Asymmetrical Information

Do you really think that it will take 50 years? I was hoping for autopilot on limited access highways in about 20 years. Mostly because it would still be useful to me then. In 50 years I'll just make my grandchildren drive me on the long trips.

The really tricky part, I think, are the highways (e.g. I86 which is also state route 17 in NY) which are limited access in parts and have lights and intersections in others, with no warning other than the concrete barriers disappearing.

I worked for a startup that was writing code to handle flocking behaviour in cars. The core issues aren't that hard - it is the edge cases (unexpected cow in the road: how does an emergency response by one car propagate through the flock? What are the differences between road conditions in NJ and TN, and how does that interact with ice?) We died in the dot.com mess, but I know that at least two other companies are working on the same code base. The big problems are (1) testing, which is really expensive, and (2) liability, insurance and regulation, which are really, really expensive.

Almost sounds like train service.

The army has been experimenting with robot convoys for a few years. Basically you have one human driver in the lead vehicle and the rest of the convoy follows. A much simpler issue that what Jane's talking about, but a start nonetheless.

If cars drive themselves, door to door taxi service could be cheap enough to replace buses for public transit. Since the major reason people do not use public transit is the inconvenants,it would increase the use of public transit, and in congested cities subway systems would expand and parking lots would go the way of buggy whips. It could have as almost much effect on cities as the invention of the car did.

Such a development is all but inevitable as maximum potential lifespan continues to grow & risk-aversion increases. I still want to know where the 21st century air-car is, though...

Forget that Jane, where's my jet-pack! I want a personal jet-pack now!

"fishbane" hit on the reason that I don't see this happening (at least not here in the US): liability.

Any company that produced such a car would be sued over each and every accident that occurred.

Too bad.

A (very) modest step toward what you desribe already exists in some expensive cars such as high-end Mercedes and BMW's. They have a so-called "active" cruise control which varies the vehicles' speeds using radar, in order to maintain proper spacing from vehicles in front.

It's a lot less than fifty years. Due to liability concerns, though, automated driving will only occur when it can be sold as being safer than manual driving. That's not as unlikely as it sounds: an automated car can easily have better sensors than a human, and better response time. Getting software better at obstacle avoidance than the brachiating monkey brain that we inherited would seem to be the big trick.

Military applications are an obvious first step. Lower liability concerns, and way more of the army drives trucks than drives tanks. Plus, if an IED blows the hell out of an automated truck, no one has to write a letter to its mom.

Jane: I don't believe that you don't understand how your toaster works. Even its insides.

When was the last time your desktop computer crashed? To assume that in some halcyon future computers will be immune to malfunction is to stretch techno-hubris an uncomfortably long way. Of course cars are packed with computers already but their malfunction is usually more injurious to your wallet rather than your tender tookus. It is absolutely correct that the legal sharks will begin feeding the first time a computerized braking system, a rollover prevention system or any other gadget that puts the driver at a further remove from the vehicles behavior goes on the fritz and kills or injures someone. Merely having the techological ability to manufacture self-driving cars does not make it a good idea. The self-drive system would have to be lot's smarter and faster than a human and dead perfect reliable. 50 years may not be enough.

Mark -- there would certainly have to be some failsafes. For example, there should always be a competent driver who can take over, and if a separate device doesn't get the "it's working" signal, it should alert the driver to take over. It's true that parts can fail, but the relevant question is whether they would fail *more often* than regular drivers. Though, I will concede that legislators won't allow them unless they can achieve impossible 100% reliability, and juries will award far higher damages to robot crashers than human crashers.

Personally -- if we can just get truckers to *try* to merge onto a freeway faster than 40 mph, I'll be happy.

It is absolutely correct that the legal sharks will begin feeding the first time a computerized braking system

You do realize that your anti-lock braking system, your power steering, your automatic transmission, possibly your suspension, and pretty much all of your car's instrumentation are computerized, right? There are already a dozen ways that a software error in your car could kill you. Reliable software for embedded systems is really expensive to create, but well within current capabilities, and getting cheaper everyday. Reliable software for general purpose systems is another problem entirely.

