The autonomous (self driving) car buyers and owners thread (1 Viewer)

[Josh Steinberg]

I'd hope for the latter. Your teacher sounds wrong to me on many levels. Absolute devotion to any rule is a bad philosophy IMO.

Agreed. In that scenario, I'd be quite comfortable being a living lawbreaker rather than a law-abiding piece of roadkill

 

[DaveF]

It actually reminds me a little of an argument I had with my driver's ed teacher back in the day. I don't remember the exact situation described, but basically the teacher was explaining how the rules of the road limit the circumstances when you can and can't change lanes, etc., and about properly obeying those rules under all circumstances. The scenario was something like this: let's say you're on a highway that goes in both directions, and a car starts coming at you the wrong way -- driving in the wrong direction in your lane towards you. If you turn to your right, you'll miss being hit, but you'll probably crash into the wall on the side of the highway which will almost certainly result in catastrophic damage to your vehicle. If you turn to your left, you'll be swerving into oncoming traffic as that's the lane going in the other direction. But in the example as the teacher presented it, for whatever reason, there was no one on the road going in the other direction. The teacher was adament that there was only one proper thing to do - to crash into the right side, because you're never allowed to drive the wrong direction on a highway. I get where she's coming from in theory. But in practice, who would do that? In real life, in a scenario where you only have a second to decide, I think most people would go with the illegal move (driving in the wrong lane for five seconds) over the legal move (crashing into a wall, possibly destroying your car, possibly getting severely injured or killed in the process).

The reasoning behind this that I've heard: the driver realizes their impending crash and swerves back into their lane to avoid. If you went into the wrong lane to avoid the accident, you both inadvertently steer into each other, and have a head-on collision in what is for you oncoming traffic. The police come to the scene, and you're to blame. You are in the wrong lane. In the case of night accident (the other driving was drifting asleep) with no witnesses, its your word against theirs, and the plain scene shows you're in the wrong. So don't veer into the other lane. It could make things worse.

If someone actually knows something about such things, I'd love to know if it could really play out that way.

What would the automatic car do in that situation? Would it be programmed to follow the rules of the road at all costs, or would it allow for an illegal move that was unquestionably the safest solution to a momentary danger?

Computerized cars avoiding accidents is intriguing. I propose we don't know what will happen, and it may not be obvious. Humans panic and make reflexive decisions. Computers do not. They have available sensor data and their algorithms. The results may not look like human behavior.

Let's add to that computer games: I recall descriptions that IBM's Deep Blue played, at times, with strange and atypical strategies, against Kasporov. That it's compute algorithms lead it down paths human players didn't take.

The new oddness now is computers analyzing food chemicals and devising new recipes: sometimes coming up with oddities not yet tried by sensible humans.

So. A self-driving car. Supremely capable of number crunching physics simulations. Exceptional knowledge of the mechanics of itself, the road. Doesn't panic.

Might it play billiards with the car? Intentionally ricochet off a concrete barrier, causing some harm, but avoiding greater damage? Maybe it could start a drift, aimed so the head-on collision converts the energy into rotation velocity, minimizing harm to the car and driver? Maybe, along with hitting the brakes, modifies the ABS while simultaneously depressing the tires to maximize tire footprint and gears into Park to stop the rotation completely, computing it can stop a critical few feet sooner, and avoid the accident.

I intrigued by the idea that the cars aren't programmed with simple, deterministic algorithms like a 1980s racing game. Instead, they may use complex heuristics and come up with unexpected results.

 

[Josh Steinberg]

The reasoning behind this that I've heard: the driver realizes their impending crash and swerves back into their lane to avoid. If you went into the wrong lane to avoid the accident, you both inadvertently steer into each other, and have a head-on collision in what is for you oncoming traffic. The police come to the scene, and you're to blame. You are in the wrong lane. In the case of night accident (the other driving was drifting asleep) with no witnesses, its your word against theirs, and the plain scene shows you're in the wrong. So don't veer into the other lane. It could make things worse.

If someone actually knows something about such things, I'd love to know if it could really play out that way.

I think that may have been why the teacher explained that that was the only proper thing to do. And I mean, in theory, I get what she's saying. It was the absolute rigidity in her explaination that bothered me. I think when driving you should follow the laws and rules of the road, that's a no-brainer -- but when you're faced with other people sharing the road and not following the law, I believe that sometimes, as much as it might not be the first choice of actions or a desireable course of action, sometimes you have to do something that's technically illegal to avoid something worth.

