Let me fasten that seatbelt for you

April 25, 2020 10:50 AM

How Cornell Disguised Drivers to Gauge the Public's Reaction to Autonomous Vehicles — It may look and smell like a prank, but dressing someone up as a car seat is a serious endeavor. A team from Cornell Tech led a study in which a person was disguised as a car seat while driving through several international cities to gauge the public's reactions to driverless vehicles.
posted by cenoxo (27 comments total)

The same thing was done in the opening stunt scene of one of the Austin Powers movies. Take a close look at that drivers seat in the Jaguar.
posted by smcameron at 11:19 AM on April 25, 2020 [9 favorites]

I remember that video faking out the autonomous vehicle crowd for a few internet cycles, there's probably a few tweets amazed how far Cornell's program jumped ahead of the SDC pack. Then deep internet groans.
posted by sammyo at 11:21 AM on April 25, 2020

Having the actual driver being disguised as a car seat in movies is pretty common. Especially when you need to film the actor driving the car in a stunt and you need the empty passenger seat in the frame. That way the actor can act, and an expert can drive the car without killing/hurting people.
posted by sideshow at 11:48 AM on April 25, 2020 [3 favorites]

Here is a fantastic gif showing how a particularly impressive shot in the Raid 2 was filmed, including a cameraperson disguised as a front passenger seat.
posted by chappell, ambrose at 12:10 PM on April 25, 2020 [12 favorites]

Wasn't there a news story from a couple of years ago where police pulled over what appeared to be an autonomous car, only to find out the driver was costumed in the same way as this study? Google apparently has no memory of it.
posted by Popular Ethics at 12:34 PM on April 25, 2020

Oh, this caused quite the minor scandal in the D.C. suburbs a few years ago. The best was when the local news reporter caught up with the car at a stoplight and tried to interview the seat. "I'm with the news, dude," was a catch-phrase for like two minutes.
posted by whitewall at 12:40 PM on April 25, 2020 [6 favorites]

Um, someone missed out on 2014 YouTube:
Driverless Drive-Thru
posted by bartleby at 12:47 PM on April 25, 2020 [5 favorites]

I've had a suspicion about this for years now. This is based on nothing but my instinct and experiences with other sorta-similar things. But here is is:

Driverless vehicles will be targets of vandalism, pranks, theft and even worse. I think putting a human being in a car (or any situation, really) leads to others reacting in a more friendly, or less hostile way in general. Obviously, road rage is a thing. But truthfully, most road rage consists of someone yelling and nothing more.

I keep thinking driverless vehicles will be targets for all kinds of stuff, up to and including outright destruction. Yeah, yeah... there will be cameras and other deterrents, but there's plenty of malcontents out there who will treat these things as targets.
posted by SoberHighland at 12:57 PM on April 25, 2020 [3 favorites]

Life imitates Knight Rider
posted by zombiedance at 12:58 PM on April 25, 2020 [3 favorites]

Yeah, yeah... there will be cameras and other deterrents, but there's plenty of malcontents out there who will treat these things as targets.

Doubly so when those malcontents are the gig economy drivers who didn't realize the companies they were contracting to were using their labor to invest in driverless technology with intent to replace them.

At a certain point, that will get figured out, and the people who lost their jobs to this will make it known exactly how much they are upset about it. Especially considering truck driving is one of the most populous jobs in the US.

I'm already imagining the inevitable first heist of a driverless truck full of some fancy new product and the hijackers get away mostly because of the lack of a real human directly involved. All devices that can transmit data like audio and video and GPS can have that transmission interrupted, and hijackers will have more desire to completely destroy anything that could contain evidence, because well, murdering a machine to make it forget it ever saw you is materially different to murdering a human driver to make them forget they saw you. I mean, at least in the eyes of the law and the courts, that falls under vandalization and property destruction, but not murder.

The reduction of the likelihood of involving or harming another human in the committing of a crime will also lead to them being targets of crime, simply because the potential of the crime to become a "violent" crime is essentially mitigated, as there is no one to commit violence against.

I can also imagine simpler crimes, like small time criminals stripping non-essential wiring from inside a driverless vehicle to be sold for scrap. Or any easily removable non-essential part, really.