It's probably only 20 years out. The technology is already able to do what's needed, so it's just a matter of popularizing it. Reliability won't be an issue because the system will be engineered by people who konw what they're doing rather than dipshit losers at Microsoft.

This seems way too conservative... Have you read Kurzweil's The Singularity is Near?

We tend to think linearly in a world that is changing at an exponential (and increasing) rate.

Sounds great. Put a sensor in the stop sign and the car will break for me. GPS will check and tripple check to make sure that there's no problems.

Here's my question. When the system breaks down, and it will sooner or later, who is responsible for the consequences? When it turns out that the people involved in the accident were poor, how do you keep the jurry from awarding 2 billion dollars in damages?

"fishbane" hit on the reason that I don't see this happening (at least not here in the US): liability.

Any company that produced such a car would be sued over each and every accident that occurred.

Too bad.

This is the same reason why you'll never see Jetsons-style flying cars, at least not in cities.

In fact, none of the Jetsons stuff is coming true lately. Never mind the flying cars...where's my robot maid, my foodarackacycle, the pneumatic tube shooting me around?

Just as well. I can do without Jet Screamer singing "Eep Opp Ork Ah-Ah!"

Joe - When it turns out that the people involved in the accident were poor, how do you keep the jurry from awarding 2 billion dollars in damages?

Shrinkwrap EULAs? Works for everything else software related. ;)

Plus, if an IED blows the hell out of an automated truck, no one has to write a letter to its mom.

Dear Mrs. Eniac,

110101100010101
010100101010100

Mark When was the last time your desktop computer crashed?

Keep in mind that a desktop computer can be made more unstable from the various conflicts that result. Netscape Navigator used to interfere with Norton Antivirus for instance. Certain drivers created conflicts. As mentioned previously, designing reliable software is expensive. But when you have control over all the software and hardware in a system you can dramatically reduce the number of variables that a programmer needs to deal with.

Besides, if data protection were that big an issue, computers would come with three-hard disk RAID arrays and external nightly backup services to an offsite location over broadband. And they'd be more expensive. But most consumers purchase cheap, lots of bells and whistles and buggy over expensive, simple and reliable.

Presuming that I can still read at eighty

When you're eighty, books will read themselves too.

Shamus: “Reliability won't be an issue because the system will be engineered by people who konw what they're doing rather than dipshit losers at Microsoft.”

Designing an embedded system intended to perform a few pre-determined functions is a much different problem that designing a general purpose system that allows unlimited user installed applications. Embedded systems allow testing methodologies that, while expensive, guarantee high reliability.

The competence of the designers really isn’t the problem.

Jane: I don't believe that you don't understand how your toaster works. Even its insides.

Yeah, really -- everyone knows about the toaster elves, who live inside your toaster and operate miniature flamethrowers. When the bread matches the color setting, they kick the release lever.

If you try to work 'em too hard, they'll kick the release lever as soon as you try to put in another slice, and make you wait until they've cooled off. Ingrates.

Hey, no one has stopped to consider a very important aspect of this issue: I happen to like driving. It's an end in itself. Very relaxing. A good time to organize your thoughts.

Personally, I think the Teamsters will mount more of a challenge than the technological problems.

Most of the DARPA challengers' navigation systems were fairly primitive*. The winning Stanford team was the only one to deploy a system that relied on cameras to any significant degree -- the four other teams that finished the course did so using only GPS and laser range-finders (which work like sonar).

Everybody but Stanford stayed away from true machine vision-type solutions because they couldn't pack enough CPUs into a car to do the image processing in real time.

But in five or ten more years...

* Not to say badly designed, but very simple compared to what we think of when we hear the word 'vision'.

Thanks for that image anony-mouse. Making toast will never be the same.

Toasters are wonderous devices. You put in bread, push down the lever, and toast comes out. No one knows where the bread goes.

In the 1950s many people would have made the same prediction, but they were wrong (actually, a lot of people had grander visions than that -- they we'd all have flying airplane-cars in our garages but they were *really* wrong).

Our oldest interstates are about 50 years old now and they look about the same as they did then. I predict they'll look and work about the same 50 years from now. The engine and fuels may be very different, but I think cars 50 years from now won't be all that different otherwise either.

