Uber's Video Shows The Arizona Crash Victim Probably Didn't Cause Crash, Human Behind The Wheel Not Paying Attention

from the everyone-error dept

In the wake of a Tempe, Arizona woman being struck and killed by an Uber autonomous vehicle, there has been a flurry of information coming out about the incident. Despite that death being one of eleven in the Phoenix area alone, and the only one involving an AV, the headlines were far closer to the "Killer Car Kills Woman" sort than they should have been. Shortly after the crash, the Tempe Police Chief went on the record suggesting that the victim had at least some culpability in the incident, having walked outside of the designated crosswalk and that the entire thing would have been difficult for either human or AI to avoid.

Strangely, now that the video from Uber's onboard cameras have been released, the Tempe police are trying to walk that back and suggest that reports of the Police Chief's comments were taken out of context. That likely is the result of the video footage showing that claims that the victim "darted out" in front of the car are completely incorrect.

Contrary to earlier reports from Tempe’s police chief that Herzberg “abruptly” darted out in front of the car, the video shows her positioned in the middle of the road lane before the crash.

Based on the exterior video clip, Herzberg comes into view—walking a bicycle across the two-lane road—at least two seconds before the collision.

Analysis from Bryan Walker Smith, a professor at the University of South Carolina that has studied autonomous vehicle technology indicates that this likely represents a failure of the AVs detection systems and that there may indeed have been enough time for the collision to be avoided, if everything had worked properly.

Walker Smith pointed out that Uber’s LIDAR and radar equipment “absolutely” should’ve detected Herzberg on the road “and classified her as something other than a stationary object.”

“If I pay close attention, I notice the victim about 2 seconds before the video stops,” he said. “This is similar to the average reaction time for a driver. That means an alert driver may have at least attempted to swerve or brake.”

The problem, of course, is that AVs are in part attractive because drivers far too often are not alert. They are texting, playing with their phones, fiddling with the radio, or looking around absently. We are human, after all, and we fail to remain attentive with stunning regularaty.

So predictable is this failure, in fact, that it shouldn't surprise you all that much that the safety operator behind the wheel of this particular Uber vehicle apparently is shown in the video to have been distracted by any number of things.

A safety operator was behind the wheel, something customary in most self-driving car tests conducted on public roads, in the event the autonomous tech fails. Prior to the crash, footage shows the driver—identified as 44-year-old Rafaela Vasquez—repeatedly glancing downward, and is seen looking away from the road right before the car strikes Herzberg.

So the machine might have failed. The human behind the wheel might have failed. The pedestrian may have been outside the crosswalk. These situations are as messy and complicated as we should all expect them to be. Even if the LIDAR system did not operate as expected, the human driver that critics of AVs want behind the wheel instead was there, and that didn't prevent the unfortunate death of this woman.

So, do we have our first pedestrian death by AV? Kinda? Maybe?

Should this one incident turn us completely off to AVs in general? Hell no.

Filed Under: ai, arizona, autonomous vehicles, driverless cars, pedestrians
Companies: uber

Slowing Down Driverless Cars Would Be A Fatal Mistake

from the people-are-already-dying dept

Unsubstantiated driverless car hype may be annoying, but that shouldn't blind us to the real cost of unnecessarily delaying autonomous vehicle (AV) deployment.

Last week, after exploring new data from the California AV disengagement reports, Ross Marchand of the Taxpayers Protection Alliance argued that we should "put driverless cars back in the slow lane." California requires AV companies testing in the state to report each time a human operator takes over for a driverless car — an event otherwise known as a "disengagement." Marchand offers some interesting analysis, but ultimately reads far too much into a limited dataset and pushes for a restrictive policy prescription that would undermine public safety. The discussion is worth fleshing out because it reveals important limits to the "precautionary principle" mindset that is so common in AV discussions.