The idea that any of that can be prevented by sensors that can be bypassed, interrupted, or destroyed seems naive to me.
posted by deadaluspark at 1:45 PM on April 25, 2020 [3 favorites]

Are we still pursuing this autonomous vehicle nonsense. Urgh.
posted by GallonOfAlan at 2:42 PM on April 25, 2020 [3 favorites]

Having a human (or a facsimile thereof) gives me something to focus my rage upon when they inevitably fuck something up.

And they will. It’s just human nature that, no matter what we do, we will fuck something up. Maybe in a small way, maybe in a large way, but there is no way we will ever not fuck it up.

That is axiom.
posted by drivingmenuts at 4:09 PM on April 25, 2020 [3 favorites]

In addition to the aforementioned Austin Powers 2 and The Raid 2 stunts using the "driver inside the seat" trick, I think Breaking Bad used it for this stunt in S04E05 "Shotgun".
posted by JauntyFedora at 4:45 PM on April 25, 2020

Ironically, one of the blocks that faces the adoption of fully autonomous vehicles, is their inability to interact with human drivers. My Tesla has few problems when it comes to rapidly picking out cars, pedestrians, kerbs, cones, traffic lights and an ever increasing set of objects. It knows where it is and where it is going, it can pick out the most plausible route to get there; it learns from the successes and problems of other cars that have been there before. But it can't make eye contact with other motorists.

Eye contact is not important for most of the time: but sometimes it is essential: negotiating progress on a narrow road with cars parked on either side, or entering a roundabout in busy traffic, for example. These are moments when drivers signal to each other with hand gestures or even with glances. If an autonomous vehicle gets involved in such circumstances then human drivers may not be able to tell that it is autonomous at all (because there could be a passenger in the driver's seat). If there is nobody in the car - then how can they signal to it - and how they know that that signal has been received and understood?
posted by rongorongo at 12:45 AM on April 26, 2020 [2 favorites]

smcameron: "The same thing was done in the opening stunt scene of one of the Austin Powers movies . Take a close look at that drivers seat in the Jaguar."

They've disguised a humorless putz as Austin Powers!
posted by chavenet at 2:08 AM on April 26, 2020

Could someone please tell me what is the point of driverless cars?

People who drive taxis and trucks will be unemployed. If every vehicle is replaced by a driverless car, perhaps with only one person riding each car just like now, cities and streets will be just as crowded. Shouldn't the tech geniuses in California concentrate on developing a better energy source for transportation, so we can stop using oil and gas as soon as possible? If every driverless car runs on gas, what have we gained?
posted by Termite at 3:26 AM on April 26, 2020 [1 favorite]

Putting Autonomous Driving Back on the Road – Founders, executives, and analysts agree that self-driving car companies are in for a bumpy ride., Bloomberg Hyperdrive, Ira Boudway; 4/21/2020:

The self-driving car was a dream deferred even before the Covid-19 outbreak brought the U.S. economy to a standstill. By early 2020, plans to launch robotaxi services had been pushed back or scaled down as excitement over driverless cars gave way to recognition of the long work and heavy expense of bringing the technology to market. Now the pandemic has forced companies to pull their cars from the road, send engineers to work from home, and look carefully at their balance sheets.

While some companies are taking the opportunity to demonstrate the promise of autonomous vehicles for contact-free delivery, these efforts serve mostly to show how far the industry remains from large-scale, truly driverless deployments. Hyperdrive spoke with a handful of industry consultants and executives about what’s ahead. There is general consensus that autonomous vehicle developers are in for a bumpy ride, and that consolidation is inevitable. But there is also widespread confidence that the project of building self-driving cars is more urgent than ever...

In the COVID-19 Era, you’ll need a safe, sealed, autonomous cocoon from your hermetic home to anywhere else you need to go.
posted by cenoxo at 6:34 AM on April 26, 2020

Could someone please tell me what is the point of driverless cars?

Like that mountain, climb it because she can.

The DARPA Grand Challenge just got a lot of research folks excited when after a dismal initial failure some researchers succeed.

Like manually running a loom for a 14 hour shift, driving work is just terrible.

Because of the Grand Challenge tech folks and venture capitalists just got really excited.

(cough) profit (as in insanely huge bigger than the internet massive profits)

Personally I think that the follow on effects of the development of great sensors at economy of scale costs will change the world in vast ways. Car accidents will be eliminated, as will factory death and injury when all the equipment become 'smart'. Long way to go but the trigger of semiconductor LIDAR at the right price point will become vastly useful. SDC's are kickstarting something much bigger. Think more pervasive than cell phone.