Cars that drive themselves with no driver? I'd be against it. I'd bet there will be more and more automated systems to *help* drivers, but that's a different issue.

I've heard it's already possible to have jetliners fly themselves from takeoff to landing.

I'm with Slocum. None of the comments so far touch upon the main reason I'd never want to "drive" something that drove me -- and incidentally, the same reason public transport is such a typical waste: *we LOVE the FREEDOM of driving our OWN car*.

A few obsessives actually love their cars, but most of us just love being able to crawl in and go where we want, under *our control*. Being driven -- either by a car, an automated train, or a driver on a bus -- takes this away from us. I've read a myriad of articles about robot cars and they all assume drivers will just give up this freedom willingly, presumably for some mythical safety and convenience.

Nuts. Robot cars are nothing but personalized public transport -- who would be in charge of where you're going and how fast, after all? I'll stick with what I've got and let the futurists go back to the drawing board.

My biggest problems with public transport are;

1. It doesn't go exactly where I want to go.
2. I don't have my car when I get there, which makes #1 worse.
3. It's more time consuming.
4. It's more confusing to use.

Outside of that, cars are a huge waste as far as I'm concerned. When I lived overseas, taxis were cheaper and that was great. If machines take over the task stateside making public taxis a cost effective alternative to car ownership, I couldn't be happier.

What would be even better would be trains that I could park my car on so I'd only have to drive
the last few miles.

I've read a myriad of articles about robot cars and they all assume drivers will just give up this freedom willingly, presumably for some mythical safety and convenience.

Nuts.

Driving time is entirely wasted time, IMO. If I want to commute from MA to NJ to visit relatives, I've got to resign myself to 3-4 hours of intensely boring behavior - sitting still, staring out a window, moving my hands a bit.

I've got about 60 outstanding books sitting at home that I want to read - everything from economics textbooks to great science fiction to medeival history...and instead, I have to waste 4 hours of my life (8, if I want to come back home afterwards) staring at asphalt and other vehicles.

I'd pay a large amount of money to delegate that idiotic task to a machine, so that I could read while being whisked to my destination.

But, your mileage may vary.

Some people like shopping for groceries, and scrubbing their own toilets.

...wherease for me, those are also on the "delegate to someone/thing else" list.

Totally agreed with TJIC. Back when I could take the bus to work, I did. It took twice as long, but I'd rather spend half an hour reading than fifteen minutes bored.

I've read a myriad of articles about robot cars and they all assume drivers will just give up this freedom willingly, presumably for some mythical safety and convenience.

Why would there be any surrender of "freedom"? I would imagine one would simply punch in the addresss of one's destination, and the robocar would do its thing. I hardly imagine one would be required to let the robocar choose the destination. Moreover, as other commenters have mentioned, one would experience an increase in freedom, such as the freedom to, say, watch a movie on the long, boring drive from Boston to Bangor, or catnap during one's morning commute. I seriously hope this stuff gets perfected before I'm in a pine box.

I'd rather spend half an hour reading than fifteen minutes bored.

Like everything else made in China, the price of basic car audio components (a CD player that can also read MP3 discs, for example) has fallen practically through the floor. Assuming you can install it yourself or bribe a friend to do the work to avoid the mandatory accessories rip-off during the store's install procedure, you can occupy any length of time you please for under $150. Start downloading podcasts and burn them to a CD-RW, or check out a few books-on-CD from your local library.

Moreover, as other commenters have mentioned, one would experience an increase in freedom,

Not if you enjoy the actual driving experience. Maybe your utility curve don't schwing that way, but some of us do enjoy it. And some of us live in flyover country, where at non-peak times of the day, we can actually drive someplace, rather than engaging in two hours of aggravated parking practice.

The navigation technology is already there, and the main thing preventing it from being tied into control systems to for automatic driving in ideal conditions is the law.

But in this case, the law isn't the only obstacle to building cars without steering wheels and manual controls. Under anything less than ideal conditions, the assumptions necessary for the car to drive itself break down. Take snow and ice, for example. No problem for an experienced handler of rust-belt roads, but while it'd be possible to program a computer to cope with them, there's no known way for a computer to reliably _detect_ them and _measure_ them, so as to engage that programming.

