What do the laws against driverless cars look like?

A few people have asked me this question, here is one example, from Falls Church City:

No person shall operate a motor vehicle upon the streets of the city without giving full time and attention to the operation of the vehicle.

Of course that wasn’t intended as a law against driverless cars per se, but that would be its practical effect.  Ask yourself the following question: let’s say you sat in the back seat, singing rap songs with your shirt off, while the computer piloted the car flawlessly.  In a 35 mile per hour zone, the car would go exactly 35 mph and you would smile and wave — with both hands — at each police officer you passed.  For effect, you could stick your two feet out the window as well.  How long could you go before they pulled you over?  How far could you get?  When would you get your car back, with computer of course?  How would they respond if you asked: “Officer, please show me where in the books this is illegal?”  In which state would that question go over best?  Worst?

Liability and public opinion issues loom larger still.  The driverless car, if it proves feasible, is most likely to come first to a smaller, higher-trust nation such as Denmark.  In some countries, if the government announces “X is safe” people believe “X is safe.”  The United States is not one of those countries.


The defense to an alleged infringement of the Falls Church City ordinance (and probably many other relevant laws) would be: "I wasn't operating the vehicle. The computer was operating it. I was just a passenger."

Hmmm. If you had a killer robot and were charged with murder, would this also be your defense?

I think the legal question would be whether the owner or the manufacturer of the car [killer robot] was responsible.

I thought that was the premise of iRobot with Will Smith

Yep, and it was considered an equipment malfunction, not murder.

Joe Jones is right. At least as to this statute. I think the more likely issue would be with vague reckless driving statutes.

The analogy to murder is inapt. Murder statutes are not limited to the "operator" of the implement of death like this statute is. Simplified a bit, murder simply requires that you start a chain of events that kills someone, and that you intended their death.

Incidentally, the "full time and attention" statute is commonly used as a lesser offense people plead to when charged with reckless driving (which is a misdemeanor in VA for speeding 20mph over, among other possible ways to qualify.) It's very rare than a police officer issues an initial citation for the offense.

Your question seems apropos if there were a lethal car accident.

But the law is merely against operating without attention, even if no accident or damage occurs.

(The other obvious response to Joe Jones is that if the computer is driving, the computer needs a driver's license.)

Fair enough.

Presumably this will get sorted out. Long term, who is liable when the thing kills someone is a good question.

Maybe the automaker would sell assumption-of-liability? Maybe they'd be required to assume it by regulation? The insurance industry would probably lobby against liability moving away from individual drivers...

Maybe you should change:

"In some countries, if the government announces “X is safe” people believe “X is safe.” The United States is not one of those countries."


In some countries, if the government announces "X is safe" then “X is safe.” The United States is not one of those countries.

In some countries, if the government announces “X is safe” then “X is safe.”

On which planet are those countries located?

Like in high trust Germany, where the government announced that vegetables are safe?

A comment like this almost invariably comes from somebody who has never spent any significant time in another country, forming impressions instead on the basis of naive accounts written by journalists who've spent a week in said country, usually via funds provided by that country's government. It's a fundamental, invariable rule: Lying is what governments do. Some political systems, like ours, are designed to take that into account; many are not.

But what if a computer drives the car, and the person on the car devotes full attention to the computer software running the car? Wouldn't that be legal? After all, the law doesn't specify how I have to devote attention.

Out of topic: you say: "The United States is". As a european non-mother tongue english speaker, I would say "United States are", just like I would use the plural for translating the sentence in my own language (italian).
So my question to Tyler and anyone else around here is: is it occasional, or is it common in US? Does this imply you consider US as a country per se, rather than a federation of States (hence why I think of US as a plural)?

It depends on what your definition of is, is.

In American English usage, "The United States" takes a singular verb: So, yes, "The United States is..." is correct. As to whether this means that we think of the country as a single entity vs. a federation of States...

Interestingly we still use "are" for the West Indies.

A federation of states is treated differently than a group of islands?

The West Indies is not a designated entity , except for Cricket.

For me it would depend on the context and whether I'm referring to the country of the US (singular) or the states that make up the US (plural) ... eg

"The United States is a country made up of ..."
"The United States have a high degree of independence in setting regulations for..."