None of the researchers or VC's predicted length of edge conditions. SDC's actually work now at a 90-99% level, that last percent turns out to be really really hard.

Um... it's just really cool.

People really like it, read that the very first google alpha testers that were allowed to use one for a commute immediate began reading on the drive. Yes a bad idea but they liked to not have hands on wheel. Even with a very few tragedies folks really seem happy with the limited Tesla autopilot, a lot of miles hands free.

Efficiency of the supply chain will be tightened, most truck have to just stop for hours a day, tune network of trucking to optimize time, or cost of fuel, or arriving at exactly the right moment.

My thought on transit, rush hour surges are not entirely predictable, wouldn't be great if a filled bus triggered an added bus, perhaps added midway on a route? Staging manned buses is obviously impractical but a few 'hot spare' robot buses could make commutes great.

With all the uber and truckdrivers out of work it'll trigger the universal basic income, robot revolution and Fully Automated Luxury Gay Space Communism! Yea.
posted by sammyo at 9:49 AM on April 26, 2020 [1 favorite]

@deadaluspark: Legislative escalation; the changing of laws to make the penalties for such crimes, in the protection of private property commonly held by corporations and the wealthy, will exacerbate existing institutionalized biases towards the ultra-rich, too. I have a feeling that such a legal shift would quickly follow the popular adoption of driverless vehicles. This also helps to continue to build precedence for unequal protection under the law, something which is already the status quo. The rich will automate every kind of labor they can, so as to create some kind of rentier-totalist economy.
posted by constantinescharity at 10:55 AM on April 26, 2020 [1 favorite]

Why driverless cars won’t work.
posted by Everyone Expects The Spanish Influenza at 12:37 PM on April 26, 2020 [1 favorite]

The Horseless Carriage Means Trouble, New York Times, H. B. Brown [WP], February 1908 (reprinted March 25, 1973)

With the advent of the automobile the courts were confronted with the proposition that a self propelled vehicle (would be] limited to no part of the highway, capable of the speed of an express train, and attended by cloud of dust and smoke, and the emission of a noisome odor.

Notwithstanding these objections, automobiles have doubtless done much to earn their popularity. They have brought suburban towns within easy access from the city; they do not run upon a fixed track, and have no monopoly of any part of the highway; they do not seriously interfere with its use by other vehicles, and afford a most convenient and expeditious method of traveling between cities and outlying villages or country seats. In the form of electric runabouts, doctor's coupes, express and delivery wagons, and other teaming, they are rapidly superseding vehicles drawn by horses. They have largely taken the place of traveling carriages with those who are desirous of speed, and are content with little more than a perfunctory view of the scenery, which, however, cannot be thoroughly “taken in” when running at a rate of over twelve miles an hour....

Who knew then — only 112 years ago — how much trouble (and transformative) they would really be?
posted by cenoxo at 2:21 PM on April 26, 2020 [3 favorites]

Could someone please tell me what is the point of driverless cars?

I think the arguments in favour could be summarised as:
1. Liberty from drudgery.
2. Costs saved by not having to employ a driver.
3. Attractions of having fewer cars.
4. Improved safety for passengers and pedestrians.

Of these:
1. I am unconvinced that people find driving all that terrible - on the contrary, most of us seem to regard it as a positive flow experience.
2. There are lots of employers who would like to save on their wage bills for drivers - but drivers do more than just drive: they help cement relationships, they sell to customers, they guard valuable goods and so on.
3. Most of us can probably appreciate the Covid-19 lockdown experience of fewer cars right now. There are lots of attractions in terms of reduced noise, car-dedicated space or pollution. The problem is that that there is a very big industry aimed at selling the world cars and other vehicles. That industry probably does not want to shrink itself.

Which leaves "Safety": the idea is that fully autonomous cars would have fewer accidents is probably true. But it is quite a hard benefit to turn into a marketing opportunity: people don't believe they are personally at that much of a risk when they are behind the wheel. Others, such as governments or employers, may look at the statistics and disagree - but still be willing to live with the status quo.

But the real contribution towards safety might not be one requiring the full elimination of edge cases that is necessary for full autonomy. For example Tesla claim that driver miles completed with its "autopilot" driver assistance feature engaged are 9 times less likely to cause fatalities versus those done without it. The features designed to move towards autonomy help drivers who would be about to change lanes into incoming traffic or doze asleep at the wheel or fail to spot a cyclist at a junction.