In 2017, Waymo — the self-driving car project formerly belonging to Google — reported driving over 350,000 miles on California roads with 63 total disengagements. Marchand claimed that, based on these data, Waymo's test vehicles are still not as safe as human drivers and that they are improving at a slower rate than those hyping AVs would have us believe. Further, he argued that until driverless cars can prove they are safer than human operators, we should keep them off public roads — and instead test them on expensive private tracks.

There are a few glaring issues with this argument. First, it overestimates how applicable and reliable the California disengagement data really are. As many commentators have pointed out, disengagement data are a poor measure of AV progress. Not only are disengagement reports an apples-to-oranges comparison across vehicle manufacturers who use different definitions, strategies and road conditions for testing, but Marchand drills down by comparing particular disengagement subcategories, leaving him with sample sizes of less than 20 — several orders-of-magnitude too small to make meaningful comparisons. Furthermore, comparing disengagements to would-be fatalities is problematic given that a safety driver's presence enables testing in regions that the vehicle is still learning to handle.

Marchand also left aside the successful testing and deployment of Waymo's fully driverless cars in Arizona. Since November 2017, hundreds of AVs have been providing free taxi services in the suburbs of Phoenix without any safety driver in the front seat. To date, there have been no reported accidents or fatalities. This suggests what we've known all along: these companies already face a host of legal, political, economic, regulatory and publicity pressures that incentivize them to prioritize safety in AV deployment. They know that every bump, scrape and crash will make headlines (regardless of who is at fault) and will slow or — if it's serious enough — completely derail their path to market. Waymo obviously feels confident enough to take its hands off the wheel, and so far has been right. Why rip AVs off the roads when no one has been harmed?

Marchand's larger argument against AV testing on public roads provided a textbook example of the precautionary principle in practice. Simply put, the precautionary principle requires innovators to prove that their new technology will not harm society, rather than placing the onus on regulators and litigators to demonstrate that an innovation actually causes harm.

And to that point, Marchand fails to specify what exactly the harm of public testing has been. Public testing has not unleashed mass fatalities on society, or even mass fender-benders. Rather, it appears to be speeding up the feedback loop of better data and more-rigorous test environments, leading to faster improvements in autonomous technology.

As a society, we can't afford to wait until we are 100-percent certain that driverless cars are statistically safer than humans before letting them on the roads. As a report from RAND highlighted, it could take several decades to accumulate enough miles on private test courses to know beyond a shadow of a doubt that AVs are safer than their human counterparts. Relying on Marchand's precautionary principle approach would mean waiting decades while nearly 40,000 people die on our roads every year. Regulatory delay of this magnitude could, conservatively-speaking, cost tens of thousands of lives.

That's not to say private test courses don't have a role to play in AV development. Indeed, Waymo already operates an extensive test track in Arizona where operators take real-world scenarios and experiment with hundreds of possible variations. This hybrid approach combines the advantages of real-world testing and private test courses. But forcing all AV testing onto private test tracks cuts off the real-world data necessary for this complementary approach and substantially raises the barrier to entry for new competitors.

To be clear, we should avoid over-hyping the progress made in AV development. Carefully taking into account the safety data will be a key part of this effort. But halting all real-world AV deployment is a heavy-handed "solution" desperately in search of a problem.

Caleb Watney (@calebwatney) is a technology policy associate at the R Street Institute. Marc Scribner (@marcscribner) is a senior fellow at the Competitive Enterprise Institute.

Filed Under: autonomous vehicles, california, driverless cars, human drivers, ross marchand

Driverless Cars: Disrupting Government Reliance On Petty Traffic Enforcement

from the uber-but-for-killing-your-small-town-speed-trap dept

Self-driving cars are on the way, and in their wake, they'll leave a variety of entities slightly less better off. Insurance companies may be the first to feel the pinch, as less-than-risk-averse drivers are replaced with Electric Grandmothers more than willing to maintain safe speed limits and the proper distance between vehicles. And as goes the car accident, so go other areas of the private sector: personal injury/DUI lawyers, hospitals, body shops, red light camera manufacturers, towing companies, etc.