The BBC seem to agree:

This is a well-known difference between American and British English. Collective nouns take the singular in the US and the plural in Britain (and many British ex-colonies).

There is a simpler answer. When "The United States" is used as the name of a country, it is not a collective noun, it is simply and solely the name of the country.

Prior to the US Civil War, the common usage was "The United States are...."

Having settled the late great unpleasantness between the states, common usage now is "The United States is..."

This is pretty close to how Google was test driving their cars; they had someone in the driver's seat paying attention and ready to assume control.

Maybe this is the intermediate step we need. I'd be willing to sit there and uselessly "drive" the car while the computer drives it if it leads to public acceptance.

you say: “The United States is”. As a european non-mother tongue english speaker, I would say “United States are”,

If you believe the aphorism, so did we until the Civil War.

The change occurred at the time of the American Civil War. Up until that time, official and unofficial usage was "the United States are." After that time, usage changed to "the United States is." The change reflected the political reality that the war ended of any real sovereignty of the individual states, which became mere subdivisions.

Put your shirt back on!

What about Chimpanzees? Up from the trunk? We are still looking for things to do with all the apes we are going to need for longevity research.

(we don't want them taking away manufacturing jobs)

Aren't we getting ahead of ourselves here?

Do we have studies that show driverless cars are as safe as non-driverless ones, and would you want such studies before you changed the law? You might want to have a discussion first on what kind of studies you would like to see, or the circumstances where using a driverless car was inappropriate or quite appropriate, rather than looking at the law in place without questioning whether driverless cars are as safe and under what circumstances.

Afterall, this is a democracy, and as much as some like to criticise all laws, we do have a say in creating them.

And, even though we have automobiles, sometimes you don't put the cart before the horse.

The "safety" of cars operated by ugly bags of mostly water isn't one constant number. It varies a lot based on factors relating to the current state of the meatbag and the world around it.

What test would you recommend? How about if the Google engineers let their kids play in the street while the car drives around them?

I think this is part of the point Tyler's making with these self-driving car posts. The more laws, regulations, etc. surround an area of life, the harder it is to introduce big innovations. At one extreme of this, you have the early days of the internet, when there wasn't much existing infrastructure or law or entrenched interests who could squawk if they didn't like what was happening. And so, we got huge rapid progress, wild innovation in millions of (mostly not all that useful, in retrospect) directions, etc. At the other extreme of this, you have innovation in medicine, which involves huge costs and regulatory barriers and laws--we do see innovation over time, but the innovation doesn't move quickly.

In an area where there's already an existing set of laws, customs, expectations, interested parties, etc., it's going to be harder to innovate. In areas where those things don't exist, innovation can happen a lot faster. (Note that *bad* innovation can happen faster, too, as with medicines that turn out to have horrible long-term effects.)

I see the point you are trying to make but comparing the innovation of drugs and self-driving to the innovation of the Internet to conclude regulation impedes innovation is a lot like concluding cigar smoking increases life expectancy. There is a third factor which tends to increase the amount of regulation and even on its own discourages innovation, which is the underlying riskiness of the activity. I'm much more comfortable trying out a new photo-sharing app than I am taking experimental drugs.

Ryan: Yeah, that's a good point.

Also, there's a problem with choosing any industries or areas where we immediately think "massive innovation" or "decades of stagnation," because then it's hard to interpret any similar areas for comparison. That is, the first examples that come to mind are comparisons between the internet and cellphones or digital TV, but since I started out thinking about how amazingly innovative the internet is, it's hard to say whether this means anything.

I'm not even sure, now that I think of it, how I'd try to check whehter this idea is right.

Dan, Think about how an insurance company would write a policy for a driverless car. When they are close to equivalent to a car driven by a human, that might be one test. You need the law of large numbers, so experimental vehicles might be the way to go for a while.

I'd be fine with letting Google self-insure while they test it. They should pay like any other insurer facing a payout, without any stuff like "reckless disregard because they dared to have a computer driving it."