Navigation is easy, compared to keeping a car safely on the road in adverse conditions.

"Not if you enjoy the actual driving experience"

There will be 'driving parks' for such people and perhaps 'manual only' corridors for them in some areas of the country.

I actually like public transportation (where practical) and am thankful it doesn't take me door to door so I get at least a little exercise walking.
Automated cars sound like the next best thing to a mid-length train trip with a good book.

One thing worth mentioning is that in addition to navigation systems being up-to-the-task, there will soon be cheap and reliable systems for detecting congestion and mean traffic speeds. They use a very clever hack of timing cell tower handoffs as drivers with cell phones move from one cell to another. Given that information from the cell towers, it's easy to infer average speeds and car densities. Feeding that information back into your car's nav system will probably be standard in 5-7 years.

"A (very) modest step toward what you desribe already exists in some expensive cars such as high-end Mercedes and BMW's. They have a so-called "active" cruise control which varies the vehicles' speeds using radar, in order to maintain proper spacing from vehicles in front."

These systems were originally called 'collision avoidance' until the lawyers got involved and it became 'active cruise control'. They use Doppler radar to determine closing speed and can apply the brakes as needed. One aspect of hydraulic brakes is there is a slight delay between when you step on the brake and when they actually are applied. The radar can detect a potential collision and 'prep' the hydraulics to eliminate that delay. Also if it determined the collision is unavoidable it can also prep the airbags which would theoretically allow them to be less 'explosive'. I think these systems run 2-3K on high end Mercedes.

The red vs. blue conflict between driving enthusiasts and train ride afficionados is surely unfounded. As with all other improvements in automotive technology, the autopilots will first appear in top of the line luxury sedans (and maybe SUVs if those don't Dodo away first) and slowly propagate downmarket. There will be a very, very long period between their first appearance and the day when the last non-equipped vehicle rusts itself into dust.

Note that the relevant laws of many states (Texas for sure) still don't require catalythic converters, seatbelts or even blinkers for sufficiently old vehicles. And some of them are still on the road. Thus, the automatic driving systems could only possibly be useful if they can coexist with manually driven cars on the same stretch of concrete. And surely the more time passes, the better AIs will be written, and by the time it would be feasible to ban manual driving there will be no utility in such a ban. Except, perhaps, to stick it to us red-staters. But if by that time the latter is still a consideration, loss of driving privileges would be the least of our problems :(

Automatic cars within 50 years were allready predicted 50 years ago. Making cars drive themselves is not that easy, especially since an automatic car will be required to outperform a manually driven one at least two orders of magnitude safetywise.

I would suggest you also look at the way railways do it. The problem of driving a train automatically is trivial compared to a car which is not bound to a guideway, and still it is not universally done. The requirements for availability and reliability make this a very expensive exercise.

And anyway, the problem of how to be able to travel medium to long distance while being able to read has been solved a long time ago. It is called "let someone else drive". Division of labour, you know. That what Capitalisme excells in. When I visited my parents over the weekend most of the driving I needed done was done by employees of "Deutsche Bahn" while I watched the rest of season two of Battlestar Galactica

As with all other improvements in automotive technology, the autopilots will first appear in top of the line luxury sedans (and maybe SUVs if those don't Dodo away first) and slowly propagate downmarket.

But it is very much harder to have a 'smart' highway with auto-piloted cars if only a few of them are on autopilot. Right now, for example, one of the things you have to do as a driver is to 'mind read' other drivers -- to notice if another driver is inattentive, erratic, or aggressive. Do we envision automated systems capable of that? I don't.

And what of the legal ramifications of determining liability if there is a collision between an auto-piloted car and a manually-driven car? How'd you like to be the manufacturer of an auto-pilot system involved in a court case like *that*?

Krist--my understanding is that the reason trains aren't automatically piloted has almost nothing to do with the technical hurdles and everything to do with the near universal fact of strong railway unions . . .

99: "I've heard it's already possible to have jetliners fly themselves from takeoff to landing."

