths/millions of injuries annually.

    AVs even in their present state are safer than probably 99% of drivers. I find it ironic most of the crashes which have happened to date with AVs were caused by a human-driven vehicle doing something an AV would never do. This is why we need to remove human-driven vehicles from the roads as soon as AVs are ready for prime time. They can't reach their safety potential so long as they share the roads with unpredictable, incompetent human drivers.

    And note I might be fine with human drivers causing deaths, so long as it was only their death. When a mistake by a human driver can cost other people their lives, it's time to figure out how to replace that driver with something better. Or have people travel by modes which are inherently safer than automobiles (i.e. trains, buses, bicycles, walking).

  8. The difference here is that cars already kill something like a million people worldwide each year. AVs aren’t ready for prime time this year, but even so they’re already better than most drivers. Give it a few more years for the bugs to be worked out. As a pedestrian and cyclist, I’d much rather share the streets with predictable robot cars which I know will never exceed the speed limit, will always yield to me, and won’t run off the road and hit people on sidewalks. Human-driven cars do all these things and more simply because humans are really bad at tasks like driving.

  9. If you’ve ever flown on a plane, you probably didn’t know that pilots handle very little of the flight with many Flight Control Systems handling both take-off and landing.

  10. Moving a couple people up and down a linear chute is a vastly different endeavor than putting AI in charge of a complex, dynamic, 4 dimensional environment.

    That’s not to say AVs can’t improve safety. I haven’t seen evidence to support this claim yet, however.

  11. These “predictable robot cars which I know will never exceed the speed limit, will always yield to me, and won’t run off the road and hit people on
    sidewalks” won’t exist for decades, though (if ever).

  12. The flight control system in an aircraft is an easy problem when compared to a driving control system.

  13. “Robot cars are in the Apple Newton stage: visionary, but full of bugs.”

    Oh, please. The fact that Uber put its AVs on public roads far too early doesn’t mean everyone else working on them is equally incompetent or negligent. More specifically, Google’s Waymo unit has been testing its own AVs for nearly a decade, and already has millions of miles racked up on them in *public* testing environments. IIRC they’ve had *one* accident, and it was the other driver’s fault.

    And while this AV START bill is ridiculous as well as premature, it’s nonetheless fact that “robot cars” are already vastly safer than vehicles piloted by humans, given that 99% of all fatal collisions are caused entirely by driver error (including DUIs), as are nearly all auto collisions generally. When all the major players reach full Level 5 autonomy, it would be unconscionable from a public-policy perspective to NOT allow AVs to be disseminated as broadly as possible.

    That said, we absolutely need to establish *very* clear sets of standards for when a given vehicle reaches that point. The only area where I concur with John Thune is that we need federal-level regulation as opposed to a patchwork of state and local rules, but those rules need to be as clearly defined as every other one stipulating the inclusion of airbags, anti-lock brakes, traction control, etc. in a vehicle. We wouldn’t let any of *those* technologies into commercially sold vehicles unless they’d been thoroughly tested, and if anything AVs need to undergo significantly *more* thorough testing.

  14. What, seriously? Btw are you even aware that about half of the tech that’ll end up in AVs is already available? And that even mid-range cars now being introduced have them available as options at the very least? They have adaptive cruise control which uses radar/sonar to maintain a precise distance from the vehicle ahead of them; lane-change systems, to follow lanes on virtually every major road in the country; and emergency braking systems that will automatically slam on the brakes if a vehicle’s about to imminently collide with a stopped vehicle (or animal or whatever) in the middle of the road.

    Going even further, do you know that computers (or microprocessors) now control almost *every* vehicular function, at least in the latest-model vehicles? Steering, brakes, accelerator, etc. (Note that “traditional” emergency brakes have all but disappeared in 2018/2019 cars: they’re now just another button to push.)

  15. Try looking into Google’s Waymo program: it’s been in the works for over a decade now, and they’ve been tested daily on public roads for quite some time. Their error rate is only a hair above zero — they’ve never had any sort of serious collision, and IIRC only a handful of minor ones (each of which was the other driver’s fault) — and they’re planning on commercially introducing their vehicles as ride-hail cars late next year.

    It would be a considerable fallacy to judge AVs generally based on a single company’s reckless and grossly negligent attempt to be first-to-market (Uber’s, obviously).

  16. Actually, it was widely reported several weeks ago that the Uber “safety” driver was watching “The Voice” when the vehicle struck and killed a pedestrian.

  17. Um, didn’t Uber just settle a lawsuit that they stole their tech from Waymo? Not sure why you think Waymo is going to be safer than another company that was basically using the same tech.

  18. People said the same thing about the segway scooter, which we are all riding to work. Ha ha ha.

  19. WTF? This isn’t even vaguely true, or at least not them “basically using the same tech.” Google’s work on AVs dates back to 2005. They’ve been testing them on public roads for nearly a decade, and had their first *driverless* test before Uber even *started* its own AV program.

