[C8]

What happens if heavy snow covers the sensors and landmarks?

Who pays if the software malfunctions and causes an accident?

Can these things be hacked?

[ennui2]

There were a host of concerns surrounding the introduction of the auto 100 years ago. It turns out some of those concerns can't be reconciled and we just collectively made the decision that the benefits outweigh the costs (including fatalities, noise pollution, localized and global air pollution, etc...). The same is going to be true here. No, it will never be perfect, since we've known since the days of Plato that perfection does not exist, but like anything else, that will not stop it from moving forward.

[Tanada]

There were a host of concerns surrounding the introduction of the auto 100 years ago. It turns out some of those concerns can't be reconciled and we just collectively made the decision that the benefits outweigh the costs (including fatalities, noise pollution, localized and global air pollution, etc...). The same is going to be true here. No, it will never be perfect, since we've known since the days of Plato that perfection does not exist, but like anything else, that will not stop it from moving forward.

Just because you personally believe the benefits outweigh the costs does not mean society as whole will agree with your opinion. There have been a great number of inventions over the last century, many of them never became mainstream and failed to achieve market penetration.

Just as examples, turboprop airliners were both quieter and more fuel efficient that either the piston propeller or turbojet aircraft of the late 1950's and early 1960's, but they never became the standard aircraft. FM radios were invented in the 1940's for consumers but did not penetrate the market until the 1970's and then only as two band radios that still received the older AM frequencies. Ammonia burning conversions for farm engines so that farmers could use the same liquid they bought as fertilizer to power their farm machinery never took off, despite the cost and maintenance advantages for the farmers. I could give more examples but I think I have made my point.

[ennui2]

Just because you personally believe the benefits outweigh the costs does not mean society as whole will agree with your opinion.

Just because you personally believe the benefits DON'T outweigh the costs does not mean society as whole will agree with your opinion.

There have been a great number of inventions over the last century, many of them never became mainstream and failed to achieve market penetration.

Nothing with this sort of forward momentum didn't in some fashion reach critical mass.

I could give more examples but I think I have made my point.

Hardly. Again, nothing with this much money and industrial momentum and cultural zeitgeist pushing it forward hasn't reached critical mass.

This is part of the greater AI/automation push. It's not a flash-in-the-pan.

[C8]

There were a host of concerns surrounding the introduction of the auto 100 years ago. It turns out some of those concerns can't be reconciled and we just collectively made the decision that the benefits outweigh the costs (including fatalities, noise pollution, localized and global air pollution, etc...). The same is going to be true here. No, it will never be perfect, since we've known since the days of Plato that perfection does not exist, but like anything else, that will not stop it from moving forward.

I agree we can't have perfection- but a computer hacking or virus in the software of a vehicle moving 70 mph may be a "cost" that people find too high for the benefit

[ennui2]

In which case the solution will be strengthening security rather than abandoning the technology.

[vox_mundi]

Tesla Model S Hacked Remotely, Just As U.S. Government Announces Self-Driving Car Policies

Tesla Model S digital panels

A team of Chinese researchers from the Keen Security Lab published a video in which they demonstrated how a Tesla Model S can be hacked remotely, even from 12 miles away. The announcement appeared roughly at the same as the U.S. government’s guidelines for self-driving vehicles

Connected cars, but more importantly, self-driving cars or cars with self-driving capabilities such as the Tesla Model S, are vulnerable to hacking. They often come with internet connectivity, which sometimes can’t even be disabled, because the manufacturers want to ensure some permanent remote capabilities--for instance, the ability to alert the authorities when there’s been a crash.

The self-driving car systems are also highly integrated to give the car owners more options for controlling their vehicles. Tesla Model S owners can remotely turn on the air conditioning from a smartphone application, for example. However, this type of connectivity and control also make these systems more vulnerable to remote attacks.

The high integration also helps the Advanced Driving Assistance System (such as Tesla’s Autopilot) control every part of the car on its own, which means that an attacker could also gain that level of control over the vehicle.

Tesla fixes security in Model S after Chinese hack

Tesla said Wednesday that it had fixed a software vulnerability in its luxury electric Model S sedan after a Chinese security team hacked a car's systems and remotely controlled it.