While that's true, I don't think passengers will fly in a pilotless aircraft. So pilots will be needed on board to take over when systems inevitably fail. Problem is, those backup pilots need extensive and recent experience landing large aircraft. It just makes sense that airlines continue allowing pilots to land their planes.

IMO, automation will continue enhancing pilots' abilities to fly their aircraft. It will not replace them - at least not in 50 years.

I also believe that's the direction for automobile automation as well. Drivers will see their abilities enhanced, but will not be replaced.

The other key point is that could reduce congestion; because they have superhuman reaction times and can coordinate with each other wirelessly, computers could pack the roads much more tightly than human drivers could at speed.

(If you *really* trust the system, you could get huge wins in traffic flow by doing away with traffic lights at intersections where only computer-operated vehicles are allowed...!)

In fact, I expect the HOV lanes will become "computer-operated vehicle only lanes", where everybody's tailgating 5 feet apart at 75 mph.

The tech is almost here (well, 10+ years for practical commercialization); liability will be the problem.

Jane,

In Europe one of the countries that is the most advanced when it comes to running trains automatically is France. Hardly a country where unions are powerless, on the contrary. But you do have a point.

The main issue is however reliability. Reliability requirements are high in railway applications, and they will be very high for automotive applications too.

When I drive I can make quite a few choices that affect the chance of me getting hurt or dying in an accident. I can choose to wear a seatbelt, be attentive, be sober and well rest. I can choose not to drive at inappropriate speeds and maintain my car properly.

When I give the responability for my safety over to someone or something else I want performance that exceeds the best I can achieve myself by a significant margin. Before I give up control I want to be doubly sure.

It is for this same reason that the travelling public in general demands safety levels in airline and rail travel that are a few orders of magnitude higher than what is accepted for cars. And rightly so.

Seeing what the challenges (and costs) are that are involved in achieving this level of safety in airline and rail travel (where the problem is acutally a lot less complex) I have my doubts it will be possible for car travel.

Automatic cars have been "a few decades away" for almost half a century by now. Don't hold your breath...

Driving time is entirely wasted time, IMO. If I want to commute from MA to NJ to visit relatives, I've got to resign myself to 3-4 hours of intensely boring behavior - sitting still, staring out a window, moving my hands a bit.

Well, actually no, you don't. Mass. to Jersey? Piece of cake: you could take almost any other form of transportation that lets you sit and read your pile of books: train, bus, air, ... Oh, wait. You want to go on your schedule? That mindless four hours starts and ends when you want? Sorry, that's the trade-off. And, sheesh, get a CD player ...

Why would there be any surrender of "freedom"? I would imagine one would simply punch in the addresss of one's destination, and the robocar would do its thing.

It's who's deciding that "thing" that concerns me. An algorithm? A company? A (gulp) government? And catnapping while the car drives itself?!? Maybe I'm just not enough of a visionary, but would that include this speeding hulk of mechanical genius avoiding things not on Yahoo maps, e.g., (blue state) potholes and (red state) farm animals? You first, my friend ...

Actually, I empathize with the down-time arguments. I occasionally get driven for my work and find it a luxurious experience -- and I get a lot done. And I really have nothing against this idea in the abstract; it just grinds against what I see as a fundamental aspect of transporting yourself where and when you want. Every scenario I run over in my mind comes back to the freedom of driving. And I'm not at all convinced that this whole business isn't just "individualized" public transport -- which is a whole other argument.

Horses are a form of self-driving transportation, as are camels and donkeys. Buggy-pulling horses on regular routes were known to stop at established destinations even if the driver was taking a nap.

The solution may be in genetic engineering -- develop a super horse capable of sustained highway speeds. No need to turn all that grass and corn into ethanol.

Let's be realistic here. Any movement towards automated traffic will be incrementally implemented. I foresee automated expressways as being the first implementation, possibly as suggested above with automated HOV lanes. Regular primary highways could be next. Secondary roads will take a long time to implement, as will driveways and local roads. I'm not sure that automating local roads will ever be practical.

As for computer glitches, control systems aren't nearly as complex as general purpose computers. And because they are much simpler devices, adequate redundancy can be built in.

