Arizona governor and Uber kept self-driving program secret, emails reveal

With some exceptions, Federal powers increased because we couldn't agree that human beings are not property, among other basic rights issues. The touchstone moments were not simple disagreements on policy, it was the pot boiling over into extreme violence. Even the Federal model requires some basic level setting for it to function. There's either a rebalancing or the whole thing falls apart.

Regardless, GP's points are notable just a local level. States have different weights assigned to executive and legislative. It appears that Arizona leans heavily towards a strong executive branch although I have to wonder if any laws were broken here. People expect corruption in US politics at this point but this incident has some staying power in the news.

I didn't mean that the federal agencies should necessarily have a role here. I would think that many US states also have agencies charged with administrating the law, something like a "Arizona Roadway Authority" or whatever.

My comment was trying to explain how even within a single hierarchy like a state's executive, which a Governor at the top and traffic engineers somewhere below, it's peculiar to see someone jump so many levels of that hierarchy to assume control of a rather low-level decision.

Maybe I'm misunderstanding your comment, but this was months ago. No rationalizing needed.

But not necessarily wrong

The Uber-Arizona story is the American economic-political dystopia in microcosm.

For-profit corporation helps elected politician maintain power.

Elected politician uses the power of the state to promote the corporation and eliminate oversight that would have been in the public interest.

All of this is done in secret, of course. The rubes who elected you don't need to know what their "representatives" are doing in their names.

The politician wins, the corporation wins, and the public is the big loser.

I don't need to believe there was a conspiracy to protect Uber here. This spot apparently has a history of pedestrian crossing, and many police departments have a habit of assigning all blame to the pedestrian in a pedestrian/vehicle collision. This blame is reinforced by the driver's "evidence" (driver didn't see anything, but she wasn't looking) and the dashcam (can't see anything, but the quality is atrocious and not at all reflective of what most drivers would see).

One thing we should take away from this incident is that we need to seriously push back on immediate "pedestrian's fault" rulings when police make them after a pedestrian/vehicle collision. It was readily apparent within hours that the story didn't add up, but people only paid so much attention because it's the first such crash involving an autonomous vehicle.

The more interesting question is how that shockingly poor dashcam footage came to be. That quality is so poor that I find it hard to believe that Uber is actively using it as a sensor input. Hell, it's worse than most dashcams I've seen on Youtube...

Imagine the police people having seen their governor (who effectively is their boss, as far as I know, so he is the "hand that feeds them") repeatedly be very, very protective and favorable towards Uber and the self-driving car testing program. Even if they never got an explicit order of the likes of "Courteously close your eyes when dealing with Uber!", they will be much more protective of Uber than they were if their boss was known as very critical regarding Uber. This is basic human instinct, and instinctive behavior often leads to decisions which might turn out to be obviously stupid in the long run.

This is extremely naive, have you ever worked in organizations?

I'm not being facetious, but a little top-down pressure can go a long way. This isn't about a grand conspiracy as much as it is about simple human nature, and the act to please those above you, keep your job, and move up yourself.

That's .. not how credibility works for the police.

These facts are only relevant if you disregard LIDAR and you believe that the camera showed what the human eye would see. Videos on youtube (https://www.youtube.com/watch?v=CRW0q8i3u6E and also https://www.youtube.com/watch?v=1XOVxSCG8u0&feature=youtu.be) and ABC news footage (http://abc7.com/video-shows-self-driving-ubers-fatal-collisi...) at the scene of the crash paint a much brighter picture. The street lights in the footage appear to be distinct while the background is dark, indicating the gain of the sensor would appear to be reduced. Try taking a picture of streetlights at night, if the street is visible, they will appear as stars. Additionally, the safety driver was not paying attention.

EDIT: In my opinion, the more you think about this crash the worse Uber looks. The LIDAR should have worked, the lighting was likely brighter than the video shows, the weather was good, no obvious distractor objects, the road was straight, no parties were impaired. The pedestrian had a big bulky bike and crossed at right angles to the car. This is nearly the most basic test of self-driving cars and it failed. The only way Uber would look worse is if it had happened in day time. On top of that, Uber disabled the Volvo auto-braking safety system. (https://www.sfgate.com/business/article/Uber-Disabled-Volvo-...)

The test for a self-driving car isn't "does it drive as well as a human," it's "does it drive as well as a self-driving car which normally beats a human should."

This might sound like it makes no difference, but the car had LIDAR. That the scene was badly lit, whether true or not, is irrelevant; the car didn't use ambient light to see. If it could hit her under these circumstances, then it could just as easily hit her at noon.