But the public sector will take the hit as well. "Flow my tears," said the policeman.

Consider the following. This past year, the City of Los Angeles generated $161 million from parking violations. Red light violations have a fee of $490. Californians caught driving under the influence are fined up to $15,649 for a first-offense misdemeanor DUI conviction and up to $22,492 for an under-21 equivalent. Cities in California collect, on average, $40 million annually in towing fees that they divide with towing firms. Simply put, the hundreds of millions of dollars generated from poor driving-related behaviors provide significant funding for transportation infrastructure and maintenance, public schools, judicial salaries, domestic violence advocacy, conservation, and many other public services.

Since California legalized driverless vehicles, Google has logged more than 1.7 million miles during the testing phase and been involved in 11 accidents, none of which were the fault of the driverless vehicle. Tesla, Mercedes, and others are not far behind. It turns out that automated vehicle technology—unlike humans—abides by the law. And that’s bad news for local government revenues. In other words, once driverless cars become mainstream, deep revenue sources acquired from driving-related violations such as speeding tickets and DUIs will decrease greatly.

Someone has to pay for the roads and other government activities, but it won't be drivers. So, as the Brookings Institution report points out, new revenue streams will have to be sought. The obvious suggestion is tax-per-mile billing, but that puts the government right in your vehicle -- an idea that's not going to gain in popularity any time soon.

While the loss of revenue will have an impact, the picture painted here is skewed. For many years, communities have treated police departments as revenue generators, rather than crime fighters. This has skewed incentives so badly that some small towns have become nothing more than profitable speed traps. That's one end of the issue: the pressure (or the willingness) to overpolice minor traffic violations to keep city governments (and the police departments themselves) funded.

But that's only part of it. The situation looks rather dire, especially if one doesn't examine what's not being said in these paragraphs. As Scott Shackford at Reason points out, the Brookings Institution report does some mighty fine cherry-picking for its list of potentially-affected government services. Without a doubt, a downturn in revenue will affect good government programs like public schools and domestic violence programs. But it will also cut back funding for far more dubious government spending.

What an interesting list of government-financed uses they've chosen. Notice they left off "Poorly made third-party database software that will stop working properly in less than three years and that was purchased from somebody belonging to the same frat as the assistant city manager," "police abuse settlements," and "blatant pension spiking."

These "losses" will also be somewhat offset by less tax revenue being spent on traffic enforcement, accident response units and other related law enforcement activities. This will also mean fewer law enforcement officers will need to be employed, which should further reduce government expeditures.

The problem is that most governments aren't capable of heading off this sort of "threat" to their livelihoods, even with years of advance notice. Trimming back unneeded public sector employees won't happen until years after it's obvious they're no longer needed and will often come accompanied with expensive severance packages. New tax revenue streams won't be explored until they can be put off no longer, and often will just be added on top of existing taxes, rather than replacing those that have slowed to a trickle.

Worse, those most affected by this sort of shift will be the same people most affected by most government tax increases: the poor. The lowest income brackets will be the last to adopt driverless vehicles, leaving them the most exposed to fines for traffic violations (fines that will likely increase as revenue dwindles), as well as new costs like per-mile taxation. They're also most likely to see support programs they rely on suffer cuts as traffic enforcement money dries up.

The report somewhat addresses this outcome with a discussion of income inequality and the "disappearance of the middle class." While some of it is accurate and some of it is mostly buzzwords in search of a point, there's no doubt that traffic enforcement revenue will mostly be collected from those who can least afford it. After all, governments have done this for years -- something that helped fuel the outrage and backlash in Ferguson after the shooting of Michael Brown.

Is Brookings actually trying to blame the gap between billionaires and the poor for the racial tension in Ferguson? Which venture capitalist was it who told the Ferguson police to step up fine collection to rake in more money for the city's coffers? Which hedge fund manager invented the bureaucratic court system in Ferguson and other St. Louis County cities designed to wring every last cent from any indigent minority who couldn't afford an attorney? Which Wall Street "fat cat" is adding additional fees to every little fine so that getting pulled over for something as simple as not signaling a turn could end up costing hundreds of dollars for somebody who could end up losing his license and his ability to even work?