Google, with their self driving car has decided a person must always been in the drivers seat when on public roads, and a human may override the system by pressing a button, the brakes, or providing steering input. I imagine a human in the driver's seat will be required for some time. They will likely initially allow talking on cells phones etc provided your eyes can monitor car travel. Gradually as the system proves its competence, you'll see the ability to read. Eventually they could be used as on-demand taxis. People with periodic vehicle needs could simply search a travel itinerary in google maps and they could show you how long the trip would take on public transit or by taxi vehicle and press a button to summon the nearest vehicle to pick you up and take you to where you need to go. It could be an amazing development as most cars sit idle the vast majority of the time; google's routing software could easily find people in the general vicinity going to the same general destination and allow carpooling and their payments system could debit each traveler accordingly. A system of this sort could be far less expensive than owning a private vehicle you have to drive/maintain yourself.

We have had aircraft that can essentially fly themselves for the last 20 years but still they all have a human in the seat watching over them.

I wonder why people think cars can skip the "monitored automation" stage?

The benefit of monitored automation is lower when the pilot is not actively engaged in monitoring. In some circumstances it is considered necessary to have the pilot actively in the control loop, because there isn't enough time between something going wrong with the automatic system and ball-of-flame for a pilot to get the necessary awareness and insert himself into the control loop and fix things, even if he is paying attention.

If car drivers are actively engaged in monitoring the car's systems, that takes away a lot of the benefit of driverless cars.

If the car can't automatically deal with a blowout swerving you towards the baby carriage about to enter the crosswalk, it's useless. Expecting the driver to look up from his text messaging, grab the wheel, and intervene is unrealistic.

Yes, but that doesn't change the lesson we have learnt from airplane systems: The best of control loops will fail in unexpected ways and there will be modes of failure we had no clue about. I see no reason why car automation technology will be any less fault prone.

Agreed that driver monitoring is unrealistic for cars; but then how will we handle catastrophic failures; especially in the severely risk averse society that we live in.

I don't think we have much disagreement. My point is that you aren't going to get good monitoring like you get with pilots. Therefore you need the car to be able to handle catastrophic failures, otherwise it's not a viable technology.

The literature on aircraft safety (and the track record) comes down strongly on the side of pilots + control systems designed to be redundant and debuggable being superior to fully automated systems or pilots with non-redundant systems. However, those pilots need to be given tools and procedures to be useful. They spend time monitoring the engines - they have sensors and displays to tell them the condition of the engines - they have, say, a backup fuel pump they can switch to if the primary pump looks wonky in their educated engineering judgment. This isn't what people envision when they talk about driverless cars. They're talking about fully automatic cars that can deal with the bad cases all by themselves.

I do think a) that it would be easy to switch off the autodrive system when entering a situation it obviously can't handle, like driving through a pedestrian crowd, b) that it's probably not possible to have a non-in-the-loop driver intervene in a helpful way in an accident, and c) that before driverless cars we'll get more great incremental improvements to technology like traction control and cruise control, that in some sense drive the car but with human-in-loop control. Pure driverless cars require a real step forward in technology. I hope it happens.

A bumper-to-bumper racing league would be a nice proof-of-concept for this technology.

A human driver probably cannot handle a catastrophic failure any better than the computer. Already the computer does a much better job using ABS to stop the car. In ideal conditions a human can realize what's going on in 160 milliseconds (that's college students prompted with an audio cue). At 70mph, that's half a car length, just to know you're in trouble.

A plane usually doesn't crash into something in 1 second. And "just stop" is never an available failure mode for a plane, which it is for a car. (It may not be the best move, especially depending on what's going on behind you. But the computer will also know that a hell of a lot better than you.)

> A human driver probably cannot handle a catastrophic failure any better than the computer.

I'm not sure I agree. A lot of what goes on in a crash is not just object avoidance, but also rapid judgment about what is and is not okay to hit, or what action will bleed energy fastest. A couple of seconds is a lot of time in an emergency.

Planes are usually manually landed, because it takes too long for the pilot to grab the yoke, recover his situational awareness, and fix the problem should something bad occur. They are not generally manually flown mid-flight because there's enough time to respond to problems. Mid-flight pilots are systems engineers. Driving is closer to landing a plane than to cruising at 30,000 feet.

I should have said airliners, rather than planes.

