Slashdot Mirror
Why Self-Driving Cars Should Never Be Fully Autonomous (roboticstrends.com)
An anonymous reader writes: David Mindell, an MIT professor, says self-driving cars should never be fully autonomous. "There's an idea that progress in robotics leads to full autonomy. That may be a valuable idea to guide research but when automated and autonomous systems get into the real world, that's not the direction they head. We need to rethink the notion of progress, not as progress toward full autonomy, but as progress toward trusted, transparent, reliable, safe autonomy that is fully interactive: The car does what I want it to do, and only when I want it to do it." Mindell writes, "Google's utopian autonomy is a more brittle, less functional solution than a rich, human-centered automation."
safe autonomy that is fully interactive: The car does what I want it to do, and only when I want it to do it."
Sure, if I own the car it should do only what I want it to when I want it to, but why should I own a car at all? I use a car only a few times a month, driving maybe 5000 miles/year total. Why should I spend $30,000 on a depreciating asset and devote 200 sq ft of space towards housing it.
I want to call a car and have it come when I want it, take me where I want to go, then go away until I need it again.
Why does this guy care? What about private roads? Private cars and private property rights? Who gives a shit?
I don't see any arguments to back his theory.
There are already systems that will warn you if you're drifting out of your lane, and systems that will warn you/apply brakes if you're in danger of collision. And of course systems that will plot a route for you and give you step by step directions to your destination have been around for quite awhile at this point.
If the goal isn't full autonomy then it doesn't really seem like we need to do much more research and development. How boring will it be to be "driving" a car that can do 99% of the driving by itself but insists on you paying attention (at least intermittently) to do the remaining 1%, instead of kicking back and enjoying your time doing something else?
(And note that anything less than full automation will provide little benefit to the biggest commercial interest, long distance trucking. Having to pay a person to ride along and babysit the automation doesn't save anything over just making that person drive in the first place.)
This Space Intentionally Left Blank
Now please relax, and watch as we show you this ad. A sedative will be provided shortly.
"First they came for the slanderers and i said nothing."
“The notion of ceding control of something as fundamental to life as driving to a big, opaque corporation - people are not comfortable with that,” -- David Mindell
I'm not sure I agree with that. Sounds similar to someone 150 years ago saying "The notion of ceding control of something as fundamental to life as growing and hunting the food to feed my family to a big, opaque corporation - people are not comfortable with that".
People get comfortable with a great number of things if you make their life significantly better even while asking them to give up a little control.
-- All that is necessary for the triumph of evil is that good men do nothing. -- Edmund Burke
In a related story, the new toaster from MindellCo was announced yesterday. It's just like a regular toaster, except when you push down the lever, it asks you if you're sure before it starts toasting, and then asks again every ten seconds.
From the article:
“[Full automation is] just proven to be a loser of an approach in a lot of other domains,” Mindell says. “I’m not arguing this from first principles. There are 40 years’ worth of examples.”
For how many of those 40 years have today's sensors, computing hardware, and AI been available?
It's possible that fully automated driving will turn out to be hard like commercial fusion power, or like commercial space travel. I think it's more likely, though, that it will turn out to be hard like speech recognition or cheap, lightweight flying drones -- each popularly regarded to be "a few years away" for decades, until suddenly it was here, courtesy of a few research advances and a great deal of exponential improvement in computer hardware.
I don't think we'll have self-driving cars right away. Instead, we'll have cars with "Enhanced Cruise Control." You get into a lane on a highway, hit Enhanced Cruise Control, and your car will stay in that lane (turning left or right as needed) keeping to the speed you set but slowing down if needed (e.g. if the car in front of you brakes). For long car trips, this would mean that a bulk of your trip would be automated. You'd still need a driver there to take control once you wanted to leave the highway and you might not be able to use this during bad weather (just like you wouldn't put cruise control on during a snowstorm), but it would be one step towards autonomous cars.
As the software gets more refined and the edge cases are dealt with better, the car will be able to handle more driving situations. For example, "automatically stop at red lights" or "keep going straight unless the driver indicates otherwise." Eventually, cars driving themselves will be the norm and human drivers will be the exception.
My sci-fi novel, Ghost Thief, is now available from Amazon.com.
Incorrect in principal and practice. It's like the angry bear vs the two people. You don't have to be faster than the bear, just faster than the other human being. The cars' don't need to be perfect - they just need to beat a human mind that is NOT an expert. If the car by itself can do better than a human without the AI, than it is sufficiently good to replace the current model that is human without the AI.
The idea that we need to achieve the maximum possible result of human+AI ignores the current situation's inherent problems of poor drivers, the elderly, drunk drivers, children, etc. etc. etc.
Parents of teenagers, children of the elderly, alcoholics and their loved ones ALL are VERY comfortable with the idea of having the car drive, not the person. They will provide the demand and market. Once their demand is met, then simple continuing research will eventually make EVERYBODY comfortable with letting the AI drive. If you are OK with the AI drive your teenager, your grandma, and your drunk cousin Joe - knowing they might be in the car next to you, then you will be OK with letting them drive you.
His comparison of other modes of transportation such as space, submarine and airplane, is also flawed.
The main reason we never automated those is that their need for accuracy was much much higher than we have for automobiles and up until recently, computers have not had the real ability to beat a trained, expert human. But in cars, they don't have to beat an expert, just a licensed and impaired human - drunk, young, elderly for example.