    What this lawsuit was about was that Uber is screwed, and its executives *know* they’re screwed: they are outgunned by Google in terms of time spent and resources available — Google has over $100 BILLION IN CASH literally at its disposal, while Uber is hemorrhaging money like a stuck pig. Consequently, they allegedly tried to take a shortcut by poaching one of Google’s earliest AV engineers — to the tune of $660 million to buy out Otto, the automated truck company he started — and making sure he snuck all of Google’s AV IP out the door before he departed.

    While it’s been conclusively proven that he copied something like 15,000 separate files off of Google’s servers — which wasn’t exactly difficult, considering he merely copied them over to his personal Dropbox account — Google’s investigators still couldn’t directly connect Uber to any of it … which is roughly akin to Robert Mueller having difficulty directly tying Trump to Russian collusion in the 2016 election, even though everyone knows that’s exactly what happened. So: they ended up settling the case.

    NONE OF THAT, however, has anything to do with the fact that Uber is so far behind Google’s Waymo unit in the “race” to develop AVs that it would constitute sheer farce if someone hadn’t literally DIED as a result of their gross negligence. To borrow the analogy from the first comment: Uber’s trying to “perfect” its Apple Newton, while Waymo’s putting the finishing touches on its iPhone X. (Or I guess Pixel 2 would be more apropos.)

  20. Actually, the Segway really *was* a brilliant invention. The problem with it was twofold: 1) It’s illegal nearly everywhere to operate motorized vehicles of any kind, including Segways, on sidewalks, plus back then cities didn’t have ubiquitous bike lanes like they do today; and 2) they cost $5K a pop.

    Today, however, we do have all those badly needed bike lanes, and much less sophisticated scooters are available for a couple hundred bucks, hence the reason Bird and LimeBike went from 0 to unicorn in all of three months.

  21. Yes, I do. I’m an Engineer. I’ve been one for 30 years. There is a difference between building a prototype and building a product for sale. The big challenge is making autonomous decisions in a crowd of autonomous vehicles, and doing so reliably and safely over tens of millions of hours of operation without killing large numbers of people. Lane following is the easy part.

  22. “Tens of millions of hours of operation”? If you’re an engineer one would *think* you’d know a million hours is approximately 115 years…

    In any event, Waymo now has 600 AVs operating 16 hours a day in real-life environments, and they certainly have tens of thousands of hours of accumulated drive time (if not hundreds of thousands). And again, they didn’t start testing them “in the wild” until at least five years of testing on private grounds. Uber, on the other hand, somehow convinced several cities to let them use their *streets* as initial testing grounds, not controlled environments, and an entirely predictable catastrophe occurred.

    “The big challenge is making autonomous decisions in a crowd of autonomous vehicles”

    Hmmm. You don’t think making autonomous decisions in a crowd of *human*-driven vehicles is more difficult?

  23. How many cars are in the United States? How many hours per day do you think they operate? How many days per year? Do the math.

    You don’t know anything about Engineering at all, do you.

  24. I as a bicyclist like my chances of being noticed by a self-driving car’s sensors much more than by some incompetent driver who is looking at his/her phone or picking his/her nose or dozing off.

    What’s more, automated vehicles will stop at stop signs, and will do so behind the stopping line rather than within the crosswalk. They will remain stopped at red lights until the light actually turns green. And, when they go, they will not exceed the speed limit.

    The danger associated with automated vehicles exists on the macro level; it consists of the possibility that more space in our cities will be given over to the use of cars. We will need to guard against that on the policy level. But each individual self-driving car that replaces a human-controlled car represents a move towards increased safety, and therefore a significant victory for the interests of bicyclists and pedestrians.

  25. By my “math” it doesn’t add up how *any* significant new technology could progress if it had to follow your curious false equivalency that a new invention somehow can’t be deemed “safe” until millions of units have logged millions of hours of operation.

    Let’s go with commercial airliners as an example (you did, after all, mention them): I don’t know about you, but — in the context of the earliest days of commercial jet service back in the late ’50s — traveling over 500mph in a pressurized tube at 30,000 feet seems a hell of a lot more intimidating than the notion of getting into a robocar today. As it turned out, commercial airlines ended up becoming by far the safest way to travel, but using your “math” the entire industry would’ve failed to get off the ground (pun somewhat intended) since obviously they didn’t have millions — or even tens of thousands — of hours in the air prior to their evolution past the prototype phase and introduction to the general public.

    Going back to your original point:

    “There is a difference between building a prototype and building a product for sale.”

    I completely agree. What I don’t think you’re getting is that AVs are at this point considerably closer to the latter than former, or that Google will have concluded nearly 15 years of exhaustive research and testing on their AVs when they’re introduced commercially in 2020.

    P.S. “Engineer” is not a proper noun — so its correct spelling is “engineer,” not “Engineer” — and there are myriad different varieties of them. Just within the ranks of my own family, we’ve had a nuclear engineer, petroleum engineer and systems engineer. Why, then, do you assume all “engineering” is somehow identical?

  26. The cost was the biggest issue. The sidewalk issue was secondary. Had the price been reasonable enough for lots of people to afford Segways doubtless there would have been a big lobbying effort to make them legal on sidewalks.