And some demanding "death is seconds away" flying tasks are, I believe, automatically flown, albeit not in airliners. I think nap-of-the-earth flying, as one might do in an attack aircraft, is generally not done by hand, particularly in night or bad weather.

A human might know that, but any better than a computer? The computer can have an order of things that are preferable to hit, like 1) small trees or flimsy street signs, 2) other cars, 3) brick walls, and then way down the line, 9) anything that might be a human. And in that other cars category, it further breaks it down by what those cars could hit, with the same logic, recursing as necessary

I can certainly imagine some situations where the meatbag would be better than the computer, like seeing that a car is full of 9 orphans and you better not touch it at all.

(I couldn't find the video I was remembering, but this one shows how much it's aware of, and it does the stunt when the pedestrian steps in front of it. http://www.youtube.com/watch?v=X0I5DHOETFE )

This is the sort of thing people do better than computers.

That said, I'm not a super-skeptic on this. And as long as regulation doesn't F- it up, it strikes me as quite amenable to incremental testing. There's no one "This better work!" moment inherent in the technology.

I think a racing series would go a long way towards convincing me. Plus you could do things you wouldn't do with human drivers, like have the cars drive upside down. :)

I would like to completely disagree with Finch. People are excellent at pattern matching, but unless people go through many crashes, they will be significantly worse than a properly design computer. I would even surmise that a computer with more input (wider frame of reference than human eyes), quicker processing (computers are orders of magnitude faster than the human brain), and proper training (yes, computers can be designed to learn almost exactly as humans do) will be much more proficient at taking the best course of action during a wreck than a human, even one with full attention and ability on the task at hand. We may be a few years away from such a spectacular design, but without much improvement in human brain power it is only a question of when it will happen.

I feel the devil's in the details. Computers are indeed good at pattern recognition but with lots of caveats. The beauty of a human is in dealing with unprogrammed situations in a decent manner.

I had several questions: Do these cars read street signs (yield, stop, speed limits) or do they depend on an internal database of speed limits, stop-signs etc.?
Would they recognize that the vehicle ahead was a school bus or an EMT and yield way? Can they identify a legal parking spot and then park.

How would they deal with toll stations and police check points (I'd be tempted to use them as a drug mule)

I feel driverless cars do have a future; but people are getting too optimistic based on one limited experiment by google.

The beauty of a human is in dealing with unprogrammed situations in a decent manner.

Definitely. If you don't specifically do work on it, the computer would have lots of tragic situations when faced with a crash. It would definitely be able to hit the breaks a lot better, but it might, say, refuse to swerve off the road.

Do these cars read street signs (yield, stop, speed limits) or do they depend on an internal database of speed limits, stop-signs etc.?

If you look at the video I posted a few above, it does have an internal database. But it does use video recognition for things like traffic lights. The reporter asks how it handles stop signs but answers it with an unrelated question, which seems really weird.

(The cool things at around 1:19 in the video is that you see its internal representation, and every car is assumed to have a block of matter in front of them that it should just consider a solid.)

How would they deal with toll stations and police check points (I’d be tempted to use them as a drug mule)

That's a really cool idea. For now you can't get it more than one tank of gas away right now and you risk losing the car.

(This TED talk mentions tool booths, but they still had a person in the car. http://www.youtube.com/watch?v=bp9KBrH8H04&feature=related )

Two reasons: first, because the average human driver is much, much worse at driving than the average professional pilot is at flying. So even if the automated driving system is absolutely terrible at handling unexpected modes of failure, it will more than make up for that in safety by not running red lights, not taking sharp turns at 70+ MPH and hitting trees, and not losing consciousness and drifting across the median. Second, just slamming the brakes is a reasonably effective strategy in a car for unexpected catastrophic failures (not the best obviously, but close enough to the average actual human reaction). There is no equivalently simple strategy with an airplane, for rather obvious reasons.

Please revisit your Bugs Bunny cartoons.

"I wonder why people think cars can skip the “monitored automation” stage?"

We don't as good of an union?

Nitpick: We have had self-flying and remote-control aircraft for quite awhile, but they're usually being used to blow things up or put things into orbit.

And they have abysmal safety and reliability records.