This is important because these are benefits that drivers (and passengers and pedestrians) can readily appreciate - also ones that can feed into cheaper insurance (note Tesla's entry into this field).
posted by rongorongo at 12:05 AM on April 27, 2020

1. I am unconvinced that people find driving all that terrible

Speak for yourself; as someone who used to have an hour-and-a-half commute through a boring interstate (no, there wasn't a public transport option available), I would have been only too happy to play my DS or read during that time had self-driving cars been safe enough to not need regular attention by the rider.

No doubt you're one of those people who insists on driving a manual transmission because you find having to constantly change gear ratios during city driving somehow thrilling. I'm not going to kinkshame, but don't assume everyone else is the same.
posted by Wandering Idiot at 1:34 AM on April 27, 2020

Some driving is terrible (commuting to work, running errands in traffic) and some driving is fun (road trips, Sunday afternoon drives, etc.). In conclusion, driving is a land of contrasts.
posted by LizBoBiz at 1:48 AM on April 27, 2020

Why driverless cars won’t work.
Why man will never fly.

The linked article is ancient history (tech scale) right after the Uber accident that killed a street person. Which is a good example, a tragic accident that would not have made local news if not for the SDC involvement was a poorly planed trial from the least responsible player who disabled code that may have prevented the accident.

Why are driverless cars taking so long.

1) Not so long for hardware.
2) Pesky edge conditions
3) Hardware, cost effective sensors
4) Uber
5) Society is harder than silicon valley expected, "fail fast' is not always the best principle

Building hardware just takes longer than software, a new blender could take years of development, just a different time frame.

Yeah roads are a mess, everywhere. The best google car and handle most situations but a rush hour drive from midtown to JFK on surface roads, or taking that quarter mile dirt road that saves 10 miles, people are still smarter.

I think sensors are the biggest actual reason, a spinning $70k delicate high rpm laser is not a production option and reliable cheep semiconductor LIDAR chips are just now entering a few limited beta tests. The computer is also an issue, it'll need to be at least as powerful as a good game level pc and much more reliable. The Tesla advanced cruise control cpu has several custom chips that are optimized for the purpose. No one knows what Wamo (google) is doing under the covers.

The Uber accident probably saved some lives and cost VC's a few billion dollars and was a wakeup call to that industry that they needed to tighten test protocols.

The reason the Uber team commented out the emergency stop code that may have prevented the Tempe accident was due to jerky behavior. I've read complaints about the Waymo cars in Chandler AZ that they can be annoying to follow and braked too soon and too often. Eye contact between drivers is a real thing that will make robot cars frustrating. Just think of how annoying it is to be behind an older driver following the law precisely. Getting robot cars to work is done, they work. Getting robot cars to "fit in" to society and play well human drivers, perhaps the ultimate edge criteria.
posted by sammyo at 9:47 AM on April 27, 2020

I think putting a human being in a car (or any situation, really) leads to others reacting in a more friendly, or less hostile way in general. Obviously, road rage is a thing. But truthfully, most road rage consists of someone yelling and nothing more.

Gosh, my instinct is exactly the reverse. Road rage is all about abusing and provoking another person. There's no point in screaming at a robot which will ignore you, there's no one inside to hurt.
posted by Western Infidels at 10:35 AM on April 27, 2020

If people are interesting in seeing what happens with current "full self drive" autopilot - then I'd recommend the videos made by various Tesla drivers who demonstrate its capabilities and failings. For example here we have DirtyTesla demonstrate the just-released capability of the car to detect stop signs and traffic lights. In a short video we can see the car accomplish some impressive feats - but also show several examples of behaviour where it would have caused an accident without there being a human to take over.

Tesla has armed itself with an FSD board capable of 144 trillion operations per second - as well as an array of cheap (ie non LIDAR) sensors. If you drive past a set of traffic cones fast, at night at night and in the rain - then the car will show you the traffic cones on its display - with the overall number and precise location of each being mapped; the thing is a sort of Rainman of traffic cones.

Tesla's AI team have talked about "chasing the 9s" in terms of the car's handling of edge cases - the need to be "99.99...%" accurate means that they have to be good at dealing not only with cones - but with a wide range of rare events: overturned cars, detaching bike racks, flocks of ducks on a minor road.... Those features which get anywhere near the required level of reliability, will have got there will many millions of miles of collected real-world data.
posted by rongorongo at 12:13 AM on April 28, 2020