While driverless cars hold a great deal of disruption potential, when it's all said and done, governments will remain largely undisrupted. Whatever changes are made in response will arrive well after they're needed and be badly implemented. The same people who suffered in the previous system will find no improvement in the next one. While one would hope the drastic reduction in traffic enforcement would result in better, smarter policing more focused on serious criminal activity, old habits die hard. Cops will just go where the driverless car ain't, rather than trim that area of law enforcement to the minimum required. And cities will cut programs deemed expendable, rather than subject their own spending habits to greater scrutiny.

Filed Under: autonomous vehicles, driverless cars, taxes, traffic enforcement, traffic fines

FBI Thinks Driverless Cars Could Be Criminals' New Best Friends

from the I,-for-one,-welcome-out-new-Robot-Death-Car-overlords dept

There have been dozens of techno-panics over the past several years, but one usually expects cooler heads to prevail in government agencies where the word "investigations" is in the title and the employees have access to plethora of cutting-edge equipment. (Note: said "cutting-edge equipment" for surveillance only. Agency computer systems remain an outdated mess and the idea of recording in-custody interviews has finally arrived nearly four decades too late.)

We are, of course, talking about the FBI. The FBI's relationship with technology is painfully one-sided. On one hand, it's pushing to gets its biometric database online and fully operational, hopefully years before its report on this database's privacy implications finally arrives. On the other hand, it has argued in court (via the DOJ) that smartphone technology vastly outpaces law enforcement's tools to access possible evidence and, therefore, should be obtainable without a warrant.

But it is an investigative agency, which would seem to indicate it has the ability to gather facts and come to informed decisions. But it seems to prefer worried conjecture to actual data. (See also: Insane Clown Posse fans are an organized criminal entity.) But here it goes again, seeing another technological development as another way for criminals to gain the upper hand.

In an unclassified but restricted report obtained by the Guardian under a public records request, the FBI predicts that autonomous cars “will have a high impact on transforming what both law enforcement and its adversaries can operationally do with a car.”

In a section called Multitasking, the report notes that “bad actors will be able to conduct tasks that require use of both hands or taking one’s eyes off the road which would be impossible today.”

The FBI looks at something that has the potential to make roads much safer and sees… autonomous vehicles loaded with gunmen, all of whom can use both hands to fire at pursuing law enforcement… or something.

Autonomy … will make mobility more efficient, but will also open up greater possibilities for dual-use applications and ways for a car to be more of a potential lethal weapon that it is today.”

Sure, the driverless vehicles could be loaded up with explosives and "told" to drive itself to its destination, but that seems like an incredibly expensive way to deliver a payload. And sure, vehicles might be hacked to ignore everything about them that makes driving safer, but that last part is nothing a human operator can't do in a normal, cheaper vehicle. And any vehicle with a driver can still carry armed criminals/explosives.

Even the FBI admits that autonomous cars present the agency with certain advantages, including the fact that the first few iterations of car AI will be able to do little more than recreate OJ Simpson's "getaway."

[A]utonomous cars would likely face many hardships with evasive driving or car chases...

But using this AI for good (and hacking it to serve its purposes) may also revolutionize the FBI and law enforcement's pursuit techniques.

"[A]lgorithms can control the distance that the patrol car is behind the target to avoid detection or intentionally have a patrol car make opposite turns at intersections, yet successfully meet up at later points with the target."

While some of this report is undoubtedly dedicated to "what if" scenarios not unlike the risk disclosures included in IPOs, there's still something ridiculous about an investigative agency being so tuned into the terror frequency that it sees criminal intent in every technological advancement. If we wanted fear-based speculation about potential havoc-wreaking by new inventions, we'd ask the MPAA.

Filed Under: autonomous vehicles, driverless cars, fbi, threats