Human drivers and other autonomous cars could have avoided this accident.

Uber's systems failed spectacularly, and should not have been on public roads.

Now we learn that Uber and the governor secretly collaborated to enable this, to neutralize any effective oversight, and to keep the public in the dark.

Blaming the victim doesn't illuminate any of these issues.

If you even believe the video is representative of what a human would see (which is totally false, as recent local videos of that stretch have shown), then the car was driving beyond its headlights. That's unsafe driving.

Deer are not reflective. Random objects that have fallen off a truck are not reflective. People are often not reflective, and there is no law requiring them to be.

And this woman was wheeling a bright pink bike, which shows up quite well. (It may also have had standard spoke reflectors, which might not show up in this video. hard to say.)

And finally, there weren't any crosswalks in that area for her to use. Yes, she misjudged the speed of the car, but her choice to cross the road at all is pretty solid.

> If a person hit the woman it would not have been homicide or manslaughter even if we didn't know the driver was on their phone.

That's because the bar for vehicular manslaughter is incredibly high. There is a huge range of shitty road behaviour that is not illegal, but you should never do, for the sake of yourself and other people.

Driving faster then you can see is one of those behaviors. Hitting someone because you had legal right of way is another. Right of way is good and all, but your first job as a driver is to avoid crashes. Right of way is just a set of guidelines for avoiding crashes... And a way to assign legal blame when one happens.

When I sea kayak, I have right of way against large boats in some situations. I'd be a complete, and dead idiot if I were to exercise my legal right to stay on course.

I would like to see the more detailed telemetry that the car gathered at the time.

For a human driving the car, the dash cam together with personal testimony would probably put the issue to rest; although the fact that the pedestrian was in the roadway before they entered the lane looks very suspicious because a human likely would have seen them because of headlights on an open road even if not illuminated by streetlights. Nonetheless, the dash cam shows what it shows, and human sense and memory is fragile.

Not so for a self-driving car -- the memory is not fragile, and should be preserved. If it is not preserved and there is no remaining copy of the car's telemetry at the time, that in itself is a detail that can point towards gross negligence on the part of Uber. If it is preserved, the question becomes, would a "reasonable person/software" have acted based on what was shown there, and that goes towards whether there is negligence or recklessness. But we (or at least, investigators) need to see that telemetry to make that decision.

> I would say the pedestrian was legally more than 50% at fault in this horrible accident.

Which is fine for Uber-as-operator if they were operating in a state which uses a contributory negligence or modified comparative negligence rule, but a number of jurisdictions—Arizona, for instance—use a pure comparative negligence rule where, e.g., an injured party that is 99% responsible can recover 1% of damages from a party that is 1% responsible.

Moreover, it's no help for Uber-as-manufacturer if strict product liability principles are applied. Now, it's not clear to me that they would apply to Uber as manufacturer as long as the product isn't sold to a third party, but even if they don't apply under current law, Uber using operating a service directly with the product rather than selling it to other operators as an end-run around liability for an unreasonably dangerous product invites a change to the law which would extend such liability to manufacturer-operators.

>But until then, the whole thing is deeply experimental. And regulation is premature

If it is deeply experimental, then it should be conducted in experimental settings, and not on public roads. If regulation is premature, then public tests are also premature. I am surprised I have to explain this. To be on public road the vehicle should pass a minimum benchmark that shows that it can consistently handle at least all the common scenarios..

There is no urgent need for self driving vehicles. It is not like they are going to solve some of the pressing issues that humanity is facing right now. It is just a curiosity, and can may be boost the profits of some corporations. That is about it.

I don't think it is worth or just to risk lives of unsuspecting public towards that end..

"Otherwise, you may be crushing potentially great ideas unknowingly."

Along with crushing some human bodies along the way.

> NHTSA strongly encourages States not to codify this Voluntary Guidance (that is, incorporate it into State statutes) as a legal requirement for any phases of development, testing, or deployment of ADSs. Allowing NHTSA alone to regulate the safety design and performance aspects of ADS technology will help avoid conflicting Federal and State laws and regulations that could impede deployment.

Can someone clarify that this please? To me it appears that this body is saying that, "States, Please don't make laws requiring minimal standards for SDVs for things, including deployment on public roads, because that would impede deployment. Allow us to handle all that"

But they didn't do it, right? SDV's were on roads even before the "safety design and performance aspects of ADS technology" was regulated, right?