Like you say... they don't want you swerving into the long lane because the driver may self-correct at the last moment, and end up causing a crash that way, and that makes sense. On the other hand, the point I was trying to make with the teacher is that the best practice may work 99.99 percent of the time, but there could be that 0.01 percent of the times where common sense and survival may depend on doing something that goes against proper procedure. What bothered me was that my teacher was unwilling to even contemplate that that was possible. She was very much, "If it's written in the book this way, then that's the correct answer, and there's not one bit of new evidence or perspective I am even willing to consider for discussion." I get worried when people adhere so rigidly to things that even the conseration of a different idea is forbidden.

In the case of an automatic car, I'd hope that it would know enough to respect the law, to try to always follow the law, but was also given some kind of internal values system so that it would known that all things being equal, that it was better to break the law to preserve lives than it would be to follow the law even if it will result in injury. I hope concepts like that are at least discussed when they're working on designing these things. I mean, I don't want each car to go around creating their own traffic rules, but I guess I want them to know that there may be an occasion where the safest possible solution to an unsafe situation is doing something that's technically forbidden, and in that case, to not get hung up on the rules.

 

[Sam Posten]

The one thing you guys are forgetting: These cars will not be entirely independent of one another, presumably there will be mechanisms where cars can talk to EACH OTHER to avoid accidents between themselves in emergencies, flashing their intended moves before they put them into action and faster than humans can react.

 

[DaveF]

That's only for the long-term situation of majority self-driving cars. A big unknown, as you linked earlier, is how they behave as the minority among human drivers.

 

[BrianW]

The rules of the road are there to keep you safe. When following them puts you in peril, then it's time to do something else.

 

[Sam Posten]

Squirrel!

Avoid creaming squirrels: While the car picks up pedestrians who may jaywalk and deer that could bolt across the road, squirrels are still too small for its sensors. The team is constantly working to pick up more and more detail, but hasn’t “done a squirrel test,” Urmson said.

http://www.wired.com/2014/05/google-self-driving-car-can-cant/

 

[KevinGress]

The one thing you guys are forgetting: These cars will not be entirely independent of one another, presumably there will be mechanisms where cars can talk to EACH OTHER to avoid accidents between themselves in emergencies, flashing their intended moves before they put them into action and faster than humans can react.

It will be interesting to see this process develop, and how it matures. (and I restate - it shouldn't be mandated...) This 'talking' will do a lot to avoid most accidents - adjust speed to maintain distance; change lanes; telegraph intentions, etc. -- basically do the same types of things that human drivers do now. I guess I fear that too many people will think that because vehicles can 'talk' to one another, that there won't be any more accidents.

Remember, a vehicle is using its processing power on navigating. Now, it's got to 'talk' to other vehicles - which will take processing power. It's got to analyze and report and wait for the other vehicle to receive and analyze and report back - all taking processing power, and all taking TIME.

I'm not saying it can't be done, but it will be interesting to see how they tackling these challenges enforced by physics.

 

[Sam Posten]

Well said Kevin, thanks!

Google takes a huge next step!!! No more human drivers at all as a test.
http://googleblog.blogspot.com/2014/05/just-press-go-designing-self-driving.html

We’re now exploring what fully self-driving vehicles would look like by building some prototypes; they’ll be designed to operate safely and autonomously without requiring human intervention. They won’t have a steering wheel, accelerator pedal, or brake pedal… because they don’t need them. Our software and sensors do all the work.

https://www.youtube.com/watch?v=CqSDWoAhvLU

 

[Sam Posten]

Tech Crunch has an amazing slideshow of the thought process behind this car, go look please if you care at all about this stuff:
http://techcrunch.com/gallery/google-self-driving-car/#/slide1

 

[Sam Posten]

Yes:
https://www.linkedin.com/today/post/article/20140528072025-142059068-don-t-laugh-the-new-google-prototype-car-has-implications-for-your-business

The reason that this is such a big deal for your business is because the simple fact that a car can drive itself is going to change how and where we all live and work. The vehicle can dynamically position itself to the point of need and then do work by moving people or goods to where we instruct. We will have a limitless supply of on-demand chauffeurs, or fully automated taxis if you prefer.

We therefore don’t really need to own a vehicle anymore if we don’t want to; or even own several as the automaker’s business model would prefer. Rather than have our multi-purpose vehicle sit idle some 90% of the day, we can hire a vehicle suitable for our need; a shared self-driving electric 2-seater for our commute (and work on the way); an SUV for the group trip at the weekend.

Some will prefer to live further from the city as the commute can now become part of their working day. Some will find it even easier to live in the urban center as they need never own a car again. Sprawl and urban densification at the same time; that will create enormous money flows from many existing business models to new business models. The impacts on land values alone could be transformative.

YES

Finally, if you are still skeptical about this whole subject and really aren’t concerned for your business, then please consider what Warren Buffet has to say given his significant Geico investment:

“That is a real threat to the auto insurance industry,” Buffett said at his company’s annual meeting here Saturday. “If [self-driving cars] prove successful and reduce accidents dramatically, it will be very good for society and very bad for auto insurers.”