The US will be first in driverless cars. Next year will be important if oil prices bring down the economy per our outlook.

A whole portfolio of laws and standards are required, including but not limited to:

- required performance levels (same or better than driver, how much better)
- method of failure (does it crash or just pull over?); required redundancy and reliability standards (these will be major cost drivers)
- interface with police
- applicable speed limits (can you make it go 80 miles an hour?)
- parking and pickup regulations (maybe your vehicle would just rather join a traffic jam rather than park in Manhattan)
- child regulations
- liability management

There will be more as well.

Good points.

People think of a single law authorizing driverless cars, but don't think of law and technology as a system, particularly as an interactive system that creates, over time, a system of laws to deal with problems as they arise and as technology changes.

What are the rules for releasing nano-particles in the environment? Do you look at one set of environmental rules, or create new rules?

Looking at the current luddite anti-atom hysteria in Germany (of all countries), I am not sure about the psychological prospects of self-driving cars.

Any crash of such car will be reported as a sensation. "OMG this technological godzilla evil robot shredded two children! Ban it NOW!"

Same as guns.

Maybe the American population is stiff enough to resist, but I do not think so. Given the ludicrous way that children are treated (can't go anywhere out, as everyone seems to believe that the world is full of evil pedophile rapists just waiting to snatch them), I think it takes one or two crashes where child fatalities are involved to ban the entire industry.

You will logically say that human drivers kill many more children on a yearly basis, but the psychological impact is not the same.

However, there is a powerful incentive pulling the other way: the convenience of not having to drive a car will be extremely attractive to people. This may outweigh concerns about safety.

The psychological impact of accidents caused by people talking on cellphones is similar. Further, this behavior, unlike theoretical driverless cars, actually makes driving more dangerous. Despite this, talking on the phone while driving has not been banned outright even after many fatal crashes. People do not want to give up convenience, even when it comes at a real cost.

...in Germany (of all countries)...

I'm curious what that is supposed to mean. Are you thinking of the Germans as particularly scientifically aware? After all, they are the people who gave us both Koch and Homeopathy. There's a strong current of wackiness (without resorting to the argumentum ad Hitlerum) in Germany.

The Germans I know are rather pragmatical and more-inclined-than-not to believe in technology and science.

Granted, my sample is not representative, but from the Euro nations that I know anecdotally, Germany seemed to me the most "rational" one.

And yes, they have strong tradition of science and technology, and they tend to teach the youth in the high schools some basics of both, in extent greater than typical for Anglo-Saxon countries. For example, young Germans who study the type of high school tracked towards university (Gymnasium) usually have in-school laboratory practice in chemistry, biology and physics.

But you're right that the stream of pseudoscience emanating from Germany is not quite thin either.

I still think you are right though; overall the Germans are a more educated, techno-friendly and rational population than many other nations.

Remember that when the Germans came up with homeopathy the rest of the world was still burning witches at the stake.

Yes, you can find anomalies but it doesn't change the fact that the Germans were historically ahead of the curve in scientific areas.

Historically speaking "Germans" are a very young ethnic group.

Uhm, one more thought ...

Fines from bad driving (or allegedly bad driving) are a substantial income for townhalls, and also a serious source of power for the police (sobriety checkpoints etc.). Here you have two powerful constituencies with incentives to push laws against driverless cars.

You are assuming driverless cars will be safer; I am not sure if this is a given. Technology can fail in unintended ways; but we will only know after we try. I am especially curious of the game theoretic aspect of maneuverer such as merging in heavy traffic or when two cars come to a stop sign at the same time.

Besides Toyota is a wealthier prey to sue than a drunk driver.

Precisely obeying the laws will be exactly the best spot for a driverless car. It will never weave out of that lane.

I am not concerned about others who want to purchase computer controlled cars. What concerns me is that making this technology common will lead to regulations requiring at least some elements of automatic override and external control, particularly so long as we live under the Ray 'Airbag-Coccoon' LaHood regeime. After all, if we prove that automatic control is safer, doesn't it just make sense to take the fallible human out of the loop? And how many lives could be saved if we eliminate high speed chases by allowing the police to remotely stop fleeing motorists or traffic lights to shut down engines until they turn green? Call me a luddite, but I simply cannot trust a technology that takes something as crucial as controlling my automobile out of my hands.