I would certainly love to be able to drive to an interstate highway, pause on the access ramp, push a button, and then not have to do anything (steer, brake, etc.) until I was safely deposited near the end of the off ramp near my destination.

Just imagine: no more 50 car pileups in the fog in Central Valley, California.

"develop a super horse capable of sustained highway speeds"

One fond memory I have of a trip to India came when I saw elephant 'drivers' heading home for the day. After a long day of giving elephant rides to tourists, all the drivers had to do at the end of the day was to climb up on the platform and take a nap. After all, the elephant knew his way home!

They weren't fast and were a significant impediment to the automobile traffic flow, but the drivers sure looked comfortable lying up there on their backs.

>> Any movement towards automated traffic will be incrementally implemented. I foresee automated expressways as being the first implementation, possibly as suggested above with automated HOV lanes.

Upgrades to road infrastructure are much more expensive than in-car improvements. So, I imagine we'll have a host of those incremental automation features on automobiles -- and not necessarily just executive rides -- long before NYC or Boston turns the first of its Lexus lanes into a zombie Lexus lane :-)

Let's say that in addition to that radar-based cruise control we get a gizmo that can follow the lane markers. Pair that to GPS navigation to predict when maneuvers may be required and give driver a heads up. Perhaps a wider angle and/or longer range radar to catch traffic clumps and those errant cows and again, warn the driver and perhaps drop the speed?

Things like that will not an autopilot make and will not absolve the driver from responsibility of, well, driving -- but may they not make highway driving easier and more comfortable and perhaps provide an additional safety margin? All that for maybe extra $5K or so, financed over 5 years -- instead of an extra billion in taxes...

But, overall, high-bandwidth communications are cheaper than ubiquitous public transportation. So let's perfect those communication bracelets first, then worry about robot cars.

Funny, I was just positing this very thing to my husband on the loooooong trip we took over Labor Day from PA to Maine. He scoffed... but I have hope, for these reasons:

We used to live in Seattle, where I was fortunate to have a ferry commute for a few months. What an absolute joy that was, to drive aboard, then either shut my eyes in the driver's seat or wander upstairs and have a cup of coffee while some of the most beautiful scenery in the world flowed past, or catch up on work, or read for pleasure. Aaahhhh... So, even though I do enjoy driving, I appreciate not having to all the time.

Then, also in Seattle, there are the "express lanes" on I-5 through downtown, which switch directions every day to improve traffic flow at rush hours. (Yes, I know these aren't unique.) Physical barriers prevent driving into the lanes in the wrong direction at the wrong time, so only cars behaving "correctly" can enter, and unlike other systems I've seen which separate express lanes from regular lanes with "turtles" that you can still drive over, the Seattle express lanes are walled off from the regular lanes, further reducing variability in driving. Within the express lanes, speed of traffic and distance between cars are remarkably constant, even using human-only judgment.

And then, I've come to rely heavily on the various online map services, because they almost never seem to make a serious error (if I have a problem getting to where I'm going using Yahoo!Maps, it seems to be because I either didn't believe what it told me or couldn't find a street sign, not because the directions were actually wrong), which indicates to me that there's a pretty darn good database of places and roads out there.

And finally, we crossed the GW Bridge after midnight on that trip to Maine and construction had the thing down to one lane; it took us half an hour to get across, during which half hour I had to listen to my husband bitch about how easy it would have been for "them" to alert drivers to take alternate bridges. True enough. But think how much more efficient it would've been if there'd been no need to alert drivers at all - just keep them moving at high speeds and low between-car distances, and reroute just the right number of drivers to optimize flow.

An idea whose time has come! Especially when it's one in the morning and traffic is at a standstill!

Assuming that the self-driving car plugged into a mass navigation network eventually replaces driving and allows cars to move quickly and safely in heavy traffic, it will be interesting to consider how the car itself will be changed.

What will it mean when car design need not include practical concessions to operating the vehicle? Might a hemispherical shape be better for safety and aerodynamics? Would people buy one-seat commuting vehicles that were more like large enclosed motorcycles? (Think of how many could fit on a road). Any other ideas?

Agree with Jamie .

Comments are Closed.