A better comparison is to look at welding. Originally people welded. Then robots came along and were better. Automated welding has taken over a large proportion of welding, we don't have humans over-riding them. Why? Because the robots are better at it than humans in most cases.
excitingthingstodo.blogspot.com
If you could RTFA, why should you? His premise is that autonomous robotic cars will be substantially worse than well-integrated robotic cars. While the "research" is focused on figuring out how to take humans out of the loop, the manufacturers are very successfully integrating robotics to tie humans more tightly into the decision making progress, with better information, faster decisions, and more fun to drive. This is a good thing. , auto-braking, lane assist and ABS ... these are all significant improvements which where limited authority automation has made the situation safer. However, things like night vision with image processing, blind spot sensors, collision sensors integrate humans better with the machine and have made huge strides in making humans with cars better drivers.
I think the technology being implemented in cars today is not really the most useful technology. Self driving cars are pretty cool from a 'sci-fi' perspective, but inherently dangerous. Personally I don't want any software driving my car! Non-moving computers can be buggy enough.
I think what should have been installed decades ago are safety systems such as proximity and speed limit sensors. These types of devices would alert the driver to potentially hazardous situations and allow them to avoid an accident.
Yet, with all of the "goodies" in new cars these days - wifi; satellite radio; gps; self-driving systems; DVD/Blue Ray players, etc. there really isn't much - if any - high tech safety features (other than ABS brakes, etc).
Most non-technical users (and some technical ones) rarely if ever update software to the latest stable version. I have seen GPS systems tell the driver to make a left hand turn on the middle of a bridge because the software was out of date!
I'll drive my own car, thanks.
Political correctness is really just herd psychology pushed by insecure people who desperately seek social conformity.
I dunno. It doesn't have to do better than me when I'm doing well as long as it can do better than me when I'm not doing well (didn't sleep well, haven't had food or coffee yet, and worried about getting stuff done). That's not a terribly high bar.
Wait: TV sets which self-align vertical and horizontal hold.
Subway systems which run completely automatically (granted with all sorts of staff there to pull the Panic switch and/or make travellers feel more 'safe')
elevators which automatically go to requested floors.
Heck, IP packets which automatically make it to their intended destination.
I think this guy has no idea where and how software is running automated control systems.
https://app.box.com/WitthoftResume Code: https://github.com/cellocgw
One, it is unclear from having read the article exactly what level of automation the author is railing against. There is a huge amount of experimentation on the part of the major players in this race and likely several levels of automation will arrive nearly simultaneously. The best approach will tend to win out in the market. Its not like it will be suddenly all totally automated cars and an “OMG we made the wrong choice” scenario.
Likely we will evolve into fully automated as more and more cars become automated. Eventually it will reach a tipping point where the government needs/wants all the human drivers off the roads for safety and efficiency -- when this happens, driving laws and automated enforcement of every minor offense will force humans to cede control to automation else be fined into the poor house. The biggest challenge to fully automated cars will be dealing with unpredictable humans. The mixed environment for the next 2-3 decades will be quite challenging for all involved.
Letter To Iran
Why Self-Driving Cars Should Never Be Fully Autonomous
I think the author meant "selectively autonomous." "Never fully autonomous" implies there are things it shouldn't do by itself, rather than the author's wish that "the car does what I want it to do, and only when I want it to do it." What if you want the car to be fully autonomous?
What does he think "fully autonomous" cars will do? Force you to go to work when you actually want to skip and go to the beach?
systemd is Roko's Basilisk.
the manufacturers are very successfully integrating robotics to tie humans more tightly into the decision making progress
So the manufacturers are adding things that take the decision making out of the stupid human's hands to more tightly integrate them into not making any decisions?
All of the automation systems are taking the stupid human out of the picture, not integrating them more. Park assist, accident avoidance, lane departure warning, blind spot warning, backup sensors. None of these more integrate the human, most of them augment the human to prevent them making stupid mistakes.
APK likes to ask for responses to the same things over and over. Maybe he just likes the responses?
It could be worse.
systemd is Roko's Basilisk.
Because people are irrational. We fear exotic deaths more than we do mundane deaths. Travel by airliner is much safer than travel by car, yet every time there's a plane crash we have a huge investigation for the purpose of figuring out what went wrong so we can prevent it from happening again. But tens of thousands of people die in car accidents, and all they get is a brief police report stating it was an accident without really delving into the cause. Why? Because death by airplane crash is more exotic than death by car crash. The same thing plagues nuclear power. Death by radiation is exotic, death by falling or getting lung cancer from soot inhalation is not. So we scrutinize and heavily regulate everything to do with nuclear power when it's already the safest power source we've ever invented in deaths per MWh of power generated, while turning a blind eye to deaths by coal (pollution), wind, and solar (primarily falls during maintenance - their diffuse nature means there's a lot more maintenance to do).
You double down on this if the accident was in your control vs out of your control. If you could've done something to prevent the accident (was driving a car) but failed to so, you say "Oopsie, I won't make that mistake again. No give me my keys back." If someone else could've done something to prevent the accident (driving a bus or piloting a plane) but failed to do so, you sue the bastard for everything he's got and try to get him banned so he never drives/flies again.
The combination of these two means autonomous cars have to become a helluva lot better than human drivers before they'll be accepted. Dying because of a typo in a line of code counts as really exotic, and the press will have a field day with it the first time it happens. And the makers of the autonomous cars will need huge insurance policies to deal with the extra liability they'll incur, since it'll likely be bigger than the sum total of all private auto liability insurance policies today (a few percent of the purchase price of the vehicle every year it's in operation).
I have literally never a comment section that didn't examine the "better than a human" part.
Sure, if I own the car it should do only what I want it to when I want it to, but why should I own a car at all? I use a car only a few times a month, driving maybe 5000 miles/year total.
Let me guess, you live somewhere on the East coast or Chicago? Or one of the few other places with public transportation? Out here in the rest of the country we tend to drive a LOT more. I routinely rack up 30-40,000 miles each year. Not because I love driving so much but because work is 20 miles each way and you cannot get anywhere else without driving there. Public transportation for all practical purposes doesn't exist where I live. The infrastructure and population density simply doesn't exist for car rentals to be economically viable and self-driving cars will not change that fact.