How do you think your ABS and traction control and engine and transmission work now? Quite a bit of the operation of your car now is very likely computer-mediated. Your computer can almost certainly kill you anytime you're driving at highway speeds. ("Say, this guy's p-ssed me off. Let's fire the airbag and pretensioners, wait half a second ignoring brake signals, and then lock the brake on the left front tire only.")

ABS, I can't avoid, but there's a reason that I drive a manual car with no traction control. It is seriously limiting my choices for the replacement.

This law doesn't outlaw delivery car/carts. Honestly I think self-driven delivery cars are a natural first step for the technology. If you are only setting out to deliver light cargo, you can have a much smaller engine and a very light chasis -- you wouldn't need to worry about safety of passengers. Hopefully it would be light enough that standard cars would not be endangered in a crash.

Anyway, this approach seems likely to skirt many legal difficulties... It might finally make home delivery of groceries and sundry viable.

I imagine an automated Segway cruising around the city, delivering pizzas.

And then I imagine college kids picking it up and putting it in the dumpster.

How much more would you be willing to pay for the driver-less perk? How much more would google and its ilk be charging for this? Is this technology economically viable?

I guess some of the cost could be offset by the fact that the driverless car wouldn't need some (most?) crash barriers like bumpers and so on, couldn't it?

Not all cars will go driverless. The ecosystem will be mixed. You need to compensate for the other fellow's idiocy.

If you think the Falls Church ordinance would be interpreted to apply to someone who is not really operating the vehicle, you have probably forgotten this case: http://blogs.wsj.com/law/2010/12/01/the-case-of-the-school-bus-the-stop-sign-and-the-missing-at/

The linguistic analysis in the article is wrong. The sentence can be parsed two ways: with intransitive "stop" and transitive "approaching" (taking "school bus" as the object), which makes sense, but is missing a comma after "approaching"; or with transitive "stop" (taking "school bus" as the object) and intransitive "approaching", which makes no sense. The court allowed the latter (nonsensical) interpretation and acquitted the driver. I thought it was an awful decision, but a very smart friend of mine, who is a lawyer (though not a criminal defense lawyer), told me it was almost a textbook case of statutory ambiguity and the court basically had no choice.

That was in Virginia, so the same judicial system that would interpret the Falls Church ordinance.

You should also note that re: the Fiorino fiasco in Philly, knowledge of the law is no excuse.

Google is trying to get Nevada to change their laws:


I predict it will happen in Japan (where trial lawyers aren't as influential and robots are beloved) and China (where life is cheaper and traffic worse) first, and then Americans will adopt it after a decade or so of mockery.

I predict that if it happens here, it will creep in through technologies like stability control, which will be increasingly willing to take over if the computer thinks an accident is imminent.

Rahul -

The market assessment shows driverless technology should be viable around $4k per vehicle for equipment (excluding any services which may be ancillary). All the main target customer groups, including seniors, soccer moms, businessmen and taxi companies, would have a payback period less than a year. That should be adequate to insure timely adoption, exclusive of concerns about safety. Thus, there is a perceptual hump to get over, but once surmounted, the market analysis suggests wide and rapid adoption.

Is the cost that low? I really am surprised (pleasantly).

Machines think they know everything.

“On January 25, 2010, the National Archives announced in the Federal Register that filming, photographing, and videotaping by the public will be prohibited in all exhibition areas in the National Archives Building, Washington, DC, beginning February 25, 2010. [. . .] In spite of a more than 30-year-old regulation explicitly stating that flash photography was prohibited, prominent signs stating the policy throughout the exhibition areas, and security guards reminding the public, Archives staff estimated that the documents were subjected to approximately 50,000 flashes a year. While enforcement of this policy has always been a National Archives priority, new cameras with automatic flash have made the policy almost impossible to enforce.” (link)

Here in the (SF) Bay Area, we call this "ghostriding the whip" -- seen here: http://www.youtube.com/watch?v=8yuLtnA_dBk&feature=fvst


How many driverless vehicles are operating in Afghanistan and Iraq?

Isn't that the best test case because the government is already paying huge amounts to truck goods through dangerous conditions where significant loss of life has occurred, and the military must provide protection on the ground to defend the drivers?

Oh, wait, we've switched to a policy based on human life being cheaper than technology.

With lots of excess labor in the US, with lots of economists arguing people should be forced to take jobs at ever lower wages, how can technology compete with cheap humans?

What problem does driverless cars solve?

As Tyler travels outside the US, I'm sure he finds lots of people employed as drivers so the owner of the car, or the user, has no need for driverless cars.

This sounds like the hype for Ginger which was announced as the Segway. Segway spent a lot of money lobbying to make the Segway legal in every State and city and town.

The market all the bloggers and tech heads predicted never appeared, but Segways have found markets as the way to zip around tourist and historic areas, and the worst accident publicized was Bush falling over when he stepped on a powered off Segway, yet the Segway is being attacked by politicians and pundits as dangerous hazards to everyone.

Why would a driverless car fair better than the Segway?

Wow these are even easier than usual.

On the war zone:
1. If you are carrying people, getting rid of the driver only reduces you count by 1, who could prove useful in a firefight anyway
2. If you are carrying valuable cargo, you still want it under guard so someone doesn't steal it.
3. The cars are designed to stop for any pedestrian. In the city this is great. In a war zone this is bad.

how can technology compete with cheap humans?

The same way I don't hire homeless people to sift through Google results for me.

What problem does driverless cars solve?

As we've already said, there are elderly and incapacitated people who can keep their independence. They tell the car where to go and it goes there.

Also, lots of people hate driving.

Why would a driverless car fair better than the Segway?

Because the Segway was different. A computer-driven car is like a meatbag-driven car, but safer for its passengers and safer for people outside of it.

Outside of airport shuttles, we barely have driverless trains without attendants to look after them, and trains ride only in a straight line with a very low probability of obstructions showing up in front of them.

With our current level of technology, I doubt that you could make a decent driverless car that uses pattern recognition or range finding within a video game much less in real life. We may knock human drivers, but our failure rate is incredibly low compared to most modern computers. If you had a decent modern personal computer being operated for 1000 hours and a human driving a car for the same amount of time (non contiguous in both cases), which one is likely will crash first?

You know we've had driverless cars driving tens of thousands of miles already, right?

You know the only accident was when the car got rear-ended at a stoplight, right?

You know these cars don't run a consumer level OS, right?

How many thousand miles have the driver-less cars been driven and how do their accident rates stack against human drivers? Is it a statistically significant improvement.

I'm finding it hard to find accident rates (since so many people want to talk about fatality rates), but I think the accident rate for humon drivers is 1 per ~180,000 miles.

That sounds really low. Don't most cars only get driven around 10,000-15,000 miles per year? That would imply one accident (including minor ones) per car per *decade*.

According to wikipedia, the longest a driverless car has been operated without human intervention of any kind is around 1000 miles. In that case it was the Google driverless car. There is a substantial thread about liability and reliability in the discussion section of that page that might be worth reading. If your estimate for human error is correct at 1 accident per ~180,000 miles then there's no evidence that driverless cars are close to reaching or exceeding that goal at present.

I am less concerned than Tyler about regulation strangling the driverless car in its cradle.

For example, NHTSA had once promulgated a rule that required a car's engine starter to remain inoperative while the car is in in gear, so as to prevent a sudden lurch when the engine was started. That made sense...until hybrids like the Prius and Insight appeared on the scene. Hybrids "work" by starting and stopping their engine while in gear. NHTSA interpreted the rule in favor of their demonstrably beneficial, and otherwise safe, operation.

See Federal Register from July 1, 2005.

When the Google Guys - and many others! - demonstrate the technology in a more-than-experimental way, then we can start fretting about law, lawyers, regulations, and regulators. Let's conquer one mountain at a time.

This shows which they last very much lengthier and thus saving you income which could otherwise are actually utilized to purchase new ones.

This shows which they last very much lengthier and thus saving you income which could otherwise are actually utilized to purchase new ones.

I'm no lawyer, but if i was one, I think I could make a few dollars arguing that if you aren't operating the car, that law doesn't apply.

Comments for this post are closed