Warren’s not laughing, and neither should you be.

And Yes:
http://www.theverge.com/2014/5/28/5758734/uber-will-eventually-replace-all-its-drivers-with-self-driving-cars

 

[Chuck Anstey]

I think the true self-driving car is still a long way off. Google has come close to the easy 80% of the problem but it is the last 20% that is always the hard part that takes 10 times the effort as the first 80%. Those cars require hyper-accurate GPS road maps and require GPS positioning, a cheat if you will. Imagine towns and states trying to create new roads or remove an existing road, fix a bridge, and the like. Right now it is a very real possibility these cars would drive right off an unfinished or partially torn down bridge. New construction of roadways will be really difficult as all the right paperwork, GPS road map and such will have to be filed and then pushed out to all the cars before any changes in roads is allowed.

These Google cars don't seem like true self-drive but rather slot cars that follow a pre-mapped route and only use their sensors to avoid common problems on the roads. If the map is wrong, the car might crash. Simple natural disasters that take out roads and bridges turn the cars into deathtraps. Also in this self-drive future, the fastest way for another country to kill a lot of Americans and cause untold amounts of economic damage is to shoot down 4 or 5 GPS satellites at once.

Right now a bad driver is a danger to themselves and those around 100% of the time but their danger to others can be mitigated somewhat by attentive drivers. In the future, a mistake in a map or a washed out bridge will be a danger to 100% of the people all the time.

 

[Sam Posten]

Planting a very mediocre first step is intel:http://venturebeat.com/2014/05/29/intel-inside-your-car-chip-giant-unveils-its-technology-for-autonomous-cars/

 

[Sam Posten]

Something seems totally off about this, too much of a hard sell maybe?https://www.indiegogo.com/projects/solar-roadways#homeIf they pull it off tho....

 

[Sam Posten]

Bwahahahahah!http://blog.caranddriver.com/5-reasons-you-should-fear-the-google-autonomous-car/

 

[Chuck Anstey]

Bwahahahahah!http://blog.caranddriver.com/5-reasons-you-should-fear-the-google-autonomous-car/

Reasons 4 and 5 (When there is one Google car, all must be Google cars AND What happens when it crashes). To me there is no doubt that the ability of every adult to have access to a vehicle to take them anywhere regardless of any physical or mental handicap that would prevent them from driving themselves is a very big deal and great for society as a whole. I do strongly worry the cost of that will be too high except no one is going to see the true cost to stop it or mitigate those costs. Stupidly simple ones are like the blog brings up: forced advertising while riding in the car and it tracks your every move and that data will be used by some 3rd party for something negative. Bigger ones like 'Road non-neutrality' have been discussed here a bit.

 

[KevinGress]

Something seems totally off about this, too much of a hard sell maybe?https://www.indiegogo.com/projects/solar-roadways#homeIf they pull it off tho....

Sam, you've got to quit posting so many thought-provoking, broad idea links!!! I can't read, then think, then post AND get all my work done!

I saw this last night and had some doubts of my own, and then I read a rebuttal. While a neat idea, it's fundamentally flawed - it'd be too expensive and wouldn't deliver on its promises. Better would be to continually improve solar panels that AREN'T driven on and improve battery capacity.

More comments on some of the other links later.

 

[Chuck Anstey]

Solar Roadways: They never bothered to actually compute the most important number of the whole thing; How much energy will be consumed to create and maintain the roadway over its lifetime? All alternative energy source only talk in dollar cost but that is the wrong unit. If it takes twice as much energy to produce and maintain a solar panel as it will produce in its lifetime, we are far worse off. And break even doesn't really help much either in most situations unless you are able to show that the traditional way of making and maintaining a road consumes something on the order of the same amount of energy.

The second question to answer is an even more important question that must be answered for any energy source; What do we do with the product and its waste once it has reached its life expectancy?

The final question that is the most important of all; What happens when things go way bad (full and complete catastrophic failure), as they always will somewhere at some point? No one ever asks this question, which is why nuclear is still on the table even though it fails this and the previous two questions. Fortunately solar roadway looks like it doesn't have a problem if it has a catastrophic failure unless there is a risk of extremely high energy electrocuting or incinerating motorists.

 

[DaveF]

Haven't read the latest links, but one had the provocative idea that autonomous cars will reduce car ownership, since cars will be available conveniently on demand.

 

[Sam Posten]

Sam, you've got to quit posting so many thought-provoking, broad idea links!!! I can't read, then think, then post AND get all my work done!

Glad you enjoy em Kevin, I know my links aren't everyone's cup of tea but I hope to good severely outweighs the bad.