    Nowadays with relatively inexpensive e-bikes readily available, along with place to ride them, most of the reasons for Segways no longer exist. Like many great inventions, it was only viable for a brief period of time. Look at how many other things were really only useful for short periods of time before technology made them obsolete:

    Hopefully, the driver in motor vehicles will be obsolete in the not too distant future.

  27. “The danger associated with automated vehicles exists on the macro level; it consists of the possibility that more space in our cities will be given over to the use of cars.”

    Actually, I’d argue quite the opposite: most theories at present on how AVs will evolve assume scenarios in which individual vehicle ownership is made functionally obsolete. The average person uses their car 4% of the time, so if cars can drive themselves around, why on earth would you “waste” them on a single owner?

    Just to cite a few possibilities: obviously automated ride-hail service (e.g. Uber & Lyft) is a given, and certainly so in the early days of the technology — though Google’s Waymo unit will beat both of them to market with an actual product — but really, many companies could in theory provide “on-demand service” of one type or another.

    We’re also likely to see subscription services evolve for AVs, much like they did with the internet (which didn’t take off until AOL introduced “all-you-can-eat” dial-up access for $20 a month, thus eliminating the by-the-hour pricing models formerly used).

    Regardless, with shared use comes a *vastly* lesser need for parking, so at some point we’re going to end up with a MASSIVE amount of excess garage and parking lot space that could be repurposed in scores of ways. And that’s even *before* considering the extent to which AVs could help solve some of the most perplexing problems relative to public transit, in particular the last-mile issue. If, say, a shared-van AV can pick you up either at your doorstep or close to it — and either drop you off at a central train or bus depot, or take you all the way to your destination — that even further eliminates the need for both car ownership generally and storage space for them specifically.

  28. a new invention somehow can’t be deemed “safe” until millions of units have logged millions of hours of operation.

    It’s not about test hours logged. It’s about reliability prediction. It’s impossible to log those kinds of hours in testing. No one does that. It’s done by analysis. Existing systems that can kill people (say, DAL A components in a commercial airliner) have reliability requirements of nine nines. You don’t realize it, but the reliability requirements on an AV will be extraordinary. Otherwise, you will get too many accidents, and accidents will be fatal to a system like this. That is why airliners have such large reliability requirements – because too many accidents would literally kill the industry. oh, btw. Those kind of reliability re

  29. I sure hope it plays out that way!

    Though we could see a dramatic increase in aggregate car usage, as people take trips in autonomous vehicles that they otherwise wouldn’t have taken. And this could result in pressure for more highways, wider streets, and so forth.

    Let’s hope that the scenario that you describe is the one that prevails.

  30. Even if aggregate car usage goes up, car ownership will still be going down. Eventually, my guess is almost nobody will want to own a car, especially once manual driving on public roads is made illegal. That will mean former parking spots can be repurposed. This will free up lots of space. You won’t need wider highways or streets, either. AVs can run virtually bumper to bumper. One lane should be able to carry the traffic 3 or 4 lanes used to carry. We may be able to shrink our highways even if more people are using them.

  31. I’ll gladly take even today’s flawed robocars over the alternative of much worse human drivers.

  32. I hope they get rid of John Thune very thune. With him in Congress, nothing reasonable is safe.

  33. Man, the pro-auto-auto crowd is out in numbers today. These things will probably end up being safer, but if and until they are fully implemented in say 20 years, it will be a huge transition, larger than when horses and carriages were replaced by cars. It will change everything, including that whole liberating idea of getting in your car and driving it yourself whereever you want. In the meantime, until they get their algorithms right, bike riders and pedestrians will probably be required to wear sensors or cede liability (one more strong emf source in a cloud of electrosmog giving everyone cancer and destroying our mitochondrial dna, killing bees, and so on) Perhaps there will be a religious exemption like Amish have for their horse & carriages for those who still insist on their God-given right to steer.
    As for the near-term, give me a real driver who can hear me yell at him and recognize the finger over an algorithm that might mistake me for a piece of paper in the road.

  34. As I recall, we the sheeple haven’t been given the “opportunity” to vote whether or not we even WANT vehicles on our public highways that drive themselves. Why is that, since doing so puts our lives at risk?

  35. Carl Jacobs 1, Captain Kirk 0.

    Nobody has yet figured out what AVs are supposed to do when a couple punk kids decide to dance in the street and hold up traffic for miles on 5th Avenue. There go all those efficiency savings. A human driver would win the game of chicken with the punk kids, sending them scurrying back to the sidewalk. But the AV will surrender every time. Instant traffic jam.

    Or here’s another fun one: slide a mannequin, or even a cardboard cutout of a person, on an expressway at rush hour. A human driver would plow through the mannequin. But the AV will go from 55 mph to a complete stop, and sit there indefinitely. Instant traffic jam.

  36. If AVs are really going to take most of the cars off the road, how is Big Auto going to stay disgustingly rich? Shouldn’t Big Auto be trying to kill AVs?

Leave a Reply

Your email address will not be published. Required fields are marked *