The newest hacking case of remote manipulation of a car served to underscore the potential dangers of vehicles that are heavily connected via wireless technologies.

"Tesla has already deployed an over-the-air software update that addresses the potential security issues" of the hack, the company said in a statement.

To put things into perspective - MOST new cars have been hacked

How Hackable Is Your Car? Consult This Handy Chart

... Is your specific car vulnerable, too?

If you own a Cadillac Escalade, a Jeep Cherokee or an Infiniti Q50, you may not like the answer.

In a talk today at the Black Hat security conference in Las Vegas—and an accompanying 92-page paper—Valasek and Miller will present the results of a broad analysis of dozens of different car makes and models, assessing the vehicles’ schematics for the signs that hint at vulnerabilities to auto-focused hackers. The result is a kind of handbook of ratings and reviews of automobiles for the potential hackability of their networked components. “For 24 different cars, we examined how a remote attack might work,” says Valasek, director of vehicle security research at the security consultancy IOActive. “It really depends on the architecture: If you hack the radio, can you send messages to the brakes or the steering? And if you can, what can you do with them?”



All the cars’ ratings were based on three factors: The first was the size of their wireless “attack surface”—features like Bluetooth, Wi-Fi, cellular network connections, keyless entry systems, and even radio-readable tire pressure monitoring systems. Any of those radio connections could potentially be used by a hacker to find a security vulnerability and gain an initial foothold onto a car’s network. Second, they examined the vehicles’ network architecture, how much access those possible footholds offered to more critical systems steering and brakes. And third, Miller and Valasek assessed what they call the cars’ “cyberphysical” features: capabilities like automated braking, parking and lane assist that could transform a few spoofed digital commands into an actual out-of-control car. ...

https://www.wired.com/2016/03/survey-fi ... jeep-hack/
https://www.wired.com/2016/03/thousands ... n-hackers/
https://www.wired.com/2016/03/fbi-warns ... real-risk/
https://www.wired.com/2016/03/uber-bug-bounties/
https://www.wired.com/2016/08/jeep-hack ... ion-hacks/
https://www.wired.com/2016/08/oh-good-n ... lkswagens/

[vox_mundi]

What happens if heavy snow covers the sensors and landmarks?

Then Ya shouldn't be driving


Walk more. It's healthy.

Who pays if the software malfunctions and causes an accident?

The responsible party.

If it's software then NHTSA will exercise its authority to force the manufacturer to recall and repair vehicles that have faulty self-driving functions

Can these things be hacked?

See previous post

[Outcast_Searcher]

What happens if heavy snow covers the sensors and landmarks?

Who pays if the software malfunctions and causes an accident?

Can these things be hacked?

These are good points. It is many issues like this which will likely make fully autonomous cars a more complex and slower process to become widespread reality than the (self-serving) pundits in related industries are proclaiming.

However, that doesn't mean the shift won't happen. It just means that just because entity "X" proclaims a target date of "X", we shouldn't hold our collective breaths and bank on that date.

OTOH, the progress and the intensity of effort and the increasing interest and debate on the issue by a large chunk of society can't be simply ignored or dismissed, with any objectivity.

There's no reason this can't continue to make progress as it has -- on an evolutionary track, vs. a sudden one.

[Outcast_Searcher]

In which case the solution will be strengthening security rather than abandoning the technology.

+1

It's not like there aren't solutions. It's just that solutions will take effort, expense, and planning. Which is an example of why one key thing that needs to happen (which takes time) is serious policy and regulation and architecture to deal with self driving cars, overall.

Of course, this is no different than a legal framework in general, insurance issues, etc. All this needs to be dealt with, and at the glacial pace of government progress, it will likely take lots of time.

[C8]

Its not just hacking that is the danger- its computer malfunction of ANY kind- a computer crash, a loss of service, a power connection issue, software glitches.