Why should I spend $30,000 on a depreciating asset and devote 200 sq ft of space towards housing it.
A reasonable question. In my case, that depreciating asset is the only means of transportation available to get me to a job that pays a lot more than $30K. Your mileage may vary. (pun intended) It also is the only way for me to get groceries and other local shopping done. It allows me to tote my three dogs without worrying about messing up someone else's property. It allows me to come and go as I please and when I please without waiting. One of my cars is actually a lot of fun to drive.
He uses the examples of planes and how humans are constantly correcting human errors. Okay, full automation would not have the human errors in the first place. Also the system would be aware that under no circumstance should a highly perilous course be taken. Actually the article more makes a point for why planes should be fully automated as most of the plane crashes have been human error. That being said, humans are still better at landing planes smoothly, but that will probably change over time.
As for cars, he says most car companies are trying to enhance driver control instead of replace it. Not so, just do a cursory search on automobile companies and find out how many already have fully computer controlled cars, or are working on it. A computer does not get tired, it can look in more directions and pay attention to them all at the same time, it does not take drugs, it does not get angry.
Now he does make one point in passing that should really be the main point, which is some people really want to drive themselves because they like driving. Some people don't trust computers. These social things are the only valid point I found in the article and it is just mentioned in passing.
Microsoft, Apple, Google, Amazon what's the difference? All steal money from devs and control with walled gardens.
I'd be happy to have a car "automated enough" to drive by itself on the boring, shitty parts of driving where you're likely to fall asleep - dull, EASY stretches of highway. That's pretty much the 80% target the automated cars are at today. I wouldn't even mind if it drove half speed, since I could be reading/working/sleeping while sitting there.
If we wait until AI has mastered the complicated, cluttered, HARD bits of driving in cities, construction zones, neighborhoods, etc it'll never happen.
It's just a problem of having appropriate systems to awake/alert the driver before the AI disengages, and/or fallbacks like pullouts where cars can go when their driver ISN'T responding.
-Styopa
You could go to Google, put five things in a "shopping cart", then jump in your car and have it take you to the places that sell that stuff. Preferably the drive-up, so you don't have to actually get out. You could even just duct tape your iPhone to the window, and have your car drive around, pay with Apple Pay and get the stuff for you, then you wouldn't even need to get out of the pool. Or the refrigerator could just tell the car that you were low on milk and beer and it would all be taken care of before you even got out of bed in the morning. This is getting better by the minute. Full autonomy! Woof woof!
If the you can make a car that would drive significantly better then a human (accidents per mile) why wouldn't you?
Nobody is saying we shouldn't but that is a HUGE if you have there. It's very much a hypothetical right now. If you could build a rocket that could get to orbit for $1/pound launched why wouldn't you? If you could build a clean fusion reactor why wouldn't you? Same sort of questions. We aren't entirely sure it is possible though it seems worth trying to find out and people are working hard on the problem.
More like it is so rare AND so many people die that the news organizations play it over and over and Over and OVER and OVER!!!
Now imagine if those news organizations gave that same coverage to every single car crash (with a fatality).
The news would be nothing but car crashes.
And people would start to be terrified of driving anywhere.
Also TFA is incredibly stupid. His examples are meaningless in this context. An autonomous car SHOULD be able to stop itself and turn control over to a human when it encounters something it cannot handle.
And, over time, those cars WILL become more popular because the people who use them will pay lower insurance rates. That is because any accident they are in SHOULD be the fault of another driver OR the programming.
Look at the airports around Thanksgiving. They will be packed with people. Because people see the value in flying. Even when they give up control to someone else and it could result in an "exotic" death. The same with autonomous cars.
I think one of the biggest problem with autonomous vehicles is directing them where you want to go. Let's say you are in a crowded parking lot and you want the car to park in the 3rd spot, 4 rows over. How do you instruct the vehicle efficiently to do that without taking control of the steering yourself? That's not an easy thing to articulate clearly. Worse, how do you tell it where to go when you don't clearly know the final destination yourself? Sometimes you don't have an address or the destination is very large like an airport.
I think autonomous vehicles might do well on major roads but I think the problem of giving specific instructions is going to be a LOT harder than many people think.
The day every cars are fully autonomous, is the day we get rid of car crashes for good.
The only to eliminate car crashes is to eliminate cars.
In the real world, just imagine what happens when Microsoft accidentally release a 'test update' on Windows For Cars Update, and millions of cars download it before they set off for work.
He uses the examples of planes and how humans are constantly correcting human errors. Okay, full automation would not have the human errors in the first place.
No, it would have it's own set of unique errors. Maybe less of them or maybe more of them. But there will be errors of some sort. Failed sensors, interference, logic errors, defective hardware, etc.
As for cars, he says most car companies are trying to enhance driver control instead of replace it.
That's because the full autonomy problem is too big. You have to break it up into bite sized pieces and solve those. Trying to eat the entire elephant in one bite simply isn't possible.
A computer does not get tired, it can look in more directions and pay attention to them all at the same time, it does not take drugs, it does not get angry.
It also is inflexible, completely literal and sometimes challenging to communicate with. I think the problem of instructing the car to take you to very specific locations will be quite challenging. How do you tell it where to park? How do you tell it to go to a place when you aren't certain of the exact destination yourself? Etc. It's much more challenging than just giving an address.
That's true only if your normal driving condition is impaired. In which case, you shouldn't be allowed to drive. The real test is are the autonomous cars better on average than human drivers, and most people are not impaired most of the time (no matter how stupid they seem).
I RTFA'd but what he's talking about and what we really want aren't the same thing. I want to be able to summon a vehicle (I can own it, but perhaps it's not parked nearby), tell it to take me from A to B, and wait. Then from B to A and wait. If it can't reach B, it takes me to the best place it can and either I take over or use some other transit. On my daily commute or trip to the grocery store, or doctors in my town, I don't want the steering wheel. I don't need it, a computer could in theory do that job without my input, better than I can do it. With the proper preconditions.
I agree with the good Dr. that it seems unlikely that any level of car automation is going to work on all road, weather, and traffic conditions, everywhere in the world. But then your average car starts to misbehave when it's off paved roads, and if you're on a dirt road in the rain, it can get bad. One day in NYC breaks the hearts and souls of human drivers, I can't imagine what a machine might think. Much like we built the existing road & traffic system to accommodate reasonably trailed human piloted vehicles, we should be constructing a traffic network that a reasonably decent autonomous vehicle system can safely navigate with exactly this level of input: where do you want to go? The car can handle the transport, I will do other things.
Perhaps this road system exists, for the foreseeable future, only in large urban centers with otherwise unsolvable traffic problems, and human drivers may have to continue to navigate traditional roads the old fashioned way for some of their journeys. The horse and buggy didn't disappear overnight, and dirt roads were common enough in the US even 40 years back. But the impact on my daily life of yielding the car to a computer for my commute even if it achieved no average decrease in commute duration, would be tremendous. Lower insurance premiums, more deterministic commute delays, more scheduling options (ex. in austin I leave before 6:25am, or after 930am, and stay at work until just about 3pm or after 7pm...this is stupid!), more time for me (rather than unpaid time supporting my job, or unrewarding time spent schlepping the family). This seems like what the goal of it all should be.
I think the Dr. is probably right that putting the entire burden on some sophisticated AI to handle conditions even humans can't read properly is very unlikely, but we also can engineer our environment to fit the needs of the AI.
I have literally never seen a comment section that did so in any substantive way. Certainly nobody in this comment section has done so. We are far enough away from automated cars being able to drive in real world conditions that we cannot even estimate when it will happen.
The real world is a chaotic system. You cannot have a 100% accurate map of it at all times. No self driving car being developed today is even trying to deal with unexpected conditions. Add an unmapped stop sign on a road, and Google's car freaks out (which means it slows to a crawl, and so does everyone behind it). Snow? Forget it. Rain? Same. Road construction, with a guy with a flag? Laughable.
We might live to see cars smart enough to drive in the real world, but there's really no reason to be optimistic.
So you'd be OK with all cars being worse drives than most human drivers are now, so long as they're better than the worst now? That seems rather . . . foolish, to me.
This, this, this. A million times over. Humans are really bad at estimating real risk, are even worse at comprehending scale, and tend to fear the exotic or unknown. Nuclear power gets the triple whammy here.
Real Risk: The health/death risk of Nuclear power to the public is so vanishingly small it disappears on any graph you can create. There have been 9 or 10 deaths in the 60 odd years of civilian nuclear energy in the US. None by radiation. About half by electrocution ( they are electric plants... ) and the half by physical events, pressure explosions, heavy things falling.
Scale: 100 Nuclear power plants have produced about 20% of the entire US electric grid for the past 60 years. In contrast, 600 coal plants produce about 40% of our needs. On top of that each coal plant needs a constant supply of coal, and much more land to operate. The land use efficiency of Nuclear is much, much better. As far as the Nuclear waste argument, here again, failing at comprehending scale. The spent fuel rods consumed for the past 60 years of nuclear power would fit into a single house.
I would like to give some credit where credit is due, and where reasonable regulations apply. Airliners are extremely safe because of the NTSB doing amazing work, getting to the bottom of every major crash and then back feeding that information into actionable fixes. The same goes for the NRC. I am hugely pro-nuclear power, but also believe the NRC is an important part of that environment.
This is all a fallacy based on the "wishes for the future". Actual automation does not work in this fashion EVER.
Repeat the following until is sinks in: "Automation of any human ability leads to humans losing that ability for themselves." (Now repeat...)
- Give children calculators and they will not be able to do math in their head.
- Everyone has a cell phone with a contact list so now people can't remember any phone numbers.
- Give someone a car with an automatic transmission and you end up with someone who aims a car instead of drives a car.
Either make the cars self driving or take all the touchscreen/infotainment crap out of the cockpit. Anything that can't be operated without taking your eyes off the road has zero business being in a human operated car.
Oh, and the idea that automation is making people better drivers is insane. Automation makes it so people can spend less time being engaged with the process of driving. (People's attention span is gone anyway.)
Auto-braking = I don't need to pay attention to what is outside that big window in front of me! Lane assist = I don't have to pay attention or even know where my 2 tons of steel is! ABS = I don't need to learn basic car control techniques.
At best all the automation comes down to an automated slap up side the head to notice what you should have been paying attention to anyway.
Planes mostly fly themselves but we still require qualified pilots in the cockpit. Software engineers are awesome but there are still bugs in code being patched every day. I want a licensed driver in the vehicle if it is on public roads.
Autonomous self-driving!
Think about thatâ¦
Cars. Why cars?
Cars provide self-sustaining R&D funding capital along the path to autonomous robotics. Cars are the highest and best embodiment for human-robotic technology development. Developing cars enables companies like GOOG, AAPL and Tesla-like competitors to affordably grease the runway for the autonomous economy future. Future progress promises autonomous action without the detrimental reliance upon an irrational unreliable primacy of man of modern economies.
MIT street-cred notwithstanding, the argument for non-autonomy is simple strawman which serves to appease the present status quo in the ascendancy of technology over human based economics.
With any of this, there needs to be an absolutely manual override. Software can get hacked or just freak out, and the passenger needs to always have the option to take control of the vehicle, especially on a large scale, like a major bug or hack. I don't see a world where law enforcement isn't going to want a backdoor that shuts down the vehicle or runs some program to pull it over or even drive it to a certain place. This hat is itchy...
X
There's a huge difference between "drive by wire" and fully autonomous. Signalling your driving intent through the wheel and pedals, and letting the computer work out how to make that happen safely, is still driving. Much like every fighter plane has worked that way since the F-16, but they're very different than autonomous drones.
An autonomous car wouldn't need a forward-facing driver's seat. The only controls would be the destination. That's a very different world from "computer-assisted driving", where you're still giving the car input moment-by-moment, and the car's just helping you overcome your limitations (puny human with no 360 degree night vision with distance calibration).
Socialism: a lie told by totalitarians and believed by fools.
it seems like you're saying that "insufficient sleep/caffeine + stress" == impaired. Is that correct? Because if so, there's a lot more impaired drivers than I think you think there are. If you're using some other definition of impaired, I'm not sure what it is or how it relates to my post. Can you please clarify?
First of all, autonomous cars, lets just call them "automobiles" (harhar) needn't drive perfect to be fully autonomous.
They only need to drive better than humans.
And - Newsflash! - they already do that!
And even just right now I'd trust a google car way more than I'd trust at least 20% of human dimwits at the wheel today.
Testing phase or not.
And that's in Germany, where driving training is a very big deal, takes long and is very expensive and elaborate.
And behaviour in traffic is compareatively civil.
Second of all, TFA says: "Yet as Mindell also observes, there are many challenges to the Google model: Its cars must identify all nearby objects correctly, need perfectly updated mapping systems, and must avoid all software glitches."
Well, no shit, dude.
"Avoid all software glitches" is called "testing" and/or "test driven development" and/or "design by contract" and/or "correct error handling". Like, for instance, warning the driver ... errrm, passenger, when there's a severe problem and they need to stop and he/she needs to get out... It's basically non-douchebag software guys doing the sort of thing any regular respectable engineer would do when designing a bridge. And, trust me, those folks at Google aren't your Type-A hobbyist/wannabe WordPress Plugin Scriptoid - they actually know what they're doing.
And now thats aside, yeah, an autonomous car needs to recognise all those many things. Well, guess what? That's exactly what an autonomous car today is by order of mangitudes *better* at than any human will ever be. For enlightenment I strongly recommend this talk by the head of Googles Autonous Car division, Chirs Urmson, "How a driverless car sees the road". Yes, it's a TED talk - you're gonna live.
Now could someone send this guy the memo?
Thanks.
We suffer more in our imagination than in reality. - Seneca
The problem is that you're comparing the current self-driving car to a *perfect* human. Yes, Google's cars might have problems in bad weather, with flag signals, etc. But they don't have problems with falling asleep, texting, being distracted by the two kids fighting in the back seat, etc. And having driven in the south after a light snow any number of times, I've seen an awful lot of humans who simply cannot be trusted with a car in the same weather conditions you're complaining that SDCs can't handle. I've seen three really bad accidents in my life- two were caused by the driver falling asleep at the wheel, the third by a person who wasn't paying attention to their right side blind spot. SDCs would have avoided all of these. I've been in two accidents bad enough to set off airbags- both were caused by a driver suddenly stopping to avoid a problem and the following car not paying enough attention to stop in time. And while I'm a pretty good driver, I caused one of these- it was a moment of inattention that a SDC wouldn't have had.
"Seven Deadly Sins? I thought it was a to-do list!"
Because I like to drive?
A fully autonomous system can only react properly to those situations which the programmer has anticipated. When something unanticipated happens, chaos breaks loose.
Even with non-autonomous vehicles, chaotic situations can happen. But at least there's a better chance of a real person being able to respond properly to unanticipated situations and therefore minimize the damage.
How do autonomous vehicles fare when an oncoming drunk driver zones in on their headlights, veers into the lane and tracks the autonomous vehicle as it tries to avoid the collision?
How will the autonomous vehicle avoid the T-bone collision from the driver that fails to stop at the red light on the cross street? Does the autonomous vehicle have peripheral scanning that will detect a cross-traffic vehicle that doesn't appear to be stopping?
How about four fully autonomous vehicles that approach a 4-way stop from four directions at the same time? Who gets to go first? Will they communicate somehow?
Give me my freedom, and I'll take care of my own security, thank you.
Do they stop randomly because the computer gets confused, resulting in other cars plowing into them for no good reason? If so, they still have a long ways to go to make the computer better. While legally that kind of accident is the fault of the other driver, realistically it's caused by the computer doing something entirely unexpected that other human drivers for the most part never do. And that's the rub. It may be better at being more conservative a driver than the average human, but it's really still not a better driver overall than the average human.
If you're only telling the car where you want it to go, and it's figuring out all of the correct maneuvering to get there safely, what's the point of making you micro-manage the operation? Just let me tell it where I want to go, and it can wake me up when I get there.
We hope your rules and wisdom choke you / Now we are one in everlasting peace
Also TFA is incredibly stupid. His examples are meaningless in this context. An autonomous car SHOULD be able to stop itself and turn control over to a human when it encounters something it cannot handle.
I don't think you (and many other people) have really thought this one through.
If one is riding along in an autonomous car, they are not going to be paying attention to their surroundings. They're going to be talking to others, texting, surfing the web, daydreaming, or even nodding off, if not entirely asleep. They might look around once in a while but they're not going to maintain the kind of focus on the road that would give them any kind of real situational awareness. It's just not going to happen. There's no reason to pay attention when your car is driving itself. And don't kid yourself. People may say "Yeah, I'd stay focused on the road." But they're full of crap. Anyone behind the wheel is not going to pay anywhere near as much attention to the road when the car is driving itself as they would when they're driving the car. If you have to pay attention to the road, why have the car drive itself in the first place?
Suddenly, the car runs into a problem that it can't deal with so it hands over control to the driver. But this "driver" has no idea what's really going on. Hell, it will probably take a fair number of seconds for the person to realize that the car has handed over to them in the first place. Reaction time will be slowed substantially by the fact that they have been lulled into a false sense of security. They are entirely unprepared to take over at that moment. Suddenly being forced to change focus is difficult and time consuming. And a lot can happen in the many seconds it takes for the average person to realize what's going on well enough to do something about it.
But here's the thing. The only reason the car would hand over control is that it is in a situation where it doesn't know what to do and that usually means there's something significantly wrong RIGHT NOW! Given that, do you really think it's a good idea for the computer to just say "screw it, I'm out" and suddenly dump control over to a passenger who has no idea what's going on around him?
No, the idea that an autonomous car would have an option to hand over control to a passenger in the car is ludicrous on it's face. There's no way in hell anyone would design a car to do that (for the consumer audience) once they spent a few minutes understanding the problem. For research cars that get used in controlled environments? Sure. But the first average consumer caught napping when their car handed off control would sue the hell out of the manufacturer if they survived the crash. Or the family of the person would sue if they didn't. Either way, that kind of liability is just not something a manufacturer would ever consider taking on when they're building for the public.
We're not going to see autonomous cars handing off control to the driver. Not once they hit the consumer market. Or at least not long after they hit the consumer market.
His main theme is, people do not want it and it did not work that way in the past. While it is usual to determine the future by past experience, it is not a totally save method to predict the future.
Shall we get off your memorized Fibonacci series, manually shifted and pushed-uphill-both-ways-in-the-snow lawn now?
Faster! Faster! Faster would be better!
This whole argument is a fallacy based on absolute ignorance of *actual* human behavior, history, and even basic linguistics.
Children have had calculators for *decades*, and still learn how to do math in their heads. (In fact, students today are actually being expressly taught the methods that work best for doing math in your head, instead of having to stumble across the methods on their own. That accounts for about 90% of the 'Common Core suxxors' arguments you see out there.)
People have had contact lists for *centuries*, and still remember names, addresses, phone numbers, birthdays, etc. Storing the list electronically is no different than storing it on paper, except that if you lose your electronic device, you can probably recover all that information instead of having to painstakingly replace it over the course of the next few years.
Apparently you think that manipulating a clutch with your foot and a lever with your hand is what 'driving' means. Strangely enough, people have driven vehicles that lacked *both* of those for *centuries* before they were invented, and are *still* driving vehicles in which manual access to those controls isn't required.
There ALWAYS needs to be a manual override in case something goes wrong
I would NEVER buy an autonomous car without an emergency manual mode
The pundits and futurists who write about autonomous cars need to spend some time talking to people who actually manage highly automated facilities
Something ALWAYS goes wrong
For the same reason the military does use (at at least doesn't admit to using) fully autonomous armed drones: software can't make judgement calls. Sure, eventually, one day, the software may get there. I think it will be less than 20 years, within well-mapped cities. In the mean time, we can keep making computer-assisted driving better, which is what most car companies are actually doing. The few who aren't will still require you to pay enough attention to take over when things go wrong, which sounds dubious to me.
And for rural driving? That's a whole different world. We won't have fully autonomous cars on dirt roads, or heavy snow, or similar conditions that most humans find challenging, any time soon.
Socialism: a lie told by totalitarians and believed by fools.
To say that full autonomy shouldn't be the goal is like saying copulation shouldn't be the goal of sexual contact. Instead, our goal should be to masturbate and then shake hands with our partners. Some people might want to do that but that isn't the goal.
What would a non-autonomous, self-driving car even look like? I would think the goal would be to tell the car where you want to go (it already knows where you are) and it goes there. If you want to take over the controls, that's a valid use-case but that isn't the development goal.
Because it is not at all obvious that an human augmented vehicle would be better. At some point if you take enough of the task away from humans they will simply lose the skills needed to drive in the first place, people already talk on cell phones, do their makeup while driving, if you make driving easier it is quite possible people will simple start paying less attention.
The article's argument states that the endpoint should augmented human control rather than fully autonomous and uses the example of current extreme environment vehicles aren't autonomous so cars shouldn't be.
This argument has some flaws, first we haven't developed any fully autonomous vehicles so of course no existing existing vehicles are autonomous. Secondly vehicles like the mars rover, or deep sea exploration vehicles have a different purpose, which is to explore, this requires a more fine grain of control, not just go to point A. Thirdly these vehicles are in unknown environments so it is harder to program them cope with unknowns, you can't send machine into a well mapped environment if its job is to map that environment.
Ultimately it probably would be an ultimate goal to send out fully autonomous robots to discover interesting things on other planets, unless we discover a means of faster than light communications.
âoeThe notion of ceding control of something as fundamental to life as driving to a big, opaque corporation - people are not comfortable with that,â he says. Additionally, other companies and research groups looking at automating cars are âoevery clearly not going for the Google approach to fully driverless cars.â
And so many of you said I was nuts; not so nuts now, am I?
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
Dark and rain are not problems for Google cars or cars using Mobileye company technology (Tesla) The big problem is snow and ice. Humans are amazingly advanced at driving on snow covered roads. I foresee trouble in the parts of the world with snowfall if a generation of drivers emerges that is very dependent on self-driving technology. You need lessons and experience to become good at snow driving and I feel self driving cars are many years away from being safe in snow and ice.
Power tends to corrupt, and absolute power corrupts absolutely.
An autonomous car wouldn't need a forward-facing driver's seat.
Yes it would. Do you envision the entire vehicle to be unusable if it's not connected or if the on-board AI breaks down? How does the occupant get out in a power failure?
You're not going to build a successful vehicle without considering those factors. Human driven it may not be, but it still needs to make affordances for the human psychology of it's human passengers.
Make sure everyone's vote counts: Verified Voting
The problem is that you're comparing the current self-driving car to a *perfect* human.
No, I'm not. I'm comparing the best self driving car to the average human driver the overwhelming majority of the time.
A car that can't be used in rain, snow, or other weather, or near a construction zone, or on a street with an unmapped pothole, or on over 99% of the roads and streets that haven't been mapped and likely never will is of zero value to the anybody other than the people developing them.
Slashdot claims that it is more signal than noise but it is becoming more shill than signal. This is another article written with a clear agenda in mind. This is to promote cars that fit the model of the traditional car companies. They are beginning to freak out at what the driverless car will really look like and are beginning to sponsor anyone who will say that it will look like more of the same.
Here is the driverless car of the future (not the next generation but a few generations of development away). First the car is on the road with all automatic cars, there are no manually driven cars because they have been proven to be more of a burden to society than we were willing to put up with. Crashes are pretty much a thing of the past. The cars have shed nearly all their safety gear and no longer have to pass onerous safety tests. The only remaining safety tests are that the cars need to go a certain number of miles while not breaking down.
So the driverless car uses one of a handful of off the shelf autonomous systems all of which are battle tested and battle hardened. The cars are of a variety of shapes and sizes with many tiny single user cars popular among commuters who are one of the last bastions of private car ownership. Most people couldn't tell you one brand of car from another as they just call them on their phone and it shows up. They no more pay attention to brand or model than people do now with uber or taxi cars. Someone might notice if the taxi were a hummer but any boring midsized sedan and they can't even tell you the manufacture let alone model.
So looking at the manufacturers they are plenty in this world of SDCs. This is because the large companies have lost their competitive advantages starting at the moment that self driving cars began to rapidly evolve. The old car companies were very good at tooling up very cost efficient assembly lines and then making roughly the same car for nearly a decade. But the SDC evolved very quickly much like the cellphone which resulted in whole assembly lines being completely out of date in less than 6 months. Also the delay of the assembly line allowed complete upstarts to pound out whole new generations of cars in less time than it took the old companies to get a single, out of date, model to market.
Then in the end with only bulk fleet buyers making up the market the traditional skill of mass marketing was just another department that needed to be shuttered by the old car companies while the new companies didn't have the same liabilities. Also the new companies integrated every modern manufacturing technology possible without "Proper" testing and review by the senior engineers. This meant that most of the upstarts failed but left knowledge in their wakes that other startups built on resulting in fantastic new cheap ways to make very high quality low cost cars.
But the worst insult to the old car companies will be that as the kids start using self driving fleet vehicles their desire to own a car or even give a crap about a car will approach zero. There won't be 16 year olds with posters of a car that the car company can sell to them when they are 50. Movies like the fast and the furious will make no sense and thus won't sell a single car.
This whole allowing people to override the robot will very quickly be proven entirely stupid when the stats will show that cars in manual mode are some massive multiple more likely to crash than cars in fully automatic mode. Also they will be able to run simulations against the manual mode accidents to show that had the car been in control there would have been no accident.
So to allow people to continue to have any control over what is really a robotic car is the rough equivalent of those steering wheels we buy children for their car seats so they can go vroom vroom and stay out of our hair while we drive.
The spent fuel rods consumed for the past 60 years of nuclear power would fit into a single house. ...
A house of several miles side length, yes.
Facepalm
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
You might want to look at the accidents the google car has been involved in. And look at them in depth. I think you'll find, as I did, that the car stopping suddenly for no apparent reason lead to accidents. While that makes it a legally "spotless record", there is still a lot of work that needs to be done to keep the car from being the cause of accidents.
Really the human in charge is the weak link in a car. It is not as simple as common errors but moods, substances, and mental issues as well as family or financial problems can lead to a crash. The entire benefit of robotic autos is getting the humans out of the loop. It is nor quite practical today but it will occur soon enough.
More like it is so rare AND so many people die that the news organizations play it over and over and Over and OVER and OVER!!!
It isn't because it is so rare, it is because an airplane "disaster" is almost always a case of people dying, or being in grave danger, who were not in control of their fate and were required to "trust the system" that pilot training and standards, and aircraft maintenance, were being upheld.
Yes, they can "control their fate" in the gross sense of "don't fly", but once you decide to fly you are in the hands of a pilot you have most likely never met. You hope the company is obeying airworthiness directives. Did everyone who is shipping cargo on that plane obey the prohibitions on unflyable cargo, and was it stowed properly?
From that view, airplane crashes are "failures of the system" and that makes them news. You almost never hear about an airplane crash that involves a private pilot, outside the local news where the pilot used to live, unless there is a "failure of the system" that creates other fatalities, or chances of same. Otherwise, the "failure" involved is the one person who made his own choices and failed because of his own failure.
That's how most car accidents are today. Someone failed under their own steam. Joe Idiot fell asleep and ran into a tree. It's news when Joe Idiot takes out innocent bystander, because bystander was trusting in the system.
Fast forward to when the system is autonomous. Today, autonomous failures are not reported often because big companies are spending big money on keeping them quiet. When the freeways are filled with these things, any crash will become a failure of the system and it won't be kept quiet for long.
It will be much less than 10_years in the long haul trucking industry. Cities within 15. The people who are really worried ate the auto insurance industry. When people only need fire and theft because accidents are covered by the manufacturer, and the car can report someone is trying to steal it, and it doesn't matter how old or experienced the driver is, don't be surprised if the manufacturers cover fire and theft as well. The only thing you'll need to cover for would be uninsured driver, vandalism,s and falling pianos. Expect vandalism to drop when the car has video of the perps.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
"Some claim that the Pipe actually controls "Bob", but .. the Pipe no more controls "Bob" than we control the cars that carry us around."
There are people alive today who will live in a world where humans aren't allowed to drive on public roads.
The Moore-Murphy Law: The number of things that will go wrong will double every 2 years.
"More fun to drive" is just marketspeak for "more tempting to do something stupid." That's why sports models cost more to insure than grannymobiles.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
Death by airplane is "someone else's" fault. Death in a car is my fault. Even if someone else was negligent, it's still my responsibility to compensate for their idiocy. This is the Merican way. Personal freedom, choice, control, and consequences. I deal with this every day as a motorcycle rider (the ultimate in personal control and consequences).
The simple fact is, on a commercial air-flight you have absolutely no control whatsoever over your death. As a driver of a car, you do, even if it's someone else's fault. In other words, flying on a commercial airplane requires no skill, driving a vehicle does.
The ratio of people to cake is too big
The whole article is just a book promotion and is full of some poorly thought through arguments.
That said, the real reason we wont have autonomous cars soon is the same reason captcha bot checks still work so well in 2015 - our ability to get computers to successfully recognize patterns is still in its infancy. The same reason why a 5 year old with no training beats the best web crawling bots and galaxy classification and analysis for many systems is crowd sourced to people with 10 minutes of training.
I would argue that we nearly have the processing power necessary already and average humans currently beat the pants off state of the art autonomous systems with inferior sensors (stereo cameras and microphones, dual 3-axis accelerometer and gyro and some tactile sensors). What we lack are the creative algorithms that will allow better than human navigation.
You don't really seem to know much about modern self-driving cars.
Google already has self-driving cars driving on roads around California and Texas. Sure, both states seem rather unreal to me at times, but I'm pretty sure they count as real-world.
You don't need to deal with ANY possible situation. If a rhinoceros comes running down the road I'm driving on, I don't need to know how to deal with it; I'll just slow down, pull to the side or evade the animal, and stare at it in confusion. A self-driving car just needs to follow similar rules of thumb; "slow down, get out of the path of traffic, call for help".
The cars are not perfect; they need more practice in rain. Snow and ice challenge most humans, so that will take a bit. But they are pretty good with guys with flags and hand gestures. Unmapped signs are fine; do you really think they have a trusted list of all signs? No, that's why they have cameras. Don't need perfect maps, though better maps are always a plus. As I said, you don't seem to know much about the technology, and I encourage you to learn more about it before loudly asserting things which are untrue.
They already drive better than humans (though in a slightly more narrow set of circumstances); that's a really easy bar to beat because humans are terrible drivers.
Humans don't need a 100% accurate map of the current system - why should an autonomous car? It will eliminate all the people driving on autopilot who suddenly realize that they're about to pass their exit.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
For the same reason the military does not use (at at least doesn't admit to using) fully autonomous armed drones: software can't make judgement calls.
Software can make judgement calls. In the case of weapons, we just don't feel comfortable allowing that. For self-driving-cars, there is no moral issue.
Sure, eventually, one day, the software may get there.
You might want to read up on what autonomous vehicles can do. They have already driven millions of miles on public roads, and have a safety record better than average human drivers.
However, an early objective was to make the missions fully autonomous, able (in theory) land on the moon and return without any contact with Earth. This was because of a concern the Soviet might try to actively jam communications in the event of the Cold War turning very very frosty.
Yuri Gagarin was a passenger on the first space flight in 1961 as his spacecraft was indeed fully automatic. It''s controls were locked out by a three-digit combination lock on the insistence of the doctors, who thought there was a chance spaceflight might make him go psychotic.
The head of the program thought this was BS, and was much more worried about an in-flight emergency that might make the controls necessary, and also kill communications with the ground. Consequently, Gagarin was quietly told what the combination was before the flight, when no doctors were around.
Do you envision the entire vehicle to be unusable if it's not connected or if the on-board AI breaks down?
There's definitely a place for cars that can't be driven manually - no controls, the car is like the back of a limo, or an old-school train car where the seats faced each other, with perhaps a table in-between. No different from "can't be driven if the engine brakes down", and moving parts fail with age in a way software doesn't.
How does the occupant get out in a power failure?
Through the ... doors?
Socialism: a lie told by totalitarians and believed by fools.
You just let the car drive along the lane and tell him: park here. Same decision process if you are driving yourself.
Not that simple. What does "park here" mean? Left or right side? Which lane? What if there are no defined parking spaces? Which parking lot? My airport has 3 parking decks with multiple levels each. How do you communicate all this nuance to the computer efficiently? The easiest way is simply to take over the driving physically because verbal communication in this case is actually quite difficult unless the computer can process information equally well as a human. Current state of the art is something like Siri which is no where near what would be needed to accurately navigate a car. Frankly I think people would get hugely pissed off trying to tell the car where to go rather than simply steering it themselves.
Finding a parking slot never was so easy, with autonomous cars tolerating and honouring the first come first principle and let the first car that 'booked' a parking slot indeed occupy it.
Plenty of parking does not involve neatly defined spaces. How do I tell it that I want to be backed up across the lawn to my front door? I don't think you are really appreciating the difficulty of the communication problem here. We have a hard time communicating this stuff to other humans. We're not going to be better at doing it with a computer.
And you can pick that slot ofc on your mobile or tablet.
Or I could take the MUCH easier approach of grabbing the steering wheel and navigating the car to my exact preferences myself. Telling it what to do on a smartphone is nothing more than an abstracted and clumsy form of driving. