Autonomous Cars Aren't As Smart as They're Cracked Up To Be (computerworld.com)

Gill Pratt, executive technical adviser at Toyota, offers a note of caution, even as more car companies start putting AI elements into their cars. Speaking in Tokyo at the announcement of a Silicon Valley AI research center that Toyota is to open in early 2016, Pratt pointed out the big shortcoming in an AI system as applied to automobile: Autonomous cars might look great in controlled tests or on pristine highways, "but soon fail when faced with tasks that human drivers find simple." From the article: Drivers, for example, can pretty much get behind the wheel of a car and drive it wherever it may be, he said. Autonomous vehicles use GPS and laser imaging sensors to figure out where they are by matching data against a complex map that goes beyond simple roads and includes details down to lane markings. The cars rely on all that data to drive, so they quickly hit problems in areas that haven't been mapped in advance. ... A truly intelligent self-driving car needs artificial intelligence that can figure out where it is even if it has no map or GPS, and manage to navigate highways and follow routes even if there are diversions or changing in lane markings, he said. I regularly drive a stretch of road that's just a few miles long, but between construction, accidents, poor marking, bicycles, and heavy traffic I'd be nervous about letting an AI system navigate. In what real-world driving scenarios would you most want humans to take over?

Comments

That's OK, I only care about bar crawls

[drinkypoo]

As long as the car can drive me home after the last bar in the line, I'm happy.

Re:That's OK, I only care about bar crawls

[Rosco P. Coltrane]

As long as the car can drive me home after the last bar in the line, I'm happy.

That's called a taxi, and it's cheaper than an autonomous car. The only downside is, if you barf on the back seat, the cab driver might smash your teeth in - something the autonomous car won't do.

Re:That's OK, I only care about bar crawls

[drinkypoo]

That's called a taxi, and it's cheaper than an autonomous car. The only downside is, if you barf on the back seat, the cab driver might smash your teeth in - something the autonomous car won't do.

I also don't like to play grabby-squeezy in a taxi, that's disgusting on multiple levels.

Re:That's OK, I only care about bar crawls

[sumdumass]

The problem with taxis (if there actually are any running that late in your area- mine there aren't) is that if you didn't start off intending to drink, it costs double because you have to retrieve your car. Of course it could be double anyways because you have to get to the bar in the first place.

It would be so much easier to just program the destination in before having a drink (say on way home from work or something) and then just push some buttons to get home. The vast majority of my drinking away from home is spur of the moment where i got invited from somewhere else or work was such a bitch i need to unwind a bit before going home.

Re: That's OK, I only care about bar crawls

[t1oracle]

Cheaper? An AI doesn't need a salary, or a medallion, and it's insurance will be cheaper. If taxis were cheap, most of us wouldn't have cars.

Re:That's OK, I only care about bar crawls

[vtcodger]

if you barf on the back seat, the cab driver might smash your teeth in - something the autonomous car won't do.

I'm confident that with a little work, cars can be programmed to express dissatisfaction with driver behavior. Maybe release the seat belt, run the speed up to 50kph, then stand on the brakes.

Just don't expect all the features to be there in V1.0 of the software.

Re:That's OK, I only care about bar crawls

[TWX]

I also don't like to play grabby-squeezy in a taxi, that's disgusting on multiple levels.

At least with an autonomous car you'll have no one there to disgust...

Re: That's OK, I only care about bar crawls

[MobileTatsu-NJG]

Why would its insurance be cheaper?

Re: That's OK, I only care about bar crawls

[mysidia]

As customer of a Taxi, you don't have to pay any of those things, only a miniscule portion of them.

The AI is more expensive, because all new cars are super-expensive, and you have to add Research and Development costs, "brand premium", And "coolness premium" the manufacturers will charge b/c the thing can drive itself.

If instead of buying a $12,000 used car that meets all your needs, you spend $60,000 on a brand new car-that-can-drive itself and lasts 10 years, plus a $50 monthly service fee for the cloud maps service, then you're paying approximately $5000 extra a year for self-driving capabilities.

That would buy you 333 $15 taxi rides.

Anyways, based on that, unless you spend more than $5000 a year on the Taxi, then it just isn't a worthwhile economic proposition.

Also, the self-driving cars are probably going to be introduced at about the $120,000 price point, not the $60,000 price point.

Then you'll also spend an extra $10,000 in vehicle loan interest per year to get the self-driving feature.

Re: That's OK, I only care about bar crawls

[peragrin]

Where do you find a $15 taxi? for $15 you can go about 3 miles. if i am that close i will just walk, and i will be sober when i get home.

The last time I got a taxi it drove me 5 miles and cost $35 plus tip. Uber helps, but taxi's aren't cheap.

Re: That's OK, I only care about bar crawls

[Maxo-Texas]

Even a tiny taxi ride in my town is at least $40.

And you may have to wait up to an hour for a taxi.

And when you really need them most (like after new year's eve), it could be four or more hours.

And one reason uber sprang up besides cost is that taxi's just don't like to go to low density or "bad" areas. They'd prefer to have a fare back as well.

Taxi's (and public transportation) make more sense in New York City and similar locations perhaps. When the city is more flat than vertical, Taxi's (and public transportation) make less sense.

Owning a vehicle is expensive. It's an extra cost in flat cities. People who don't have a car do get extra spending money. But in areas where you must drive 30+ miles every day, taxi's are not an effective option (and neither is public transportation).

I think what the parent poster needs is less of a street legal car and more of an automated bicycle type vehicle that could get him home drunk in an hour or less while he's plastered.

Re: That's OK, I only care about bar crawls

[mysidia]

Where do you find a $15 taxi? for $15 you can go about 3 miles.

In N.O., a cab is about $4 to start, plus $2.50 a mile, so 5 miles is about $16.50 plus $1.70 tax. You could easily do some walking, take a streetcar to get across town, and then find a cab ride for the last few miles.

Even if you had to pay $30 for that cab ride home, though... it's still not going to justify the expense of a self-driving car economically. Seriously.... who goes out and get drunk at a bar more than two or three nights a week, and is still employable.? sounds like an alcohol addiction. The price of booze far exceeds the price per ride of transportation, at that point.

The reasons to purchase a self-driving car would still be comfort and convenience. (Not that there's anything wrong with that)

Re: That's OK, I only care about bar crawls

[mysidia]

And when you really need them most (like after new year's eve), it could be four or more hours.

Convenience. Trade $$$ for time. These are good non-financial reasons to want a self-driving car.

Even a tiny taxi ride in my town is at least $40.

This is still workable. Suppose you go out 2 nights a week; 3 weeks per month.

4 taxi rides X 3 X 12 X $40 = $5,760/Year

That still looks better financially than the figured $48,000 upfront + $600/Yr, that comes to $5400/Year.

What we haven't mentioned.... is that when you take Taxi rides, you are not committed upfront; there is a certain value in not having to front money, and pay $40 as you go.... most of the money you would spend on taxis is delayed upfront, so you can be earning 2% CD interest, or 10% stock market returns on most of that cash for more than 5 years in the Taxi scenario.

Another thing is that it's an unrealistically optimistic assumption that your brand new self-driving car will last 10 years, most cars won't last beyond 5, a self-driving car is more complicated, and therefore, likely more fragile/fault-prone; we have not considered the additional maintenance repair costs, and equipment replacements that will be required to keep this self-driving car working with all the sophisticated added sensors, GPS antennae, 4G LTE radios for internet, and computers needed for it to function.

With conventional cars, drivers typically pay more than $5000 a year in maintenance costs; it's not far fetched to consider that the additional sensor arrays may add at least a few more thousand bucks worth.

We haven't talked about the additional expected towing costs and inconvenience, when the car with no steering wheel needs a tow, because one of the sensors had a rock crack the glass, resulting in "safety shutdown".

Re: That's OK, I only care about bar crawls

[non0score]

Isn't the Google self driving car supposed to be cheap?

Re: That's OK, I only care about bar crawls

[Your.Master]

Aside from the usual arguments about AI drivers likely driving safely -- an automated taxi has 1 fewer human inside of it to be injured. Even if AI drivers generally had a slightly worse record than humans in this hypothetical future, an AI taxi might have cheaper insurance :).

Re: That's OK, I only care about bar crawls

[Jack Griffin]

Who said insurance will be cheaper? When people learn to steal or vandalise these things, insurance is going to make them economically unviable.

Re: That's OK, I only care about bar crawls

[mysidia]

Isn't the Google self driving car supposed to be cheap?

I think the Google self-driving car is more like a proof of concept they're working on, and Google will probably be more like an "arms dealer" attempting to license technology and designs to be incorporated by existing car manufacturers --- who always add a huge price premium to new models, especially when new features are included.

I see "self driving" as being a luxury feature not included in base models, at least initially.

I just don't see Google ever becoming an auto maker, that's so far removed from their core business.

Re: That's OK, I only care about bar crawls

[stoatwblr]

Or, you find that the taxi company buys AI cars and dispenses with drivers (which are the single largest expense)

Re: That's OK, I only care about bar crawls

[Maxo-Texas]

Holy moley, your figures are way off. Per consumer reports, "Most Lexus models have relatively high maintenance and repair costs (primarily due to maintenance), despite their excellent reliability. The Lexus ES350 racks up an average of $2,300 in maintenance and repair in the first five years, about twice what you'd pay on a Buick LaCrosse." A full order of magnitude lower than your proposed cost at only $460 a year. Your figure is closer to the "total cost of ownership".

Most cars "last" well beyond 5 years these days. It's well documented that the average age of cars in the usa is over 11 years.

I could see the new unproven auto-driving parts failing sooner so I'd want a warranty to cover them for 8 years. I have an 8 year/120,000 mile warranty on my current car.

Similarly, a recent study showed for the last 10, 20, and 30 years the average investor has only made between 2 and 3% annualized return in the stock market after taxes. (Forbes) Likewise, I'd love to know where you can find a 2% CD from a trustworthy safe source. Best I can find are all 1% or lower.

Similarly- you use your car for things besides going out. Shopping, vacation travel, emergencies, visiting friends and relatives.

So you need to add in a lot more taxi rides and a couple weeks of outright car rental.

I get it, you are enthusiastic- but you are shading every fact you want to be true to be larger than reality while discounting or outright omitting every fact that undercuts your position.

Just 4 taxi rides a week for a year is close to the total cost of ownership for a car. So you need to add the cost for public transportation to work twice a day. And to be fair, you should value your time spent waiting for a bus/train/taxi at your hourly rate. And you need to add the cost for car/truck rentals (and the time it takes to rent them). Finally, there are intangibles like waiting in miserable weather outside the club after it closed for a taxi to come get you. Or when your parents/syblings/best friend/you have an emergency and you sit waiting for a taxi instead of dealing with it immediately. If you had to take an ambulance instead of a taxi, that's closer to $1200 (or more) for just one ride.

I agree the are cases (as I said- vertical cities) where taxi's work well. Taxi's are not a panacea and their utility comes and goes in cycles. Some years, it gets pretty bad (especially no-shows or refusal to pick up certain people padding the fair with a circuitous route, etc.).

Re: That's OK, I only care about bar crawls

[dave420]

Insurance companies don't just make wild guesses as you just did - they will spend a long time calculating their rates. And it's probably not a good idea to steal or vandalize hardware covered in cameras with an internet connection and precise knowledge about where it is, so I doubt that would be as big a problem as you seem to think it is.

Re: That's OK, I only care about bar crawls

[Sylver Dragon]

I just don't see Google ever becoming an auto maker, that's so far removed from their core business.

Considering their re-organization into Alphabet Inc., they may well be considering pushing even further afield than they already have. Would that include a car division? Probably not, I wouldn't be at all surprised if they sought out an established partner; but, they have show a willingness to jump into markets with large, entrenched players (e.g.: Google Fiber).

Re: That's OK, I only care about bar crawls

[Jack Griffin]

And it's probably not a good idea to steal or vandalize hardware covered in cameras with an internet connection and precise knowledge about where it is, so I doubt that would be as big a problem as you seem to think it is.

You mean like every single bank in the world that ever got robbed,or every 7eleven, or the entire City of London?
Your opinion of the crime deterrence of cameras seems to have no basis in reality.

Silly

[umghhh]

This Gll dude is just silly - he does it all wrong, if he made an intelligent car with 3d printer as discussed here many times this would solve all the problems. Even more - if the whole world had been made with a giant 3d printer then there would be no problem with white spots on maps used by autonomous cars.
At least at the beginning when the model and reality were still almost the same.

It will be like service areas for phones

[Immerial]

You'll look up where you'll be using the car to see if it's covered. I think for 85% of the folks, it would work fine for them: commuting the same paths to work, to the stores, to friends and family houses, etc. There would also be plenty of cases where an autonomous could wouldn't work well. [sarcasm]What?! A solution doesn't work for everyone?![/sarcasm] :P

Re: It will be like service areas for phones

[ememisya]

I think the really jammed up and extremely non-adventurous city traffic would appreciate auto driving IF every car was equipped with it. Because then that invisible bottleneck where car A moves first, car B moves, then car C wouldn't happen. Cars A, B, and C would stop and move at the same time. Other than that I don't want one. In rural VA it's really fun to drive, plus or minus a few insane drivers.

So just have the cars drive where it is easy

[LetterRip]

You don't have to have the car drive everywhere, 95% of the places you drive will probably have all of the factors needed for the car to navigate easily. Just don't have the car drive in areas where it can readily get in trouble.

You don't start teens off in ambiguous hard to drive conditions, but rather low traffic side streets or empty parking lots, etc.

We don't need self driving cars that are perfect from the start, merely good enough to drive us most places most of the time, and do not have accidents in the areas that are suitable for it to drive.

Re:So just have the cars drive where it is easy

[Anonymous Coward]

Don't know where you live, but here they change lanes around, close lanes, and reroute traffic on a nearly daily basis. Its been non-stop doing that for over 10 years straight. If auto driving cars depend on GPS that means I can't use it to just go to work even, not to mention going somewhere else.

What happens when they close part of the city for a parade and a lot of streets are closed? Your car just stops and waits the 4 hours for it to pass before continuing? Sure you could take over, but if your drunk and you do that you committed a DUI, the reason you bought a self driving car to start.

Re:So just have the cars drive where it is easy

[Anonymous Coward]

If you are constantly drunk during day(parade and all...) maybe it is time to talk to someone in AA?

Re:So just have the cars drive where it is easy

[tompaulco]

The safe bet is to only use an autonomous car in the laboratory where it was designed. Just like a lot of software, it only really works in clean room, optimized conditions with hand picked data.

Re:So just have the cars drive where it is easy

[tmosley]

Waze figures that out pretty quickly.

Big data, son.

Re:So just have the cars drive where it is easy

[AthanasiusKircher]

You don't start teens off in ambiguous hard to drive conditions, but rather low traffic side streets or empty parking lots, etc.

We don't need self driving cars that are perfect from the start, merely good enough to drive us most places most of the time, and do not have accidents in the areas that are suitable for it to drive.

This might be true if we lived in a "rational" society. We don't. We live in a society whose concerns are driven by media hype. These days, the media seems to be on the side of self-driving vehicles because they seem really cool and awesome as a concept.

But the media is fickle and could change its mind the moment something more sensational happens.

We've already seen issues with idiots using the Tesla "autopilot" feature in ways it wasn't intended, and the company is starting to rein in its use to prevent idiots from using it in cases where it's likely to fail.

The thing most people seem to be forgetting is that most cars are effectively multi-ton projectiles that easily have the potential to kill. And it's not feasible to expect human drivers to take over in a split second to stop an accident when the software fails.

All it takes is one idiot who puts a car into self-driving mode in a place where the software can't handle it yet, and that car inadvertently causes an accident involving a schoolbus and kills a couple kids (or even hits a family traveling in a van on vacation or whatever). Can you imagine the fallout??

It doesn't matter whether a human driver could have avoided the accident or not. The headlines will just read "AI CAR KILLS KIDS." The (probably rich) guy who drove it will be horrified as he gets sued, leading other rich guys to avoid buying the cars. The car company, engineers who designed the software, etc. will also get sued. And heaven help them if they built in some small choice in their software that led the car to drive into the bus rather than effectively "committing suicide" if the collision was avoidable... suddenly all those ethical debates about whether it's okay to kill kids on a bus rather than doing a potentially suicidal maneuver to avoid it will come into play.

By this point, Congress and state leaders will step in and start banning these cars until safety issues can be worked out. Self-driving cars will be set back at least a decade or two as they are subjected to ridiculous regulatory constraints that human drivers would never have to deal with.

Will this nightmare scenario come true? I'm sure every executive who is betting on self-driving cars hopes to heck it doesn't. And any of them who have any good sense are trying their best to ensure the cars are as nearly "perfect from the start" as they can make them, to avoid this nightmare scenario.

Of course, in additional to living in a society driven by media hysteria, we also live in an economy driven by greed -- so despite safety concerns, I'm also sure that many of these companies are trying to beat their competitors in getting these self-driving systems out.

But your analogy with a "teen driver" doesn't work at all. People who own these cars will use them inappropriately, even if warned not to. And if AI cars kill someone (even if it couldn't be prevented), it's not going to be a teen losing a license or going to prison... it could ruin the whole industry for decades.

Re:So just have the cars drive where it is easy

[TWX]

Don't know where you live, but here they change lanes around, close lanes, and reroute traffic on a nearly daily basis. Its been non-stop doing that for over 10 years straight. If auto driving cars depend on GPS that means I can't use it to just go to work even, not to mention going somewhere else.

What happens when they close part of the city for a parade and a lot of streets are closed? Your car just stops and waits the 4 hours for it to pass before continuing? Sure you could take over, but if your drunk and you do that you committed a DUI, the reason you bought a self driving car to start.

I expect that in addition to cars being able to self-determine routes and find barriers there will need to be intelligent barriers that the cars can detect and follow the instructions of. These kinds of barriers would be used by construction crews, emergency responders, and perhaps even as a function of the four-way hazards when a car is stopped on the side of the road. Call it a more precise means for the autonomous car to determine what it should do or what the expectation is in a complex situation.

Just as an example, in long-term highway construction projects it's not uncommon to take a two-lane-single-direction stretch of Interstate and to route both directions on it, one going the natural way, the other driving what would normally be opposed, while the other two-lane stretch is being worked on. In cases like this there needs to be a way for the construction barriers themselves to notify the vehicles both that something has overridden the expected behavior, and that this particular path is the override. The car will in-turn have to account for this deviation in the path and to know that it's not actually trying to go the wrong-way even though its default programming would say that it is, and it would have to understand that while one lane is now no longer the wrong, way, the other lane still is the wrong way and to not try to use it.

Other construction-related examples include the ability to follow a pilot car and the ability to pay attention to flag-men. The flag-men method is a variation of the one-lane bridge in many cases with the addition of a very spontaneous control (ie, the switch from slow to stop and stop to slow comes without warning from the flag-man himself, so the vehicle must pay attention to the flow of traffic in addition to somehow figuring out the sign or receiving a signal from the sign), and the nature of pilot cars means that there has to be some means for cars to be subordinate to other vehicles, which leads into the next example...

...emergency responders. Cars will need to respect things like fire trucks blocking the road, or police cars blocking the road, or tow-trucks blocking the road, or any other sort of obstruction that will be present for awhile and indicates that it isn't safe to be within a certain area. Cars may also have to react to barriers placed by these responders, and it may make sense for those barriers to have some kind of component that lets them more intelligently broadcast so that the cars don't have to figure out what they are visually. Obviously if the police are attempting to close a stretch of road due to an accident investigation they want to keep cars out of that area so that the evidence is not disturbed. If firefighters are working on a structure fire they need to keep cars out of the immediate staging area and from driving down the road that the firemen may be crossing regularly without notice. They also need to keep a wide berth when a tow truck driver is working with a disabled vehicle, wherever that vehicle is disabled and whatever is wrong (ie, difference between an overturned vehicle on the highway, a stalled vehicle on the highway, and a stalled vehicle on the median or shoulder). These are all complex situations that happen all of the time, and cars need to be able to handle them.

I think the first application for autonomous cars will be open-highway dr

Re:So just have the cars drive where it is easy

[ArylAkamov]

Don't know where you live, but here they change lanes around, close lanes, and reroute traffic on a nearly daily basis. Its been non-stop doing that for over 10 years straight.

Hi Seattle!

Re:So just have the cars drive where it is easy

[Jack Griffin]

Don't know where you live, but here they change lanes around, close lanes, and reroute traffic on a nearly daily basis.

A few scenarios I've had this week:
A major set of lights was down, with cops directing traffic. Can robot cars deal with this, or do they just become paper weights when faced with unknown circumstances?
A major crash closed a road, so cops were directing traffic over a grass verge, around the accident site, then back onto the road again. I can't imagine any amount of AI will know how to deal with this.
On the main road is what looks either 20kg weight plate (from a gym), or maybe a base from a sun umbrella, right in the middle of the road. It's been there a week and everyone just drives over it as it fits inbetween your wheels and is low enough to clear the car. Will AI be able to handle that type of obstacle? Or will it think it's a cat?
Robot car has fail written all over it.

Re:So just have the cars drive where it is easy

[bingoUV]

These kinds of barriers would be used by construction crews, emergency responders, and perhaps even as a function

And vandals, black hats, hackers, script kiddies, drunkards, people out to have fun.

Re:So just have the cars drive where it is easy

[dave420]

Let me sum up every single one of your questions:

"I don't know how this works. I have absolutely no idea. I also seem to think that if i don't understand something, no one can. I am the be-all and end-all of human knowledge."

So just because you can't figure out how to make a car follow directions (or other cars), doesn't mean the countless people brighter than you can't.

Re:So just have the cars drive where it is easy

[Jack Griffin]

Let me sum up every single one of your questions:

"I don't know how this works.

Same goes for flying cars yeah?

So far it's me 1, Flying cars and robot cars 0.
Let me know when the robot cars take over the world, I won't hold my breath.

Two camps

[transporter_ii]

There seem to be two camps of people. Those that think we will be living on mars and have fully autonomous cars in a couple of years, and those that actually look into it and see how hard it is going to be. For some reason, the media seems to prefer the first one. Reality prefers the second one.

Re:Two camps

[zippthorne]

We're in the future right now - I'll lay even odds you have in your pocket or nearby right now, a personal magic map that nearly always knows where you are and can show you how to get to where you want to go, and you can even use it to communicate with people over a distance!

Re:Two camps

[gurps_npc]

Bull shit. No one I know of thinks we will be living on Mars in 5, 10 or even 15 year.

You are a cynical pessimist that is unaware of the state of current technology and the deep need for it.

Yes, if the choice was between a sober, experienced, 30 year old driver familiar with the roads and a self-driving car, no self-driving cars would ever be created. That is NOT the market for them.

The market for self driving cars will start out being wealthy parents with kids that have a history of drinking alcohol, upper class 70 year olds that can't afford a chauffeur, but love their retirement in the country, and municipalities with weak unions/ (garbage trucks, public busses, etc.)

We are effectively in the testing phase for these products. They will not be perfect, but they will be BETTER THAN EXISTING human drivers. They don't have to work in the rain, snow, or other bad conditions.

But over the next 5-10 years we will begin legally selling self-driving vehicles with limited licenses for specific areas/times.

Then the companies will re-invest the profits, driving huge improvements in the technology. Within 20 years, it will be ready for the mass market.

Re:Two camps

[Jack Griffin]

The first camp are also people who believe flying cars are only a couple of years away. We all know how that one is working out...

Re:Two camps

[dave420]

You also forgot the third and fourth 'camps':

3. Those who understand that it is possible for autonomous cars to be better than human drivers (and that they already are in a wide variety of situations), and who see that the market is going to drip these automation features into new models of cars as the years progress, with some manufacturers opting to introduce more features more quickly compared to others.

4. People who make ridiculous oversimplifications in order to tortuously make a dig at a group of people they don't agree with by concocting a complete misrepresentation of their concerns and understanding, and then attacking it.

You just claimed the world's experts on autonomous vehicles think autonomous vehicles simply can't work. Clearly that's not the case, so you must be wrong. I guess that's the danger of massive oversimplifications and generalizations - they're so frequently wrong it makes the person making the claim look an absolute gibbering idiot for ever thinking it was a coherent, intelligent thing to say.

Re:Two camps

[Cro Magnon]

Maybe we'll have autonomous cars on Mars. Should be easier, without existing traffic and people to worry about.

Re:That's nothing

[belthize]

I don't see the issue.

For many scenarios where a human will have to fave that decision the autonomous car never will because it would have chosen option C, avoid situation long before it became an issue. Take the Oklahoma parade a few weeks ago, an autonomous car would not have had to decide should I kill the drunk driver or plow into the parade.

But lets pretend that the car has to decide and chooses poorly and wipes out 10 people in a crowd including little kids. How is that a problem if across the set of all autonomous cars 100 such occurrences were avoided. Choices should be made on relative merits, X is better than Y. Not X is perfect or X is imperfect.

Are you trolling or just boring?

[drinkypoo]

People keep saying this, but the truth is that the car is going to [be programmed to] follow the law. That means it's going to approach intersections at safe speeds, and it's going to avoid hitting pedestrians in crosswalks but will simply murderize them even if there's ten of them in your lane, and a cancer-ridden octagenarian driving a yugo in the other lane — even if the car has enough sensors to smell cancer, it's still going to run right into those pedestrians like you've gone bowling rather than deviate from the marked lane. It's going to make a good-faith best effort to stop. But remember, it's not going to go around a blind curve at a speed at which it can't stop if there's an obstacle. It's simply going to decelerate for the curve, and then accelerate again on the other side. If someone is in the road, it won't hit them, because it's not driving for fun. It's driving to minimize risk.

Re:Are you trolling or just boring?

[Kjella]

People keep saying this, but the truth is that the car is going to [be programmed to] follow the law. That means it's going to approach intersections at safe speeds, and it's going to avoid hitting pedestrians in crosswalks but will simply murderize them even if there's ten of them in your lane, and a cancer-ridden octagenarian driving a yugo in the other lane â" even if the car has enough sensors to smell cancer, it's still going to run right into those pedestrians like you've gone bowling rather than deviate from the marked lane. It's going to make a good-faith best effort to stop. But remember, it's not going to go around a blind curve at a speed at which it can't stop if there's an obstacle. It's simply going to decelerate for the curve, and then accelerate again on the other side. If someone is in the road, it won't hit them, because it's not driving for fun. It's driving to minimize risk.

Yes, and apart from the law-abiding aspect there's also the whole "is it reasonable to kill someone else to save you from your own stupidity" argument. Let's for the sake of argument say you're going 55 mph and two people drop down from an overpass smack in front of the car and the car can either run them over or slam into a concrete wall, two lives versus one since you have no passengers. Or alternatively, that you could mow down one person on the sidewalk. Sure you could save lives but I'd say it's the people "at fault" that deserve to die, not anyone else. Or deserve is too strong a word, but they don't deserve that anyone else is sacrificed to save them. It's actually harder if you say the other way around, if we could drive the car into the ditch sustaining light to no injury do we have the right to mow them down anyway. I'd say legally you'd probably not charge a human driver in that situation, even if it's not optimal just hitting the breaks is an acceptable response. And that should be enough for a computer too.

Re:Are you trolling or just boring?

[kheldan]

Sure. But is it going to slow down or come screeching to a halt when some dumb cat or dog runs out into the road, and cause the vehicle behind it to rear-end it, or is it going to know that while squishing Fluffy is sad and regrettable, it's preferable to causing an accident in which property will be damaged, and more to the point, where human beings might be injured or killed? It won't, because it won't differentiate between a cat, a dog, a squirrel, or a human child, or some kids' balloon that drifts on the breeze from his birthday party into the roadway. Please don't talk about how this won't be a problem because all cars will be autonomous cars and never operated by a human being because that's about as likely to happen in our lifetime as practical and widespread use of fusion reactors to supply electricity is likely to happen in our lifetime. Also there will always be bicyclists and motorcycles, and you can't make those 'autonomous' so there's always going to be other variables in the traffic equation that you can't control and can't accurately predict the behavior of, and before you say it, claiming that we'll outlaw those things is at best unrealistic. Also, I don't want to live in a world where you can't 'drive for fun', and I know I'm far from being alone in that sentiment. Honestly, the whole 'autonomous car' thing is so completely overblown that I don't know how anyone can take it seriously anymore. It's just not going to happen anywhere near like some people think it's going to happen, and I don't understand why some people can't see that.

Re:Are you trolling or just boring?

[BoberFett]

I would guess that the car will be traveling at the posted speed limit meaning it will likely have time to slow down, and if it absolutely doesn't have time to slow down, it won't. Furthermore, the automated car behind it will be traveling at a safe distance behind it, not tailgating as asshole humans are known to do, so it will have plenty of time to slow down as well.

The point is, cars should never have to make that kind of decision because they aren't shitty drivers to begin with.

Re:Are you trolling or just boring?

[kheldan]

There you go, assuming that every car on the road is magically going to be driven by a computer 100% of the time, meanwhile I keep telling you people: It ain't gonna happen. Dream all you want, friend, it's going to remain science fiction for at least the rest of your life, so please be sure your driving skills are adequate, you're going to need them.

Re:Are you trolling or just boring?

[BoberFett]

Whatever you say, Nostradamus.

Re:Are you trolling or just boring?

[drinkypoo]

Sure. But is it going to slow down or come screeching to a halt when some dumb cat or dog runs out into the road, and cause the vehicle behind it to rear-end it,

No, because rear-endings are caused by insufficient following distance.

It won't, because it won't differentiate between a cat, a dog, a squirrel, or a human child, or some kids' balloon that drifts on the breeze from his birthday party into the roadway.

They're going to see that stuff before it even enters the roadway, so they don't have to hit it to begin with. In situations where the sides of the roadway are obscured, the vehicles will simply reduce their speed — the same thing a human driver should be doing.

Also there will always be bicyclists

In California, we're already required to stay three feet away from them at all times. That helps account for their inherent variability. A self-driving car won't get pissed off when it gets held up by some cyclist who won't pull over to permit passing, as the law demands.

and motorcycles,

Not only can you reasonably make a self-driving motorcycle eventually, but they'll eventually be prohibited from lane splitting anywhere, which will make them easier to predict.

Also, I don't want to live in a world where you can't 'drive for fun',

Support your local racing organization.

Re:Are you trolling or just boring?

[Catbeller]

Well done.
I think Ian Malcolm, the chaos theoretician in Jurassic Park gave the best (in the novel) explanation as to why we can't outsmart the universe. Mathematically, chaos wins. Life/Chaos beats the best-laid plans of mice and men. Humans are best at all-purpose threat response in unusual situations. The sheer audacity of the belief that we can build a billion two-ton computers and let them loose is staggering.

I stand watching a computer-dispatched elevator sit on a floor, opening and closing its doors, and no one can make it stop. They have to call Otis every time. And it happens a lot. That's a damned simple sensor and interface, and it fails. No one pays much attention because there are five other elevators. If the elevators were doing seventy among hundreds of other elevators, it would be more tragic.

A car would consist of two or more networks, dozens of sensors, hundreds of millions of lines of code that absolutely must work, and would work best if all the other cars were running the same code and are integrated into a worldwide network. This assumes the sensors don't fail eventually, or that the owner doesn't, and this might shock some Americans, trade in their car every two years, but instead keep their cars for a decade or more. They do that because new cars last that long, easily. A robot car would fail in hundreds of ways in that ten years, because of normal failure of all that incredibly delicate self-driving equipment. I guess everyone HAS to buy new every few years, as the manufacturers really, really would like it if they do, so I assume code updates won't be all that forthcoming on older models. Bricked or new, your choice. Cha-ching. No wonder they're so very, very happy to build robot cars. It's a license to drain all the damned wallets.

And of course, we can't ignore that car AIs with outside network connectivity are, by definition, remote-control ready. If anyone thinks this isn't a power dream, really? Recall also that America isn't the only country in the world. Remote monitoring and override of cars will be used by secret police and evil bastards in nearly every nation on the planet to nail journalists and malcontents. Damn, this is gonna be a well-behaved world when they are done.

Re:Are you trolling or just boring?

[Catbeller]

Skills don't matter when the car yanks the wheel away from your hands. That's what self-driving means.
Do you all REALLY WANT an app to drive your cars? Don't lie to yourselves.

Re:Are you trolling or just boring?

[Catbeller]

Planes have slammed people into the ceilings. Planes have crashed themselves. Self-driving cars, v .1, crash more often and cause more injuries per crash. Computers are toasters. They are not intelligent. Computers can't deal with chaotic situations. "Life finds a way" is another way of stating it; it is impossible to build an AI that does what humans can do. Driving isn't a video game. You can't program for chaotic systems. If you can, go get your Nobel.

Re:Are you trolling or just boring?

[aberglas]

Um, Rule 43589:-

If animal-like thing appears in front then
    If nothing is behind and road is safe then
        maximum brake
    Else if other car is far behind
        maximum brake but monitor closing distance of car and release brakes if too close
    Else if other car is close behind
        moderate brake and monitor other car's reaction
    Endif
Endif

The above is a simplification, but that type of reasoning is not very hard. Recognizing fluffy is an animal and not just a pot hole, now that is tricky.

And how many people when seeing fluffy carefully monitor the car behind them when applying brakes?!

Anthony

Re:Are you trolling or just boring?

[omnirealm]

One tweak to maximize safety in these cases:

If other car is close behind
        reduce speed in proportion to how close it is
Endif

That's the algorithm I go with when driving. It's saved my bacon a few times.

Re:Are you trolling or just boring?

[Your.Master]

I wouldn't be convinced that the first generation won't change lanes when safe to avoid obstacles. After all, that's something they have to do even in normal driving scenarios. Suppose there is a car stopped on the road up ahead after an accident. The car isn't going to stop and wait for hours for the obstruction to clear, it's just going to change lanes. The computer is constantly taking in information from multiple vectors so decision-making time isn't a factor. If a sudden obstruction appears ahead, it seems like brake + lane change is going to happen.

Re:Are you trolling or just boring?

[kheldan]

Nice to see someone on here who is thinking straight.

If and when the day comes that we really, truly understand, top to bottom, how the cognitive processes of the human brain works, then we might be able to create an artificial intelligence that can match, if not surpass, us. Of course that's an entirely different can of worms to open, one which I won't bother covering here because science fiction authors have already covered the what ifs of that more than sufficiently. Of course the problem with trying to discover how the cognitive centers of the human brain work, is because you can't dismantle the system and examine it's components, then be able to put it back together and still have it function, unlike non-living machines. So far as I know we've still only got the vaguest ideas how a living brain works at all, let alone how to build the equivalent of it, so as such real AI is still quite a ways off.

Re:Are you trolling or just boring?

[Jack Griffin]

It's simply going to decelerate for the curve, and then accelerate again on the other side.

And this is the problem. A human accepts a certain amount of risk to get from A to B, but the AI can't, it has to slow down, making the journey unbearable.
While I scoot around the bend at 80km/h, the stupid little egg car with have to slow to 30, just in case.
Riding in one you may as well put a retard sign on your head, because that's what you will look like getting everywhere extremely slowly.

Re:Are you trolling or just boring?

[Jack Griffin]

They're going to see that stuff before it even enters the roadway,

What if it's already in the middle of the road? What if it isn't a cat after all, just a piece of cat-sized rubbish? Does it stop stationary in the middle of the road indefinitely?
Just for a laugh, can I bring the city's traffic to it's knees with a handful of small cardboard boxes strategically placed on certain roads?

Re:Are you trolling or just boring?

[drinkypoo]

Just for a laugh, can I bring the city's traffic to it's knees with a handful of small cardboard boxes strategically placed on certain roads?

No. If they're small enough to drive over, then the cars will just do that. If not, then they'll stop normal traffic, too. If people can't go around the box, they will have to get out and move it. Before you raise an objection to self-driving cars, be sure that it does not apply to normal cars.

Re:Are you trolling or just boring?

[bingoUV]

No, because rear-endings are caused by insufficient following distance.

And the human driver behind has exactly that - insufficient following distance.

They're going to see that stuff before it even enters the roadway, so they don't have to hit it to begin with. In situations where the sides of the roadway are obscured, the vehicles will simply reduce their speed â" the same thing a human driver should be doing.

In most cities that means driving under average speed of 10 mph at all times not just traffic peaks. Autonomous cars might make a nice ride for tourists, but anyone who has work to do will choose a human driven car then.

Remember - Top speed of dog - 50 mph especially if kicked. Angle of approach - unknown. Origin - possibly boot of the car in front with insufficient following distance. Because it cut between the autonomous car and the earlier car in front because autonomous car was driving too slow, keeping too much following distance and holding up traffic.

A self-driving car won't get pissed off when it gets held up by some cyclist who won't pull over to permit passing, as the law demands.

The occupant/owner/rider will. If he has anything important to do with his own life, he will sell this autonomous car at a low price and buy a car he can drive humanly.

Re:Are you trolling or just boring?

[GuB-42]

Sure. But is it going to slow down or come screeching to a halt when some dumb cat or dog runs out into the road, and cause the vehicle behind it to rear-end it

It is going to stop to a halt, which shouldn't be a problem for the car behind it if safety distances are observed. And if the car behind is an autonomous car or have some kind of assistance like some today's cars, it may be even less of a problem.

Re:Are you trolling or just boring?

[stoatwblr]

"But is it going to slow down or come screeching to a halt when some dumb cat or dog runs out into the road, and cause the vehicle behind it to rear-end it,"

That kind of thing would be a rare event if the tailgating driver was criminally charged - as he/she should be.

There's a reason for safe following distance laws and what you've described is an argument in favour of more AI in charge of cars, not less.

Re:Are you trolling or just boring?

[stoatwblr]

"What if it's already in the middle of the road? What if it isn't a cat after all, just a piece of cat-sized rubbish? Does it stop stationary in the middle of the road indefinitely?"

No, it changes lanes and goes around it.

You're also assuming the monkey can't get out of the car and kick the offending object to the kerb.

Re:Are you trolling or just boring?

[Jack Griffin]

What if the cardboard box is filled with concrete?
What happens when the car doesn't have a person it, or the person is mobility impaired?
The reason I raise these objections is because AI is so far behind actual intelligence, it seems ridiculous that otherwise smart people put so much faith in it. Do you know of any real examples where AI has proven itself? I can give you plenty where it has failed (eg Siri/Google voice search, Captcha, Google targeted search etc) The only time it works is in completely controlled environments (railways, elevators, mine sites). As soon as you add any unknowns it goes to shit. How will this be any different?

Re:Are you trolling or just boring?

[Jack Griffin]

No, it changes lanes and goes around it.

What if it's a single lane road?
What if the other lane has high density moving traffic and the AI deems it not safe to change lanes? Most meatbags will simply push in, but the AI won't be allowed to do that.
The reason this will fail, is the public road requires risk to operate, but AI won't be allowed to accept any.

You're also assuming the monkey can't get out of the car and kick the offending object to the kerb.

I'm assuming some monkeys won't be in a position to do so (too young, unsure, too old, disabled, too drunk). And do you really think a robot car will be allowed to instruct it's passenger to get out into the middle of busy traffic potentially endangering their lives?
Risk is why it can't work.

Re:Are you trolling or just boring?

[kheldan]

Do you all REALLY WANT an app to drive your cars? Don't lie to yourselves.

I'm not convinced anyone is in denial here. What it seems like, is every time I get into a 'discussion' on here about this subject, they're just not considering all the things that can go wrong with the technology, which probably stems from how complex a task it is, and that people have been doing it for so long that they take it for granted. Case-in-point: I remember seeing some documentary (On PBS, maybe? Nova?) about robotics in general. Some University was trying to create a robot that could do 'simple' household chores. Specifically, they were trying to teach it to fold shirts. It couldn't do it anything like reliably. Tiny little changes would completely throw it off. The human brain is an incredibly complex thing. Consider how long it's taken just to build a robot that can stand and walk on two legs without falling over at the slightest difference in the floor, or an obstacle put in it's way, or (gasp!) trying to climb stairs. Or, as I mentioned elsewhere, how we still can't get an 'AI' to pass the Turing Test. In the category of 'denial', though, there is that people can't or won't realize that when they step into an 'autonomous car', it's not really the computer that's driving them, it's the programmers that wrote the software that's driving them. That fact alone scares the hell out of me. Couple that with how many things we're discovering that automakers have covered up because they were too expensive to fix? Disaster waiting to happen, if you ask me. No thanks, I'll just keep driving myself.

Re:Are you trolling or just boring?

[drinkypoo]

What if the cardboard box is filled with concrete?

I didn't say "run over", I said "drive over". You fail at reading comprehension in your rush to make me wrong.

What happens when the car doesn't have a person it, or the person is mobility impaired?

In the first case, a driver will probably have to operate the vehicle remotely, to pilot it around the obstruction. In the second... the same thing that happens when the same person is driving a normal car. Again, you are raising objections that apply equally to normal cars. It's stupidly boring. Stop.

Re:Are you trolling or just boring?

[dave420]

The current automatic cars are doing a lot better than people, so I don't know what your point is. I also suspect you don't know what your point is either.

Re:Are you trolling or just boring?

[dave420]

The automated cars are already better than people, and they're relatively new technology. You don't really have an argument. That's not to say you can't find a good argument for your side, just that you've not found it yet.

Re:Are you trolling or just boring?

[david_thornley]

The car companies won't be concerned because a rear-ending is almost always the fault of the car in back, and they'll have no liability. Having been in a couple of rear-end collisions personally, I wasn't happy about my back being in pain for a week and not fully functional for months.

Re:Are you trolling or just boring?

[Jack Griffin]

I didn't say "run over", I said "drive over". You fail at reading comprehension in your rush to make me wrong.

Er, the cardboard box will be taller than the clearance of the vehicle. Your only choice will be to run over, or go around (did I really have to explain that?)

In the first case, a driver will probably have to operate the vehicle remotely, to pilot it around the obstruction..

What if the person is drunk? ie one of the big selling points of robot cars.

Again, you are raising objections that apply equally to normal cars. It's stupidly boring. Stop

I have just raised examples that do not apply to normal cars, but you fail to address that. But sure if calling names makes you feel better about avoiding the point, more power to you.

Re:Are you trolling or just boring?

[Jack Griffin]

The automated cars are already better than people,

I say otherwise, and eagerly await evidence of your bold claim.

Re:That's nothing

[belthize]

Quick follow up because editing isn't an option. I sort of jumped the gun and replied to a statement you weren't making.

I agree that the day probably will happen and agree that there will be outrage and hand wringing. My post above was actually picking a side in that discussion and not a response to whether there would be such a discussion.

Re:That's nothing

[Rosco P. Coltrane]

The issue is, human drivers have a strong instinct of self-preservation. Someone who has to decide between the parade and the tree in a split second will probably avoid the tree out of sheer instinct.

Now then, you might think the cool-headed computerized car will make the right decision and kill its occupant. But I can just imagine the following court case: "Your honor, my father's car killed him wilfully. I therefore sue Toyota/BMW/Honda/Google for murder, and for 100 kajillion dollars in damage".

One such court case - especially in the US - will do enormous damage to the entire industry, and might kill it off entirely. And no, the argument that autonomous car create fewer accidents overall won't fly, because somebody's property isn't supposed to kill its owner on purpose. You can bet emotions will run high, and emotions aren't good for rational debates.

Not to mention of course, people will have second thoughts about buying a vehicle that they know can decide to put them in danger for the greater good.

"It has to be perfect before it'll work"

[paradxum]

"A truly intelligent self-driving car needs artificial intelligence that can figure out where it is even if it has no map or GPS, and manage to navigate highways and follow routes even if there are diversions or changing in lane markings, he said." - from tfa

Frakly this is BS... I drive a large portion of my day for work (not a trucker, IT guy going to clients.) I run into "diversions or chaning in lane markings" and have to stop and think about what to do at times too! Why should an AI have to understand the intentions of a road worker/civil engineer better than we do before it can be accepted as intelligent?

" that can figure out where it is even if it has no map or GPS" ... OK, I'm going to drop you off in the middle of Kentucky mountain area with no GPS and no map, leave you stranded with noone to talk to and you should just magically know where you are.... sorry but NO. Unless I had been there before (i.e. prior knowledge or.... mapping) I will have no clue where I am and will have to basically start driving in one direction (which these cars can do) until I figure out where I am.

I know the media hype's this up, but he's going the other way and just being all doom and gloom.

Re:"It has to be perfect before it'll work"

[Gaygirlie]

Frakly this is BS... I drive a large portion of my day for work (not a trucker, IT guy going to clients.) I run into "diversions or chaning in lane markings" and have to stop and think about what to do at times too! Why should an AI have to understand the intentions of a road worker/civil engineer better than we do before it can be accepted as intelligent?

I'm not seeing anyone in TFA saying it'd have to be better at it than us, just that it'd have to be able to do it at all would be a good start. As things stand autonomous cars are not anywhere near of being capable of doing that on their own.

I will have no clue where I am and will have to basically start driving in one direction (which these cars can do) until I figure out where I am.

No, they can't. That's the whole point here: as long as they rely on GPS and very detailed mappings for navigation they won't be able to do that -- they need to know where they are to be able to start driving at all. The author wasn't saying the car should be able to magically instantly know where it is even when no mappings or GPS was available, just that the car should still be able to try and figure it out -- quite possibly doing the exact thing you suggested and trying to find a roadsign or two. The issue here is that these cars won't know even how to get off the god damn parking lot without GPS and mappings, let alone going out and figuring their own surroundings on their own without some very extensive AI.

Re:"It has to be perfect before it'll work"

[AthanasiusKircher]

Frakly this is BS... I drive a large portion of my day for work (not a trucker, IT guy going to clients.) I run into "diversions or chaning in lane markings" and have to stop and think about what to do at times too! Why should an AI have to understand the intentions of a road worker/civil engineer better than we do before it can be accepted as intelligent?

As long as it is feasible and SAFE for it "to stop and think about what to do" in these situations, that's fine. When you're on a highway traveling in a pack of bumper-to-bumper traffic at 60mph+ between concrete barriers on both sides in a construction zone and the lane changes and signs come suddenly, I don't think just stopping in the middle of the road seems like a good idea.

Almost every time I travel any significant distance on highways, I end up driving through such construction zones (including impromptu ones put up with cones) at some point, sometimes where things can get very confusing. A self-driving car needs to be confident enough to get through such scenarios safely and without messing up the rest of the traffic around it. (Or, at a bare minimum, it needs to know enough that will definitely warn and hand off driving to a human at such a point with sufficient warning that it's reasonable.)

" that can figure out where it is even if it has no map or GPS" ... OK, I'm going to drop you off in the middle of Kentucky mountain area with no GPS and no map, leave you stranded with noone to talk to and you should just magically know where you are.... sorry but NO. Unless I had been there before (i.e. prior knowledge or.... mapping) I will have no clue where I am and will have to basically start driving in one direction (which these cars can do) until I figure out where I am.

I don't think that's what TFA is talking about. The issue is that many self-driving cars depend on mapping not only for routes and location, but also for knowing the details of the roads they are on, what's coming up, etc. to assist them in driving safely. For the same reason that self-driving cars find it hard to adapt to sudden construction signs and deviations, it's generally much harder for them to deal with basic driving on an unknown road.

The equivalent for a human driver would be something like if you got lost on a road, your driving skills suddenly reverted to the level of a teenage driver on a learning permit.

Re:"It has to be perfect before it'll work"

[phantomfive]

The issue here is that these cars won't know even how to get off the god damn parking lot without GPS and mappings

That's a good example

Re:"It has to be perfect before it'll work"

[RR]

Frakly this is BS... I drive a large portion of my day for work (not a trucker, IT guy going to clients.) I run into "diversions or chaning in lane markings" and have to stop and think about what to do at times too! Why should an AI have to understand the intentions of a road worker/civil engineer better than we do before it can be accepted as intelligent?

As long as it is feasible and SAFE for it "to stop and think about what to do" in these situations, that's fine. When you're on a highway traveling in a pack of bumper-to-bumper traffic at 60mph+ between concrete barriers on both sides in a construction zone and the lane changes and signs come suddenly, I don't think just stopping in the middle of the road seems like a good idea.

What are you talking about? If you are in a construction zone and the lane changes suddenly, I think you will find yourself in a traffic jam. For that matter, I don’t think you will be driving “60mph+” in a construction zone; I think you will already be in a traffic jam.

Re:"It has to be perfect before it'll work"

[vovin]

Regardless. If the cars in front know where to go ... the route is by definition mapped and the data shared into the pool of knowledge about the route change and how to handle it.
Done and Done.

Re:"It has to be perfect before it'll work"

[gnasher719]

" that can figure out where it is even if it has no map or GPS" ... OK, I'm going to drop you off in the middle of Kentucky mountain area with no GPS and no map, leave you stranded with noone to talk to and you should just magically know where you are.... sorry but NO. Unless I had been there before (i.e. prior knowledge or.... mapping) I will have no clue where I am and will have to basically start driving in one direction (which these cars can do) until I figure out where I am.

Note that a self-driven car will almost always have GPS and now its exact location. It will just sometimes not have an accurate map of the area directly around it. Both self driven car and car driven by me will proceed to the nearest road, then make a guess which direction to turn. The difference is that the self driven care will always know where it is and what direction it is going. It can't get lost. If it returns to a place where it was before it can take that into account.

Re:"It has to be perfect before it'll work"

[sl149q]

If you have used Google Maps / Waze / Apple Maps recently you will have noticed that they do a pretty good job of showing congestion for your route in real time.

For any obstruction on the highway, the FIRST car may have to figure something out, the SECOND car will simply have an updated "map" saying that there is an obstruction use the left lane and pay attention for a flagger.

Re:"It has to be perfect before it'll work"

[stoatwblr]

"When you're on a highway traveling in a pack of bumper-to-bumper traffic at 60mph+ between concrete barriers on both side"

Then you and every other idiot on the road is probably following too close and you all need to redo your Driver's Ed classes

In any case it's not just "Maps and GPS", it's also intertial nav - my 2003 Nissan has this as an integral part of the Navigation system to cope with urban canyons - and real time traffic updates broadcast over broadcast FM subcarriers that my aforementioned Nissan also uses to reroute around blockages.

12 years on from that technology you can add real time map updates (iridium, etc in addition to mobile data services) and feedback from other drivers on the road (waze and friends mean the data can be democratised and is less susceptable to rogue operators), although this is starting to be automated and drivers are merely adding commentary.

In another decade you can add direct V2V communications so that when there's a problem around the corner your car knows about it within seconds, not minutes. Whether that's a monkey or a machine at the wheel, it makes decisions easier when you know what 's ahead.

90% of drivers do it because they have to, not because they enjoy it. They will happily let the computers take over.

90%+ of crashes and dings are the fault of one or more drivers. Insurance companies know that and the most likely long-term trend will be _punitive_ insurance premiums if you insist on manually driving your own vehicle.

Re:"It has to be perfect before it'll work"

[dave420]

So much guesswork! So much vitriol! It looks like an argument, but it reads like a rant against how you imagine these cars to operate, with no thought given to them actually improving over time (as most technology seems to do).

Toyota getting left behind

[monkeyxpress]

Translation: Toyota is woefully behind in autonomous car development, and rather worried about it.

The FUD begins.

Re:Toyota getting left behind

[Chuq]

Yep. It's hard to take someone seriously when they are the executive technical adviser of a company that is pushing hydrogen cars over battery electric.

Re:Toyota getting left behind

[JustAnotherOldGuy]

Yep. It's hard to take someone seriously when they are the executive technical adviser of a company that is pushing hydrogen cars over battery electric.

Hydrogen cars? What could possib- *KABOOM* (debris and body parts rain down)

Re:Toyota getting left behind

[Anonymous Coward]

Yep. It's hard to take someone seriously when they are the executive technical adviser of a company that is pushing hydrogen cars over battery electric.

Hydrogen cars? What could possib- *KABOOM* (debris and body parts rain down)

Oh, the humanity!

Re:Toyota getting left behind

[RR]

Forget autonomous car development, Toyota is woefully behind in computer-controlled car development. Random relevant article: Toyota's killer firmware: Bad design and its consequences.

Re:Toyota getting left behind

[NotThisMind]

Still, one of the most reliable car companies, pick one feature.

Re:Toyota getting left behind

[stoatwblr]

Bear in mind that Toyota spent 30 years developing 60kW gas turbine vehicles before finally admitting they were impractical and finally giving up in the mid 1990s.

However the multiple-input continuously-variable gearbox, associated drivetrain and computer control technology were directly inherited by the 2nd generation Prius (the first gen was more closely related to railway diesel-electric tech than anything else)

The technological children of "failed" projects are oftimes more important than the fact of the failure.

Hydrogen tech is unlikely to ever be practical but materials technology driven by its development may well be a gamechanger in future.

Re:Toyota getting left behind

[JustAnotherOldGuy]

Hydrogen tech is unlikely to ever be practical but materials technology driven by its development may well be a gamechanger in future.

Sure, and that's why I keep hoping for flying cars. I doubt it'll ever happen but the spin-offs may turn out to have some practical benefit.

we automate routine tasks

[NostalgiaForInfinity]

The cars rely on all that data to drive, so they quickly hit problems in areas that haven't been mapped in advance.

I don't see that as a problem. If it works on most of the roads people drive every day, that's good enough. As with all automation, we automate routine tasks and let humans do the rest.

A truly intelligent self-driving car needs

But we don't need "truly intelligent self-driving cars" for self-driving cars to be very useful any more than we need "truly intelligent factory robots" for factory automation to be very useful.

I regularly drive a stretch of road that's just a few miles long, but between construction, accidents, poor marking, bicycles, and heavy traffic I'd be nervous about letting an AI system navigate.

Well, then don't. In fact, your AI driver would probably simply avoid that route altogether precisely for those reasons and still get you to your destination safely and efficiently. Nobody says that an automated driver needs to take the same route as you do; after all, bikes, motorcycles, buses and light rail probably don't either.

Re:we automate routine tasks

[Jack Griffin]

In fact, your AI driver would probably simply avoid that route altogether precisely for those reasons and still get you to your destination safely and efficiently.

It'll be safe, the law will require that, but don't for one moment think it will be efficient. Based on the amount of risk on your average road, each trip will slow as all hell, so take a LOT longer to get everywhere.

So they aren't as smart as they're supposed to be.

[hey!]

Well, neither are human drivers.

Re:That's nothing

[14erCleaner]

The real test of artificial intelligence will come when the self-driving vehicle will have to decide between plowing into a crowd of people to protect the driver, and smashing into a tree to protect the crowd of people - but killing the driver, when the accident is inevitable.

Yeah, I hate it when that happens (and it happens all the time). So far I've always chosen the crowd of people - why can't we just program the car to do the same?

Last quarter mile navigation

[sjbe]

The biggest problems aren't actually going to be point to point navigation or even obstacle avoidance, though those aren't trivial. Navigation when you know the destinations is a solved problem and we've got a pretty good idea how to handle obstacle avoidance and terrain following though there is progress to be made

Possibly the hardest problem to solve it you want completely autonomous cars will be navigation in the last quarter mile and for destinations where you aren't actually sure exactly where you are going. This is a human interface problem and those are always challenging. In those circumstances it is REALLY hard to instruct a computer efficiently without actually taking the controls yourself. For example how do you explain to the computer that you want the parking space 3 places over but you want to back in? Or that you don't want to block in the car so park next to it on the lawn? Sounds easy but it really isn't - not yet anyway. Humans can do it mostly competently but we don't have any computer that is anywhere close to human level processing of verbal commands. Stuff like parking lots will be surprisingly hard to automate in a way that will be pleasing to most people. There are solutions but they are going to take a long time and require a lot of infrastructure. Probably several decades away at minimum. Sort of how we had autopilot for planes many year before we had the ability to do autonomous takeoffs and landings. (and the aviation problem is arguably easier as it has fewer variables)

I think we will see semi-autonomous systems relatively soon particularly for stuff like highway driving. But I think there is going to remain driver controls for quite some time because steering into that parking space or instructing the car to back up to the front door is actually pretty hard to do well. What will happen is that you'll program in your destination, the car will take you close to where you want to go and then you'll probably drive the last little bit yourself in a lot of cases. I think this piece of navigation will be solved last if at all.

Re: Last quarter mile navigation

[uradu]

I've been thinking about that problem too. A good example would be unstructured situations, such as spill over event parking on lawns or gravel lots. Given existing cars cars already there certain pattern formation algorithms can be applied, such as continue this or that line of cars. Otherwise things get even trickier and some spacial user interface will have to let the passenger point out in an overhead view where exactly to place the car and in what orientation. None of these are unsolvable problems per se, but a lot of research is still ahead for sensible user interface and car interaction paradigms that can be elevated to standardized levels across makes so one can hop from one car to another and still operate it reliably. Perhaps even in completely autonomous vehicles such as the ones Google is experimenting with a fallback controller of some sort will always be needed.

Interesting times, for sure, and I'm glad I'm getting to live through them.

Re:Last quarter mile navigation

[Moof123]

The real killer for adoption is going to be the couple minutes you have to spend punching in a destination before you get going. We are so used to just getting in and going where we want that such delays will be very frustrating for short trips.

Worse will be the times where you know where you are headed, but don't really know the address or proper name such as that italian place downtown, you know the one with the good meatballs. Or the soccer field just past the railroad tracks. Judging by how awful some of the built in navigation systems are I don't think we can trust car companies not to foul up the one for an autonmous interface.

I see a lot of people tiring of such small delays and hassles that they just don't use it much of the time. The likely heavy adopters will be the wackos with long commutes who prefer to spend on tech instead of a closer address.

Re:Last quarter mile navigation

[Gavagai80]

The real killer for adoption is going to be the couple minutes you have to spend punching in a destination before you get going. We are so used to just getting in and going where we want that such delays will be very frustrating for short trips.

People already spend a couple minutes punching in a destination on their phone / gps to get directions. Add a "send to my car" button on the phone, problem solved. Frequent destinations can be saved as favorites and assigned a number or letter to start in one keypress.

Re:Last quarter mile navigation

[stoatwblr]

"but don't really know the address or proper name such as that italian place downtown, you know the one with the good meatballs"

'OK google. search restaurants, italian, downtown, meatballs

OK google Navigate to result N, go.'

and that's with existing technology.

Re:Last quarter mile navigation

[Cro Magnon]

The last time I used my phone to find a route, it gave me something which I didn't think looked right. So, I went the way I thought was right, and found that though the phone's directions would have gotten me there, my directions were more direct and quicker.

Oh noes

[JustAnotherOldGuy]

"Autonomous Cars Aren't As Smart as They're Cracked Up To Be"

Let me be the first to say, "No shit."

Re:That's nothing

[NostalgiaForInfinity]

The real test of artificial intelligence will come when the self-driving vehicle will have to decide between plowing into a crowd of people to protect the driver, and smashing into a tree to protect the crowd of people - but killing the driver, when the accident is inevitable. Computers just aren't good at all at that sort of thing.

In actual fact, computers are very good at making that kind of decision, and unlike humans, they can make it consistently and reliably, without panic, anger, or selfishness.

So, since you're human and presumably good at making these decisions, just share with the rest of us: what rule should the car follow? Personally, the rule I think it should follow is protect bystanders over protecting the life of the driver, unless the bystanders are doing something illegal. Of course, your preferences may differ.

I can't wait to see the legal and moral discussions that will ensue after the first such accident occur.

Well, and you should be thankful that we can actually, for the first time, have these discussions meaningfully, because until now, such discussions were pointless.

Re: That's nothing

[t1oracle]

The AI wouldn't drive into a tree in the first place. If you jump front of it, it may run you over, but that's your fault. We can't expect AI to make ridiculous calculations on who to kill. The AI's priority is it's occupants. People outside of the vehicle are responsible for themselves. It's exactly the same as when a human drives. When you and your kid jump in front of a car and die, your surviving family doesn't get to sue the driver for not choosing suicide instead.

Autonomous cars

[penguinoid]

No, autonomous automobiles.

Good Summary Until...

[Thunderf00t]

The columnist inserted their own concerns, concerns which read like FUD more than anything substantive.

Autonomous vehicles are pretty good when it comes to avoidance (accidents, heavy traffic, and bikers, for example) -- they often detect things that people may miss -- the problem they face is, well, exactly what the article noted: location awareness. Put an autonomous vehicle in a parking garage with a layout that is different than its heuristics have encountered, and take away precise mapping, and it may simply decide to park itself in the middle of the lane ;)

Re: Good Summary Until...

[uradu]

Yep, the unstructured garage situation is a much better example than the poorly marked highway with missing markings or potential conduction areas. The sudden lack of GPS and maps is highly contrived anyway (!), and these conditions can be coped with fairly easily by analyzing traffic if other cars are around, staying on the right half of unmarked roads (dynamically calculating virtual lanes, something the software does anyway), recognizing obstacles and objects in motion around construction sites. These are all highly predictable situation that engineers can more or less easily integrate into their models, so to raise them as top level concerns seems very naive, especially for a supposed expert in the field.

Re: Good Summary Until...

[sl149q]

People who own parking garages own them so that people can park in them. It follows that if any substantial portion of the potential market for your parking structure cannot use it because of X (in this case because the mapping company for the auto-car industry has not mapped your garage) then you will solve for X (in this case get the mapping company in.)

In other words... there will be new businesses that are created to solve these problems. You need to repaint the lines in your parking lot because the municipality says you need to change things... most likely the paint company will also have a contact at the mapping company and arrange to have them come out and remap it before the paint dries.

Did the TFA also mention the imaging lasers?

[Nutria]

How much fog/rain/snow/smoke does it take to degrade the sensing level?

Re:Did the TFA also mention the imaging lasers?

[jabuzz]

Thing is an autonomous car is not limited to the narrow visible light range of the EM spectrum that we humans are. Just because you can't see in fog does not mean the autonomous car is unable to see.

Not that I have any special knowledge of the sensors being used by Google et. al. but if they are restricting themselves to the visible spectrum then they are pretty stupid. At a bare minimum I would be using infra-red as well, probably thermal and UV for good measure.

I would also expect them to be listening to the environment as well, as it is essential for knowing there is an emergency vehicle in your vicinity.

Re:That's nothing

[tompaulco]

The real test of artificial intelligence will come when the self-driving vehicle will have to decide between plowing into a crowd of people to protect the driver, and smashing into a tree to protect the crowd of people - but killing the driver, when the accident is inevitable.

I don't believe AI now or in the near future is capable of the sort of recognition needed to determine to make such a determination. It would probably pick the tree because the tree is a single object of unknown mass and density which is moving slightly (due to wind, but the AI doesn't know that) whereas the crowd of people is a bunch of objects of unknown mass and density that are sitting still, moving slightly, or moving a lot. AI also is not capable of determining the damage that may be caused to the tree or the crowd of humans, nor of making the philosophical decision of what the damage cost is in moral terms.

Snow, ice, etc.

[Anonymous Coward]

When there's snow and/or ice on the roads, the lane markings might as well be missing. Same for the edge markings, ditches beside road, pathways beside road, etc. For pathways beside the road, we're not so concerned with pedestrians as with the several inch step up. A snowstorm or blizzard is possibly worse.

Human drivers can manage quite well under these circumstances (provided they slow down a little),

P.S. Captcha was "bloodied", appropriately enough.

Re:Snow, ice, etc.

[unencode200x]

This. I was driving someone's new Ford Explorer and turned on the lane keep assist or whatever it was called. It would keep us in the lane but it did so by sort of jerking back and forth. I thought to myself that if there was ice on the roads we'd be screwed.

Re:Snow, ice, etc.

[Bing Tsher E]

Added as a problem to the issue of hazardous road conditions is the fact that when said conditions occur, if there is a manual override on the autonomous vehicle, the driver will be less experienced in their general driving skills because they've not been driving. So in bad weather conditions, more responsibility to manually drive safely will be dumped on drivers with less experience in doing so.

Re:Snow, ice, etc.

[Mattcelt]

I think this highlights the basic problem: roads (including hazard warning design and all markers and visual elements) are designed with human consumers in mind. An AI is, by its nature, playing a losing game by trying to translate non-native (i.e., human) elements into machine language and adaptation.

What will happen as autonomous vehicles become more ubiquitous is inclusion of machine-consumable elements into road design. Wireless lane markers, inter-vehicle (mesh) information sharing, and other technologies will be incorporated, making the 'I' in 'AI' a lot more unnecessary in that sense.

Re: Snow, ice, etc.

[pfg23]

More vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) technology will evolve.

It's what I've been saying all along...

[QuietLagoon]

... Autonomous cars might look great in controlled tests or on pristine highways, "but soon fail when faced with tasks that human drivers find simple." ...

I want to see those so-called self-driving cars navigate a New England winter, or the pothole-filled roads that occur after said New England winter.

Re:It's what I've been saying all along...

[sl149q]

So you live in the 20% of the 80:20 rule. We'll solve for the 80% for great benefit. And let the 20% hang out to flap in the wind. It simply doesn't matter that we can't drive there. We probably don't want to drive there. And you can be like the Amish driving around in their horse drawn carriages.

Re:That's nothing

[monkeyxpress]

Agree. People keep bringing up this scenario, but it really is very unrealistic. If you are travelling at motorway speeds, then why is there a crowd of people near the motorway? If you are travelling at urban speeds (50kph) then a car can come to a stop on dry tarmac within 15m. At 30kph (the actual speed in many busy urban areas) the stopping distance is only 5m. Most of the stopping distance you normally have to leave is due to the really rubbish reaction time of humans (> 1 second). If the Google car's 3D scanner can detect the wall of people at 15m out, then it can just slam on the brakes and come to a stop. I would imagine the Google car can see a bit further than 15m which means it can either go faster, or not have to use full emergency braking so you can avoid hurtling into the crowd in comfort as well.

You say it like it's a bad thing

[Kjella]

Fact is, roads and road markings aren't supposed to just pop up out of nowhere. Here in Norway every public road (and many private roads, pedestrain/bike roads, forest roads closed for general traffic and soon) is mapped out in NVDB (Norwegian Road Database), and it's supposed to be authoritative guide on speed limits, road signs, pedestrian crossings, speed bumps, bridges, tunnels, road classification including lane types and weight restrictions, railing and so on. This is all public data, I'm looking at it right now. From the looks of it, everything that needs a permit could be in this system though I'm not sure if it is today so any planned detour, lane closure etc. could be mapped out here. And they're in GPS coordinates good enough to fit on a map, they don't have the structural details in this view but they do say there's more for entrepreneurs doing work. Let's face it, road marking sucks but often people drive like they're there anyway. Like right now on the way to work they've laid new asphalt in a roundabout, there's no pedestrian signs just zebra stripes in the road, well now there's not but people walk like they're there and people drive like they're there.

I don't see this as a problem, I see it more like a double validation. You're going to have cameras scanning the road, if they don't find what's expected according to the database and/or past mapping they're going to assume the safe path like the sign says 55 mph, the database says 45 mph let's drive in 45 mph and send an "uh oh what's wrong here" message to headquarters. Maybe you have to have a human driver or a "safe mode" drive the first time but honestly I don't think that is a problem as long as the camera is in the production model. I'd be more than happy to volunteer driving to the places I want to go, upload the footage and let a Google engineer play "spot the signs and road markings" so it can drive itself next time. That would get crowdsourced real quick.

Re:That's nothing

[tmosley]

Actually, it would just hit the breaks. Or, you know, swerve the other way.

Computers can hit the breaks a lot faster than a human, and better too.

In other news, Toyota researcher finds...

[BenJeremy]

Toyota researcher finds autonomous driving technology is hard to do, beyond the autonomous accelerator pedal.

On a side note, this stuff has been worked on for ages. I worked with a company in 2000, doing image recognition for lane departure warning systems and other subsystems that are currently in use today. The technology is there, but not all companies are happy that many of those technologies are tied to patents and would rather be able to use in-house sources. Developing those sources now is a bit late in the game.

In 5 years, comments like Pratt's will be completely laughable. The only reason he's taken remotely serious now is because it isn't ubiquitous yet. Consumers do not have serious experience with autonomous driving, so his FUD is accepted at face value. In reality, he's just faced with a tremendous uphill battle to catch his company up in the game, and it's overwhelmed Toyota, to the point they are sowing caution to the masses, mostly in the hopes to catch a breather in the court of public opinion.

This was always completely obvious

[DrXym]

Every day driving is filled with problems that would be intractable for a computer. Anyone who wasn't drinking the self drive koolaid would realise this in a second.

The best chance self drive has is on closed loops, e.g. airport terminal transfers where vehicles can drive separately than the other traffic in mostly predictable conditions. Even there there'll probably be some guy in a booth whose job it is to takeover if the car gets stuck, confused or breaks down.

On the public roads it would be better for vehicles to focus on advanced driver assistance - smart cruise control, emergency collision braking, hazard avoidance, lane tracking and niceties such as parallel park etc.

Re:This was always completely obvious

[BoberFett]

Everyday driving is filled with problems that intractable for humans. Computers will be better at some things, people at others. How about using the strengths of both?

Re:This was always completely obvious

[DrXym]

No, every day driving isn't filled with problems intractable for humans. Quite the opposite. The things which are hard for humans are reacting in time to apply brakes, anti-skid maneuvers and the usual foibles of tailgating, driving too fast etc. Which is why I said advanced driver assist is a better route to go and which does use the strengths of both human and computer.

Pessimist

[monkeyxpress]

There seem to be two camps of people. Those that think we will be living on mars and have fully autonomous cars in a couple of years, and those that actually look into it and see how hard it is going to be. For some reason, the media seems to prefer the first one. Reality prefers the second one.

I'm guessing you're either a taxi driver or a martian. Reality prefers the first one, but boy would life be more exciting if it was the second.

Three anecdotes.

[vtcodger]

Just a few of the many things I've encountered in 60 years of driving that are going to be a problem for computers.

1. GPS? My wife and I bought a new GPS on sale at a local mall a few years ago. First thing we did when we got in the car was to program the thing to take us home. We hit GO. It thought a while and then told us that home was 2700 odd miles away and that the trip might take a while. Guess what? GPSen don't work in parking garages. It apparently thought it was still in Sunnyvale where last it was turned off, and it was contemplating a trip across the continent.

2. A couple of days ago I was using that same GPS to navigate through a rural area in Vermont. Seeking the shortest route, it put me on a (dirt) road that ran about a half mile, turned a corner, and ended in someone's barn. Care to try your hand at a program to recognize and deal with that situation?

3. Many years ago while traveling up the (dirt) road to an obscure National Monument out West, I came around a corner and found myself in a large herd of sheep. Couldn't see the road. Or the ditches. Or anything but sheep. What now Kit?

Not that cars a few decades from now won't be able to deal with thousands of situations like that. But it'll take a while I think.

Re:Three anecdotes.

[Catbeller]

They don't respond well to "road closed" signs, either. GPS sort of just insists you drive through.

Re:Three anecdotes.

[jrumney]

Seeking the shortest route, it put me on a (dirt) road that ran about a half mile, turned a corner, and ended in someone's barn.

This is why the default setting for almost all GPS Navigation devices is quickest, not shortest route.

Re:Three anecdotes.

[Jack Griffin]

A few decades? It'll take three seconds.

1. GPS not working. Connect to cell towers, detects available wi-fi networks, talks to your phone, checks a glyph printed on the wall of the parking garage, and talks to other cars to determine its location.

Cool, becasue that always works too. If there's one thing we know about technology, it never ever fails....

Re:Three anecdotes.

[stoatwblr]

Bullshit. My car's nav system has a specific "road ahead closed" function that's one button on the dashboard.

It's 12 years old and the tech has only improved since then.

Re:Three anecdotes.

[dave420]

That's your argument - technology fails, so we shouldn't/can't ever use it? Brilliant work, sparky!

Re:Three anecdotes.

[Jack Griffin]

That's your argument - technology fails, so we shouldn't/can't ever use it? Brilliant work, sparky!

Er no. Logic isn't your strong point is it?

Answer

[Anonymous Coward]

So nobody is answering the last question in the post - about when would you want to take over. I say never. The whole point of a self driving car is so that I don't have to watch the road, don't have to pay attention, and can sleep, read, or work while the device drives me like a taxi would. If it can't do that, it is less than worthless. Less than worthless because why would I want to pay for a fancy AI in my car if I have to keep my hands free, near the wheel, and my eyes on the road? I do that NOW. If I have to pay attention and be ready to take over then the thing is just a toy and they can keep it. Once it can actually drive me without using my eyes, ears, hands, and brain as a backup then I might want one.

Whats this guys definiton of real world?

[Dereck1701]

"Autonomous cars might look great in controlled tests or on pristine highways"

Hasn't Google been testing out their cars in the real world? And if the wiki article is right they've driven over a million miles and only had 14 minor traffic accidents, none of which were the fault of the autonomous system (at least according to Google). If that is true and if my math is correct that puts their accidents per mile ratio at about 1 / 71,400. Again if my math is correct your average human vehicle experiences accidents at a rate of 1 / 66,700. Suggesting Googles autonomous vehicle is safer. Admittedly there are probably limitations, letting one drive in torrential rain or snow/ice covered roads may result in far less advantageous statistics, the roads do have to be pre-mapped and there are almost certainly situations they can't handle. But most of those situations go for any vehicle/driver, I've driven in a variety of terrible weather and I've never been in an accident that was my fault, I have siblings who have been in a half dozen accidents most of which were in good weather. Most humans generally do well when encountering road work areas, I've seen others driving in oncoming lanes because they failed to notice the gigantic signs pointing them somewhere else. Some people are going to be safer drivers than these autonomous vehicles, some people should be encouraged to let the vehicle drive instead.

Re:Whats this guys definiton of real world?

[Kjella]

Hasn't Google been testing out their cars in the real world? And if the wiki article is right they've driven over a million miles and only had 14 minor traffic accidents, none of which were the fault of the autonomous system (at least according to Google). If that is true and if my math is correct that puts their accidents per mile ratio at about 1 / 71,400. Again if my math is correct your average human vehicle experiences accidents at a rate of 1 / 66,700. Suggesting Googles autonomous vehicle is safer.

A much more interesting statistic that Google hasn't released is how often their engineers have either intervened or pre-emptively taken over control in order to avoid a potentially bad situation. Not to oversimplify what Google has done, but if I think of my commute to work and map it out to a computer with relatively simple rules like here's the lanes, there's an intersection and there's the light, there's a crossing that you don't pass until it's clear I would say at least 95/100 times it'd get by on very simple rules of the road. Of course since I make my commute ~200*2 = 400 times a year, I'd crash twenty times a year with that accuracy. So I don't really trust the car to handle all sorts of weird until they've told me that we kept the hands off the wheel and let it handle all sorts of weird, even if that looked like a really bad idea.

Re:Whats this guys definiton of real world?

[Catbeller]

A university study was just published last week that self-driving cars crash about 4-8 times more often than humans, and the crashes cause more serious injuries.
The publishers were at pains to point out that only a few million miles were on record, so they hedged the obvious conclusion that humans drive better. Utter nonsense. People REALLY want to believe that robots drive better. A million miles or hundred million, the numbers are in. Self-driving cars aren't working. A million miles are more than enough data.

The real thrust of getting this onto the road is to fire truck drivers. End of story - they want to keep all da money. Uber outright admits it wants to fire the Uber "contractors" and replace them with robots. This is about money, bub. Hundreds of billions of fat, juicy dollars for businesses, and armies of newly unemployed. They hate paying people money to do jobs when they can keep it all.

Re:Whats this guys definiton of real world?

[Jack Griffin]

Hasn't Google been testing out their cars in the real world?

Only once you accept that their version of "Real World" is a lot different yours and mine.
Tests are heavily controlled, in extremely well known and mapped out areas. I have heard of no real, 'real-world' test that involved giving a car to a novice to take whereever they wanted indefintely. Until I see that, it's not really 'Real-world' is it?

Re:That's nothing

[leftover]

I think you are spot-on the crux of the issue. This will determine the future of fully-autonomous vehicles. Even if there are legislative/regulatory decisions in place before the first fatality, the ambulance-chasers will do their best/worst to get around them.

Privately owned vehicles can be covered by existing case law -- if it's yours then you are responsible even if your were sleeping | drunk | makin' whoopie.

Driverless taxis or their taxpaying equivalents are the new issue. One solution would be a broad legislative indemnification for decisions made by approved software. [ Definition of "approved software" is left as an exercise for the reader.]

Since both the automobile and ambulance-chaser lobbyists will oppose this with every fiber of their wallets, I don't see any such legislative action happening anytime soon. Without it, I don't think the autonomous taxi will happen at meaningful size.

Re:That's nothing

[Dutch Gun]

decide between plowing into a crowd of people to protect the driver, and smashing into a tree to protect the crowd of people

People hold up ridiculous scenarios like this as some sort of hypothetical metric, but how well would a human do with an insane choice like this, presumably with only a split second to make the decision? Not very well, I'd imagine. Don't put AI up against ridiculous situations. Put them up against realistic obstacles, which we might actually have a chance of seeing in our lifetimes. Road construction. Temporary obstacles with police directing traffic. Blizzards. Temporarily flooded road. Parking lots or garages.

There's also this false dicotomy presented, wherein some people seem to think that unless an AI can can handle ALL situations possible, it can't possibly work. I'll tell you what will happen in many situations. The AI will come to a controlled stop and tell the human "Hey, I don't know what's happening. Take over the wheel, please." That seems perfectly reasonable for crazy scenarios that only rarely occur.

The answer to what would likely happen, by the way, is that the AI in the car would have long ago started braking, so as to avoid the problem in the first place.

Re:That's nothing

[unencode200x]

Agreed. But, just like the doctors, the car companies will find a way to bend the law their way and have limits on how much they can be sued for.

Re: That's nothing

[unencode200x]

Have you seen iRobot where the robot choses to save Will Smith (the police officer) instead of the little girl because the robot calculates the little girl's probability of survival was lower? I thought that was an interesting take on decisions like this.

Re:That's nothing

[Dutch Gun]

No, the correct decision of an car's AI is to always prioritize the life and safety of it's occupant. And you can bet that's what every vehicle will be programmed to do. People on the outside can take care of themselves.

Note that this doesn't mean speeding recklessly and then plowing into a crowd to save the driver. That only occurs because of previously made poor choices. Obstacles don't magically teleport in front of cars. It only appears that way to human drivers because we have a bad habit of not paying attention. Computers don't have that little flaw, and so will be braking the car before the human occupant even realizes there's a potential situation ahead.

Car manufacturers are not exactly strangers to litigation. The notion that any single court case will doom an industry is overstating things, I believe.

The point of self-driving cars....

[mark-t]

.... is not to literally drive the car, it is to prevent accidents.

Any autonomy that an automobile might appear to exhibit should be seen as a side effect of that goal, and not a direct manifestation of intent.

Stop saying "Artificial Intellgence"

[mbeckman]

The first step to an intelligent debate on autonomous cars is to eliminate the phrase "artificial intelligence" from the discussion. Autonomous cars are just that: cars that navigate roads without human intervention. They are not intelligent, artificially or otherwise, anymore then a 1940s autopilot in a Beechcraft D-18 is.

"Autonomous" is the perfect adjective, because these cars are automatons, not conscious, thinking beings. Because we have only the foggiest definition of "intelligence", we are in no position to create an artificial one. If someday we do have that knowledge, what will we call artificial intelligence when we actually make one? That'll be a problem if we sully the term today with myth and superstition.

Re:Stop saying "Artificial Intellgence"

[Wain13001]

Actually "artificial intelligence" is exactly the correct term to be using here. It is not real intelligence, it is ersatz. This pedantic argument has been brought up for years with a proper answer proposed a long time ago; That when a "real" AI is created, we refer to it as Machine Intelligence...since there will be nothing artificial about it.

Re:Stop saying "Artificial Intellgence"

[mbeckman]

This pedantic argument has been brought up for years with a proper answer proposed a long time ago; That when a "real" AI is created, we refer to it as Machine Intelligence...since there will be nothing artificial about it.

Really? I find no such terminology in the extant AI literature. The eponymous Machine Intelligence Reaearch Institute, major proponents of the AI super intelligence religion, appear to make no distinction.

Citation or retraction, please.

Re:That's nothing

[dissy]

Computers just aren't good at all at that sort of thing.

It's a simple trivial problem that doesn't need any AI to solve. One variable and a couple IF statements is all that is required.

Have the car reset a counter at midnight to zero. For each person the car runs over and kills, increment the counter by one.
Once that counter reaches 2999, shut down the engine and refuse to start until the next day.

If self driving cars limit themselves to less than 3000 people killed on the road per day, they will already be safer and kill less humans that our current situation with human driven cars, so the self-driving cars empirically win.

Re: That's nothing

[unencode200x]

Is that a world-wide number? US annual traffic deaths are around 30,000 a year.

Re:That's nothing

[Anonymous Coward]

I think you are spot-on the crux of the issue. This will determine the future of fully-autonomous vehicles. Even if there are legislative/regulatory decisions in place before the first fatality, the ambulance-chasers will do their best/worst to get around them.

This is clearly not a show-stopper.
Volvo says it will accept full liability for accidents involving its driverless cars, making it "one of the first" car companies to do so.
There will be some legal issues if ambulance-chasers start to abuse this, but they will be going up against a large car manufacturer with a legal department, not another driver without court experience.

Turning Test

[kheldan]

We can't even, after decades of trying, create an 'artificial intelligence' that can pass the Turing Test, and that's just text on a screen. What makes any of you so sure that 'autonomous cars' were ever so close to being a reality? Even then, as I've said in the past and will keep saying, there's always going to be a full set of manual controls, by law, and you'll always still be required, by law, to be fully educated, trained, tested, licensed, and insured in order to be behind the wheel of any vehicle, regardless of any so-called 'self-driving' feature it might have, because when all it said and done, when human safety and lives are at stake, a human being must be the final 'backup system'. Furthermore since we all know that any skill that isn't used often tends to atrophy, you'll likely be required to be re-tested by the government more often than you are right now, to ensure that you're still competent to be operating a motor vehicle. So get over it: You're still going to be driving yourselves around for a good long time to come, probably the rest of your lives, or at least until you're too old to be a competent operator of a motor vehicle.

Now, then, for all of you with all your complaints about 'other drivers' being so bad: Hush up already, you're probably at least as bad as the ones you're complaining about. That being said, what we need to do in this country is to improve driver training and education, and tighten up testing procedures and frequency to improve the overall competence of drivers on the roads, and exclude the ones who can't (or won't) show an acceptable and consistent level of competence. This should include tougher and longer-lasting penalties for individuals convicted of DUI. Furthermore any use of any kind of any mobile wireless device while driving should be strictly prohibited and punished severely; I think a six-month suspension of driving privilege with a hefty fine should be sufficient.

Meanwhile, auto industry, please do continue to develop and produce collision-avoidance systems that warn the driver when they're screwing up.

Re:Turning Test

[Mantrid42]

We can't even, after decades of trying, create an 'artificial intelligence' that can pass the Turing Test, and that's just text on a screen. What makes any of you so sure that 'autonomous cars' were ever so close to being a reality?

Because those are two wildly different problems?

Re:Turning Test

[kheldan]

No, see, they're not: If you can't even create software that can think enough like a human being to have a casual conversation with someone, over text on a screen, which is a relatively simple thing with relatively few variables in it, then how the hell do you expect something as complex and dynamic as operating a vehicle in the real world, where anything can happen at any time, completely at random? You're expecting us, in a few years, to come up with the equivalent of the human brain, that took millions of years to evolve to the state it's currently in. At best this is still science fiction. I find Google and it's toys to be laughable at best, and I'm obviously not the only one. Better keep your driving skills up to snuff, friend, you're going to need them for probably the rest of your life.

Re:Turning Test

[dave420]

What makes humans terrible drivers is our ability to think like a human. It's not a great gift when driving. The best drivers remove as much of the "human" in "human driver" as possible, and just stick to using the inputs provided by the car in order to move it around.

Your argument fell apart when you assumed that humans thinking is a good thing in cars at all. It's demonstrably not.

Re:Turning Test

[david_thornley]

Being able to have a conversation requires an immense amount of knowledge, easily retrievable, and the ability to make connections. It would have to do all of this in context. (One example, I think from Hofstadter: in a crowded and busy restaurant, a man comments that he's glad he's not a waitress here tonight. The correct response is to ignore the sex change, occupation change, etc., and focus on what's specifically hard about being a waitress.)

Being able to drive a car requires modeling of the immediate environment and making appropriate responses. That's far, far simpler.

Magic roundabout

[unencode200x]

I'm sure these cars will do great around the Magic Roundabout: http://basementgeographer.com/...

Re:Magic roundabout

[stoatwblr]

Humans don't do well around those things the first few times they encounter them. Some never do master them.

A human-driven taxi is better right now.

[fyngyrz]

The "AI" we're talking about presently and in the near future is mostly "A" and almost no "I" at all.

The kind of thing marketroids and the naive are calling AI today, including the tech that's beginning to show up in vehicles, is so one-dimensional in its "intelligence" as to be on about the same level as a toaster that "knows" not to burn my toast, or a chess playing program that can kick my ass at chess. The toaster "AI" couldn't control a robot vacuum cleaner, and the chess "AI" can't even play checkers, much less deal with anything further out of it's 1D "I" zone of competence, including not burn my toast.

In order for a vehicle to be able to "know where it is" and "know what to do about it", it will have to be more than one dimensional; it will have to be able to read signs, it will have to know destinations as things other than map references and paths other than mapped roads (parking lots, unmarked roads, etc.), it will have to make decisions based on extremely vague inputs and be able to do things such as ask for, and locate sources for, directions and understand them in pretty much whatever form they are provided. It will have to deal with the various situations that come up when the maps don't match the roads, too. Judging by my GPS, that's a lot more common than one might otherwise assume. It changes over time in random, unpredictable ways, too.

A general intelligence system designed for service (by which I mean to imply not conscious... otherwise we're talking about slavery, and we should know better than that by now) is not that close as yet. Frankly -- and I'm speaking with my AI researcher hat on now -- I think we'll get to a conscious general purpose intelligence well before we get to an unconscious one. We have a great deal of experience with imparting information to consciousnesses and we have considerable information available to us about what comprises one in our study of the human brain, whereas we have almost none about building a general purpose non-conscious intelligence, other than stacking multiple one-dimensional intelligences one upon another, which approach is approximately equivalent to solving the problem of multiplying by a million by adding one to an initial value of zero a million times. In other words, it'll eventually get the answer, but it's not in any way efficient.

As far as AI goes, all we really have right now is AI research, and various (not insignificant) benefits from the various tech insights and advances that fall out of that process. We don't have AI at all, at least not in the sense that is even slightly worthy of the term. The way AI is being used today, you'd want to be very careful telling your kid they were "intelligent", because they're likely to take away the idea that you think you just told them they're about as bright as the toaster. Not to mention the fact that when an actual AI is finally brought to light, we're not going to have anything useful left to call it. At that point, "AI" would be an insult. Not a great way to start a conversation with a new entity, IMHO.

The whole "it's AI!" meme reminds me strongly of the whole "3D TV" debacle. Again, marketroids and the ignorant built and propagated that appellation as a supposedly appropriate designation for fixed-viewpoint stereo vision, where fixed-viewpoint stereo vision is constrained, even by a relatively coarse and generous measure using whole-number degrees, to about 2 and 1/64800D or 2.000015432...D, whichever notation you prefer, leaving the viewer with something that in very few ways indeed resembles an actual 3D perception. When trying to describe actual 3D imaging, one is left with no accurate terminology. Unlike AI, we even actually have some low-performance versions of real 3D imaging now, so the linguistic problem is already on the roost, so to speak.

Sure, language evolves, that's a legitimate and real thing, but language also devolves, and that's what we're seeing in both these cases. I'm going with it, but I'm going kicking and screaming about the word-crap the marketroids are leaving on my lawn. Goddamn kids and their unleashed word-mutts. Where'd I leave my shotgun, anyway?

Re:A human-driven taxi is better right now.

[aberglas]

Things do not need to have human level intelligence to be able to work in the real world. And being able to read road signs is very doable today. The original post seemed to think that use a GPS is AI which it clearly is not, but being able to drive down most roads is not a particularly difficult task, even if it is much more difficult than playing chess, say.

Consider a wasp, with a nervous system smaller than a pin head, and all the complex behaviors it exhibits in the natural world. Driving a car is simpler than that.

Have a look at

http://www.computersthink.com/

for some ideas on this.

Re:A human-driven taxi is better right now.

[david_thornley]

Driving on regular roads should not require strong AI. People mark the things in mostly regular ways, that should be simple for an AI to pick up. Get off paved public roads, and it gets a lot harder.

Re:A human-driven taxi is better right now.

[fyngyrz]

Things do not need to have human level intelligence to be able to work in the real world.

No, they don't. That's why your text editor works. But your text editor isn't intelligent. Nor is your spreadsheet, the speech-to-text in your phone, a Tesla vehicle, or Google's search engine.

And being able to read road signs is very doable today.

Even if the ability to read all appropriate signage, no matter what kind, how positioned, or how occluded were a current capability -- and it most certainly is not -- it is still a decidedly one-dimensional application. It is not intelligence. Would such a thing be useful? Of course. But useful is not a synonym for "intelligence" or "intelligent actor."

but being able to drive down most roads is not a particularly difficult task

Again, you are incorrect. For any automated driving solution today, there exist larger sets of situations within which they cannot drive, than the sets of situations within which they can drive. Many depend on maps, and all depend upon road markings. These types of solutions do not provide competence even close to the level of a human driver. If the task was not "particularly difficult", these challenges would not be an issue. But they are, which fact falsifies your assertion.

Stripped of all the marketing hype and the glittering generalities, the technologies that have come out of AI research to date remain minimal, which is to say they incorporate competence in extremely limited domains of both types of information and relevant decision-making.

When technology produces an intelligent result, we'll know it -- it'll be outright obvious. It'll be able to learn to drive, play chess, make your toast correctly, and explain why you just tripped over your shoelace. Not because the information was programmed into it, but because it will be able to reason.

Without the general ability to reason, without induction, without the ability to learn from any evidence, what you have is not intelligence. All you have is a glorified if-then construct of some type.

I expect technology to (eventually) produce two types of intelligence: non-conscious intelligence, and conscious intelligence. I strongly suspect the former is going to be much harder than the latter, and will come much later. Time will tell.

Re:A human-driven taxi is better right now.

[fyngyrz]

However, people do not drive in regular ways, obstacles, errors, equipment failure and right of way damage occur in and around roadways in an almost infinitely variable number of ways, and markings are not always timely, legible (or even correct.) That's all true even of the "regular roads" you mention. As you indicate, away from those it becomes even more challenging.

btw, your sig is fabulous. :)

Re:A human-driven taxi is better right now.

[RyoShin]

Perhaps computer science should create an "indistinguishable from magic" term for programs/processes that are not truly intelligent (in that they learn and adapt with no further human tampering of the programming) but give the appearance of being intelligent in a narrow field and, as far as the common person is concerned, is "smart" (like Siri).

Simulated Intelligence?

Nope

[fyngyrz]

No, it wasn't:

...and it's insurance will be cheaper

That is equivalent to the following, just as you were told:

...and it is insurance will be cheaper

The correct usage for that sentence is:

...and its insurance will be cheaper

"its" is possessive. "it's" means "it is."

The mistake is usually made (and I make it as well, though I certainly know better) because in English, the general rule is that the apostrophe followed by "s" indicates possessive; but English is also riddled with exceptions. The "it's" / "its" issue is one of those exceptions. This particular one comes about because English also uses the apostrophe to create single words out of several:

o you will = you'll
o I have = I've
o she is = she's
o it is = it's

The last one simply takes priority over the usage where the apostrophe indicates possession. Instead, it indicates the compounding of the two words. So while it is an exception to the possessive usage, it is fully in line with the compounding usage.

We now return you to non-pedantic mode. :)

Re:Nope

[Maxo-Texas]

The important thing is not to loose your cool when playing fast and lose with grammar on the internet.

Re:Nope

[sjames]

The horse is dead.

Re:Nope

[Dog-Cow]

In English, the pronouns all have possessive forms that do not rely on apostrophes. Thus, the presence of an apostrophe in a word with a pronoun always indicates a contraction. It's actually rather simple, if one learns that simple rule.

Re:Nope

[fyngyrz]

The false assumption you're making is that everyone knows what a pronoun is. :)

Re:That's nothing

[dgatwood]

No, that's clearly not what the GP meant. The GP meant that the car would cut a corner, drive across the beach, and crash into the underground coral reefs. I would have thought that was obvious from context....

Re:That's nothing

[Hognoxious]

Someone who has to decide between the parade and the tree in a split second will probably avoid the tree out of sheer instinct.

More likely they'll try to miss the tree, but overcontrol and end up in a skid and hit anyway, sideways. Or clip it, spin, and hit the parade too.

But if you're in the situation where you have to make that decision haven't you already crossed the border into Fuckupland?

Re:That's nothing

[Anonymous Coward]

never drive faster than you can stop in the distance you can see - otherwise, that is driving blind and unsafe.

dee du dee du doo de doo, dum dee dum dee dum ...

[Hognoxious]

I don't want to live in a world where you can't 'drive for fun', and I know I'm far from being alone in that sentiment.

Do you have an uncle with a country place that no one knows about? Perhaps a former farm?

Re:That's nothing

[Mantrid42]

People keep bringing up this argument, but why would the AI be piloting the car at breakneck speed toward a crowd of people in the first place?

Re:That's nothing

[gnasher719]

The real test of artificial intelligence will come when the self-driving vehicle will have to decide between plowing into a crowd of people to protect the driver, and smashing into a tree to protect the crowd of people - but killing the driver, when the accident is inevitable.

The real test of driver intelligence and ethics will come when the driver of a human-driven vehicle will have to decide between plowing into a crowd of people to protect himsel, and smashing into a tree to protect the crowd of people but killing himself, when the accident is inevitable.

On the other hand, if you as the driver ever allow such a situation to arise, and you plow into the crowd, you'll hopefully go to jail for a very long time.

Re:That's nothing

[gnasher719]

The issue is, human drivers have a strong instinct of self-preservation. Someone who has to decide between the parade and the tree in a split second will probably avoid the tree out of sheer instinct.

Actually, the most likely thing that person would do is nothing (and the car goes wherever it goes). The second most likely thing is that the person avoids whatever is happening without action. If the car aims at people, that's the immediate danger and the driver will try to avoid that without thinking about the secondary danger of killing himself. If the car aims at the tree, that's the immediate danger and the driver will try to avoid that without thinking about the secondary danger of killing many others.

Re:That's nothing

[gnasher719]

Agree. People keep bringing up this scenario, but it really is very unrealistic. If you are travelling at motorway speeds, then why is there a crowd of people near the motorway?

I was told by British police that four percent of road deaths happen on the motorway (which makes it the safest place to drive by far). Of these four percent, 20% are pedestrians. Which makes the motorway an awfully dangerous place for pedestrians.

Re:That's nothing

[Computershack]

I don't see the issue.

For many scenarios where a human will have to fave that decision the autonomous car never will because it would have chosen option C, avoid situation long before it became an issue.

Autonomous cars are subject to the same laws of physics as any other car. If something steps out in front of it at a distance too close to be able to stop in time then depending on what other traffic is on the road and what is around the vehicle its choice is going to be hit that thing that has just stepped out in front of it or hit something else. As the occupant of that car I want it to be the option that does the least harm to me. Problem is that the AI may not share the same concern for me as I do.

Re:That's nothing

[Computershack]

Actually, it would just hit the breaks.

What if the braking distance is too long?

Or, you know, swerve the other way.

Into oncoming traffic?

Computers can hit the breaks a lot faster than a human, and better too.

That isn't my experience of the system fitted to my 44 tonne truck. The times its chosen to hit the brakes because it "thinks" there's going to be a collision even though there isn't, if I'd been carrying any load other than the one I was at the time or being empty, there would've been a serious accident.

Re:Whats this guys definiton of real world

[Moof123]

Part of Google's testing involved lending out Beta cars to employees. Sure enough, they whipped out laptops, went to sleep, and otherwise were in no position to take over if HAL gave up and handed over control. These were well educated folks who knew they were in Beta cars (and who should have been fired for such negligence). So as far as the general public goes we can expect zero backup for the system from the human inside. So the system needs to be truly autonmous in every sense before it gets released to the public.

Re: That's nothing

[Catbeller]

That happens because we drive cars. Cars are very stupid, expensive and dangerous solutions to the simple problem of moving people around. They've killed and crippled more people than wars. That's because moving tanks at high speeds on flat open strips results in accidents. That will not change. Want safety? Build trains. Get rid of the cars.

What's next,

[thinkwaitfast]

junk food isn't to blame for America's obesity epidemic ?

Re:That's nothing

[aberglas]

+1. Except that the human need not actually be in the car, but may be in a third world transport control centre.

For more ideas on this see

http://www.computersthink.com/

Re:That's nothing

[BradleyUffner]

I don't see the issue.

For many scenarios where a human will have to fave that decision the autonomous car never will because it would have chosen option C, avoid situation long before it became an issue.

Autonomous cars are subject to the same laws of physics as any other car. If something steps out in front of it at a distance too close to be able to stop in time then depending on what other traffic is on the road and what is around the vehicle its choice is going to be hit that thing that has just stepped out in front of it or hit something else. As the occupant of that car I want it to be the option that does the least harm to me. Problem is that the AI may not share the same concern for me as I do.

Unless you live in Middle Earth, trees "suddenly stepping out at a distance too close to be able to stop in time" isn't really a problem that autonomous cars are going to have to deal with.

Re: That's nothing

[Jack Griffin]

The AI wouldn't drive into a tree in the first place.

Where I live, we have trees literally a couple of metres from roads that are zoned up to 80km/h. Should (when) something go(es) wrong, the car can be propelled towards one of those trees without the means to stop itself.

Re:That's nothing

[Jack Griffin]

...the AI in the car would have long ago started braking, so as to avoid the problem in the first place.

So it sounds like a typical journey in these things will be extremely slow with cautious braking all the time. Gee you've got me excited....

Just take the wheel

[sjbe]

How about a touchscreen interface showing the area around you (based on sensor data), which allows you to drag a representation of the car to the place you want it to go?

That's basically the same thing as handing someone the controls to the car. Plus it really would be hard to do well, presuming it's even possible at all. Might as well drive at that point. It would be a LOT quicker and less annoying. The car's sensors can keep you from hitting things.

Black ice

[pfg23]

Black ice on a highway curve at night and if a sink hole suddenly emerges in my lane.

Things Moving Smart Cars Can Do

[LifesABeach]

1. They can veer right.
2. They can veer left.
3. They can stop.
Simple, the A.I. can then choose which will cost the Insurance company less.

Surrounded

[pfg23]

When a bunch of car jackers surround my vehicle.

Skynet

[pfg23]

All autonomous vehicles should constantly upload and retrieve information from a publicity owned database in the cloud, absolutely secured of course, so all vehicles have up-to-the minute awareness of road conditions and status. Let the chaos begin?

Re:That's nothing

[stoatwblr]

In most parts of the world the law is that you must be able to stop in the distance of clearly visible road in front of you, or half that if there is no centreline (narrow roads).

Speed limits are secondary to that - they're the velocity you must not exceed (most are arbitrary) - and you _can_ be charged with speeding if you're driving too fast for the conditions but below the posted limits, or dangerous driving if you're being completely stupid and doing things like driving 60mph in fog with 20 foot visibility.

The vast majority of road crashes are completely avoidable, which is why they shouldn't be called accidents.

Most crashes happen because cars know the laws of physics far better than the monkey at the controls.

An autonomous vehicle is a far more cautious driver than any human will be because if they're outside their parameters they'll either stop or go very slowly (overconfidence on the monkey's part is a big contributor) and they don't get impatient (which is a huge failing on the monkey's part and one which is inherent to their operating system), nor do they miss things like spotting the feet of the kid about to step out from between 2 parked cars 40 feet ahead (monkeys have terrible trouble coping with more than 2-3 inputs)

If the monkey is allowed to direct an autonomous vehicle but not given direct control, having hard limits imposed would reduce crash rates dramatically. The outcry about "loss of control" is the paradox, because the vehicle has _more_ control.

Re:That's nothing

[stoatwblr]

"If something steps out in front of it at a distance too close to be able to stop in time"

Then the driver hasn't been paying attention. Nothing "just steps out" and an observant driver will have seen the pedestrian heading for the roadway long before it sets foot on it.

An AI controlled car is far more likely to have been slowing down long before the human even noticed.

Re:That's nothing

[stoatwblr]

"Don't put AI up against ridiculous situations. Put them up against realistic obstacles, which we might actually have a chance of seeing in our lifetimes. Road construction. Temporary obstacles with police directing traffic. Blizzards. Temporarily flooded road. Parking lots or garages."

The most likely thing in such cases will be a button for "override, proceed - SLOWLY - until obstacle passed"

As for parking and parking lots, computers have already demonstrated they're far better at handling these than most monkeys.

Re: That's nothing

[pfg23]

Where you live, you're screwed no matter who's driving.

Re:That's nothing

[tmosley]

"What if the braking distance is too long?"

So, are these crowds of people just teleporting in front of the cars or what? You aren't likely to be running into people when you are going more than 30 mph, and you can stop within 10 meters or so. They would have to put a LOT of effort into hiding from the thing, and jump out at the last second.

"That isn't my experience of the system fitted to my 44 tonne truck"

I have no idea what you are talking about, but unless you JUST bought that truck within the last few months (when Mercedes started offering partial self driving functionality on big trucks), then you are talking about some other kind of system. And besides, we are talking about cars here, not big rigs. Cars are a lot easier to stop.

It's not about me, it's about me

[DriveDog]

The real question is not when I'd want to be in control of my own vehicle, the answer to that is anytime I think I should be. The real question is when I'd want the other guy to be in control of his car, and the answer is very infrequently.

Re:That's nothing

[david_thornley]

A few years back, we were driving down the street towards an intersection that (a) had a green light on, and (b) had had it on a significant period of time. There was a bus pulled off to the right, and of course we had no way of seeing through the bus.

A family stepped out from behind the bus, crossing against the light and appearing in front of traffic without warning. My wife did a panic stop, and avoided hitting them. (They looked disapprovingly at us, like we should always stop for a green light or something.)

Fortunately, we were not in the lane just to the left of the bus, but the lane to the left of that. Had we been in the lane next to the bus, there would have been no way to avoid the collision.

Of course, here we're back in the "sure the AI will run them over, but a human would under the same circumstances" situation.

Re:That's nothing

[stoatwblr]

"There was a bus pulled off to the right, and of course we had no way of seeing through the bus"

Defensive driving 101 - ALWAYS assume that a pedestrian will leap out from behind the object you can't see through.

Your wife was arguably driving too fast for the conditions and she was also likely to be concentrating on the green light (and speeding up, green lights have a tendency to cause that in humans) than on the possibility that someone might step out in front of her.

If you're driving at 30mph and are at the rear of a bus you're passing when someone steps out in front of it, _you_ as a human will run them over before you even get close to touching the brake pedal. An AI will react a lot faster and any defensive driver knows to be slowing down/foot already hovering on the brake pedal in such a situation because sooner or later it happens (in my home country, 100 children a year used to be killed in exactly such circumstances. It became a specific part of drivers' education and that figure is now a lot lower)

In other words: You're bringing up a situation caused by your wife's poor driving skills and claiming an autonomous vehicle wouldn't cope - it doesn't have to cope if it's avoided the situation by slowing down to a safe speed already.