These are all nuisances on a laptop sitting in a room

Going 70 mph on a mild curve overlooking a valley even a 1 second loss of control can mean death

I don't think people are really thinking this out- computers are simply too unstable for this level of risk

Its one thing on a jumbo jet where a well trained pilot (who is paid to pay attention) can recover in mid air before things get out of hand

Its entirely another thing when you are on a road and just a one second problem can kill you. There will be millions of cars sold with this software- sold to people who are not well trained, reading a book, and not paying attention. Does anybody really think this will end well? Look at all the glitches and crashes on computers every day.

I think this self driving world thing is going to hit cold reality as the cars move from the controlled tests to the real world

If millions of these things are made there will be serious quality issues and massive security breaches- its just how it is

This hyped up zeppelin is gonna crash

[ennui2]

And yet people seem to disregard how many human-caused crashes occur on a daily basis. And how about lower-tech recall issues like sudden-acceleration?

[vox_mundi]

Autonomous Cars’ Effect on Auto, Cyber Insurance

Autonomous vehicles are expected to shake up the property/casualty market, as personal auto insurance premiums drop and cyber insurance steps in.

... According to Aon, personal auto insurance currently accounts for 40 percent of global insurance premiums and a reduction in that could drive up market volatility in the property/casualty sector.

“The autonomous technology is going to be a catalyst for change in…how people purchase transportation,” said Stankard. The shift will be toward ridesharing/mobility services, hence a greater need for commercial auto insurance to cover fleet owners, he said.

As far as changes in claims, Stankard said that because the majority of car accidents are the result of human error – the National Highway Traffic Safety Administration estimates driver error is the cause of 94 percent of all accidents – he expects that there could be a nearly 90 percent drop in auto accidents, which will result in the processing of fewer claims. ... Liability may be more clear cut in these cases, said Stankard, since electronic data recorders will likely be more sophisticated, tracking a variety of data.

While the personal auto insurance market is expected to shrink, the cyber insurance market will likely step in to provide more coverage related to auto cyber claims, he said.

A Munich Re survey, conducted on site at the Risk and Insurance Management Society Conference in April 2016, found that 55 percent of risk managers feel that cyber security is the greatest insurance concern associated with autonomous vehicles. This was followed by allocation of liability when autonomous/non-autonomous vehicles share roads (27 percent), economic disruption (seven percent), safety (seven percent) and the cost of technology to repair damaged vehicles (four percent). From a cyber security perspective, respondents believe the greatest threats to autonomous vehicles are auto theft by an unknown individual hacking into and overtaking vehicle data systems (42 percent) and the failure of smart road infrastructure technologies (36 percent).

Volkswagen building new unit to protect autonomous cars

German automaker Volkswagen and three Israeli cybersecurity experts established a new automotive cybersecurity firm on Wednesday, called CYMOTIVE Technologies.

The company, based in Herzliya, Israel and Wolfsburg, Germany, is a response to the growing epidemic of connected cars on the road with insecure software that is readily hackable.

Tesla update halts automatic steering if driver inattentive

Tesla Motors says a software update to its Autopilot system will disable automatic steering if drivers don't keep their hands on the wheel.

The update also adds multiple features, including improved radar, better voice commands and an industry-first temperature control system that helps prevent kids and pets from overheating.

Tesla started moving the update to Model X SUV and Model S sedan owners Wednesday night over the internet.

Tesla says the software update should help avoid crashes, since it will enhance the radar system and make Tesla's vehicles rely more on radar signals—which can see through snow, bright sun and other weather conditions—than cameras. The new radar can detect braking in cars up to two lengths ahead and has a clearer picture of the road than the previous version. The company also redesigned its indicator lights to more clearly show when Autopilot is engaged.

Robot rides may force error-prone human motorists off the road

New rules of the road for robot cars coming out of Washington this week could lead to the eventual extinction of one of the defining archetypes of the past century: the human driver.

While banning people from driving may seem like something from a Kurt Vonnegut short story, it’s the logical endgame of a technology that could dramatically reduce -- or even eliminate -- the 1.25 million road deaths a year globally. Human error is the cause of 94 percent of roadway fatalities, U.S. safety regulators say, and robot drivers never get drunk, sleepy or distracted.

Autonomous cars already have “superhuman intelligence” that allows them to see around corners and avoid crashes, said Danny Shapiro, senior director of automotive at Nvidia Corp., a maker of high-speed processors for self-driving cars.

“Long term, these vehicles will drive better than any human possibly can,” Shapiro said. “We’re not there yet, but we will get there sooner than we believe.”

This week, technology industry veterans proposed a ban on human drivers on a 150-mile (241-kilometer) stretch of Interstate 5 from Seattle to Vancouver. Within five years, human driving could be outlawed in congested city centers like London, on college campuses and at airports, said Kristin Schondorf, executive director of automotive transportation at consultant EY.

... With mobile devices an added distraction, U.S. highway fatalities rose 8 percent last year, the biggest increase in 50 years. Some 38,300 people were killed on the road in 2015 and 4.4 million were seriously injured, according to the National Safety Council. Globally, 1.25 million people die in car crashes annually, according to the World Health Organization.

U.S. Annual Vehicle Deaths = 911 x 12
U.S. Annual Vehicle Injured = WWII wounded X 7

“Behind the wheel, we are only human and we are expected to screw up,” said Raj Rajkumar, co-director of the General Motors-Carnegie Mellon Autonomous Driving Collaborative Research Lab in Pittsburgh. “There will come a point in time where we should not be allowed to drive.”

8 big challenges remain for self-driving car makers

'Reinforcement Learning' is Teaching Robots Faster Than Ever

Robots are learning how to complete tasks in sped-up virtual worlds, developing skills in a matter of hours that might otherwise take months. Simulated deep reinforcement learning (or Deep RL) means a skill that would normally take 55 days for an A.I. to learn in the real world takes only a day in the hyper-accelerated classroom.

“It’s got the potential to really revolutionize what we can do in the robotics domain,” Raia Hadsell, a research scientist with Google DeepMind, said at the Re-Work Deep Learning Summit in London on Thursday. “We can learn human level skills.”

Cheap Lidar: The Key to Making Self-Driving Cars Affordable

... Many autonomous cars have relied on the HDL-64E lidar sensor from Silicon Valley–based Velodyne, which scans 2.2 million data points in its field of view each second and can pinpoint the location of objects up to 120 meters away with centimeter accuracy. But the sensor itself weighs more than 13 kilograms and costs US $80,000. This year, Velodyne announced the VLP-32A, which offers a 200-meter range in a 600-gram package. With a target cost of $500 (at automotive scale production), the VLP-32A would be two orders of magnitude cheaper than its predecessor but still too expensive to be integrated into driverless cars intended for the consumer market.

At the CES 2016 electronics show, Quanergy Systems, in Sunnyvale, Calif., demonstrated a prototype solid-state lidar sensor designed for driverless cars. It uses an optical phased array to steer laser pulses rather than a rotating system of mirrors and lenses. Quanergy projects that its sensor will cost $250 in volume production, and it should be available to automotive original equipment manufacturers in early 2017.

Meanwhile, two startups are working on $100 automotive lidar systems they both say will be released in 2018. ... MIT researchers have leveraged silicon photonics to condense a functional lidar system onto a single 0.5- by 6-millimeter chip that can be fabricated in commercial CMOS foundries. MIT’s prototype has a range of just a few meters, but there is a clear development path toward a 100-meter range and a per-chip cost of just $10.

Google’s self-driving car will automatically pull over for police

Can Self-Driving Cars Protect Black People From Police Violence?

Right now, for too many African Americans, the simple act of driving a car is a risky option whenever there is any kind of a police encounter. If automated vehicles can change and save so many lives, we should talk seriously about how they can protect black lives.

... The point here is that, while the car represents freedom and mobility for many people, it has become the locus of police-driven killings for too many African Americans, regardless of whether laws and safety guidelines are being followed. Philando Castile was wearing his seatbelt when police killed him. Terence Crutcher’s hands were visibly up when police shot him. In the Keith Lamont Scott case in Charlotte, even if Scott had a gun as the police say, North Carolina is an open carry state. If, in any of those scenarios, an automated car would have helped them avoid death, then maybe this new technology would truly be a game-changer.

Why Hire a Lawyer When a Robot Will Do?

[ennui2]

Human error is the cause of 94 percent of roadway fatalities, U.S. safety regulators say, and robot drivers never get drunk, sleepy or distracted.

Where is the indignation from the likes of PStarr over this? Oh, right, peak oil, end of happy motoring and suburbia...um...when will that be exactly?

Could it be the aversion to this tech isn't due to any genuine safety concerns but merely the role it may play in EXTENDING BAU??? Sorry to say that without this tech, we'll just get more of the same, more distracted drivers, more accidents, more dead kids who get run over while chasing balls (which is, well, not a very big statistic, but has become the poster-child scenario for self-driving car FUD).

[vox_mundi]

In 2005 Hirsch warned that we would need a paradigm shift in Transportation to weather Peak Oil. A 90% drop in fossil fuelled vechiles at least.

Strange how that's now the same goal of the autonomous vehicle industry.

[Outcast_Searcher]

Its not just hacking that is the danger- its computer malfunction of ANY kind- a computer crash, a loss of service, a power connection issue, software glitches.

These are all nuisances on a laptop sitting in a room

Going 70 mph on a mild curve overlooking a valley even a 1 second loss of control can mean death

I don't think people are really thinking this out- computers are simply too unstable for this level of risk

...

This hyped up zeppelin is gonna crash

No, you're not thinking this out. Ever hear of redundancy? We already know autonomous cars will have multiple inputs from things like cameras, radar, lidar, GPS, electronic signals of various kinds (and no telling what else, as the technology develops).

So you presume there can be no redundancy in the computer tech? Wrong.

NASA had that in the Apollo series, when computer tech cost tens of thousands of times what it does today (and it was thousands of times cruder and MUCH less reliable). That was 50ish years ago.

It won't be that big a deal to have, say, 3 or 5 master computers in the car, and take the vote of at least two when in doubt. And override a single manfunctioning signal. And given what the chips in a smart phone cost, it won't be all that expensive.

You keep acting like one challenge means the technology is impossible. If that were true, smart phones, databases, and even spreadsheets, which metaphorically cover the globe, wouldn't exist.

[C8]

I have considered redundancy- and its amazing how many redundant systems crash anyway. In fact, the complexity added by all the redundancy can be a source of the crash (and provide many, many extra entry points for hackers). There is nobody who depends on computers more than Delta Airlines- see what happened to them below

Why do computer crashes keep bringing major airlines to their knees, leaving hundreds of thousands of passengers stranded at airports? Human error. Mistakes. Good old fashioned screw ups.
That's the explanation offered by airline computer experts Monday, after Delta Air Lines scrambled to deal with a huge computer snafu. The world's second largest airline was forced to delay of all its flights on the ground for at least six hours worldwide.

Delta (DAL) blamed the problem on a power outage. But that alone should not have brought the system down -- there are backups that should have kept Delta's system up and running.

"You're basically saying, 'We had power failure in a location but unfortunately we were unable to continue operations from a secondary data center despite the fact that we spent hundreds of millions of dollars on it.," said Gil Hecht, founder and chief executive of Continuity Software, an expert in computer disaster recovery.
That is essentially an admission of human error, added Hecht.

A Delta spokesman wouldn't comment on whether the airline had a backup independent power supply -- but experts told CNN that it's certain the company has one.

It's not just Delta. These glitches happen a lot.
Monday's computer crash came about three weeks after a computer outage at Southwest Airlines (LUV), which led to the cancellation of more than 1,000 flights. In May, JetBlue (JBLU) computer issues forced passengers to be checked in manually at some airports. Computer problems delayed United Airlines flights worldwide in 2015.

Why do these airline computer failures keep happening?
"Complexity in the data center gets out of hand," Hecht said.

The airlines, by building layers upon layers of systems -- each one with a different configuration and a different purpose -- accidentally create the threat of something going down in their computer networks.

"Somehow, someone created a threat in the Delta Air Line situation that caused their disaster recovery not to work. How do I know it? Because their disaster recovery system should have worked. And it didn't."

Airline experts say there are three reasons why systems go down.

1) No redundancy. An airline might have chosen not to protect itself with a backup system. That's unlikely for a major carrier like Delta.

2) Hacking. The crash was caused by a malicious attacker. That's not likely the cause of Monday's Delta computer failure, said Hecht, because a malicious hack into Delta's system would probably have been isolated and the system would have been brought back to life more quickly, he said.

3) Human error. Layers and layers of systems that pile up over time create some kind of glitch and suddenly the whole thing comes crashing down. That's the most likely explanation for Delta.

Meanwhile we can't go back to paper check-ins during an emergency like this. It's just not feasible anymore, experts say -- especially on international flights -- because airline computers are linked to security networks like government No Fly Lists and visa document systems.

So what can airlines do to prevent these computer failures from happening so often?
They can install more automated checkup systems. They can perform emergency drills by taking their systems offline during slow periods and going to their secondary and backup systems to make sure they are working properly.

http://money.cnn.com/2016/08/08/technol ... r-failure/

Of course car navigation systems do not have the luxury of being taken off-line to troubleshoot problems- people will be using the system 24/7 and expect it to work continuously- any sudden take down of the system for diagnostics would lead to multiple crashes

The irony of all this is that we are discussing this on a website that repeatedly crashes!

[kublikhan]

I don't think people are really thinking this out- computers are simply too unstable for this level of risk

Compared to humans who get fatigued, drunk, texting, heart attack, etc? I admit computers can fail but the failure rate is much lower than the failure rate for humans. I am more concerned with how computers lack common sense and can't adjust to unexpected situations rather than a hardware failure.

If millions of these things are made there will be serious quality issues and massive security breaches- its just how it is

There are already millions of things on the road with serious quality issues and massive security breaches. That's what recalls are for.

There will be millions of cars sold with this software- sold to people who are not well trained, reading a book, and not paying attention. Does anybody really think this will end well?

This one I agree with you. The Tesla feature is not self drive but rather driver assist. Some blame here should fall on Tesla for overselling the feature. But also on people for using the feature in a way Tesla did not intend.

Who pays if the software malfunctions and causes an accident?

I am interested in this is well. I want to see how liability for crashes plays out here.

I think this self driving world thing is going to hit cold reality as the cars move from the controlled tests to the real world

I share your concerns. Vox_Mundi posted some articles on self-driving vehicles leaving their controlled routes and hitting real streets. It will be interesting to see how things play out. Although this test is a still an in-between stage test. A driver is present who could take over at any time, controlled routes, average speed of 6 mph, etc. Baby steps:

Helsinki is making headlines for hosting two of the world’s first self-driving buses. For the next month, a pair of electric-powered Easymile EZ-10 vehicles are carrying up to 12 public passengers along a fixed route in Helsinki’s Hernesaari neighborhood. The buses were previously tested on closed roads in the Netherlands and in a small Finnish town just north of Helsinki. But this trial—with autonomous buses carrying riders along public urban streets—is one of the first of its kind anywhere on the globe. Members of the public can hop on and off at pre-defined points along the route.

Finland is one of the only countries in the world that does not legally require every vehicle on public roadways to contain a driver. Because of this feature, the country is fast becoming a popular testing site for self-driving technology.

This test project is being coordinated by Helsinki’s Metropolia University of Applied Sciences with the aim of better understanding how the vehicles fare in real-world traffic. The buses are limited to traveling at an average speed of just 6 miles an hour (10 km/hour) to minimize risk as they share the open road with other cars and pedestrians for the first time.

Self-driving buses are now on the road in Helsinki

[toolpush]

Has anyone thought of what fun a human driver could have, being totally surrounded autonomous driven cars on a busy freeway. Where the human knows the autonomous driven cars are not allowed to run into anything.
Just imagine doing wild serves from one side of the freeway to the other, and watching the auto cars having heart attacks. By rights the human could just go anywhere, and the autos would just give way. Police presence on the roads should be a lot less as there would not be enough vehicles speeding or doing any other revenue generating violations, to make it worthwhile to spend their time on the road.
Fun times ahead for the devious thinker.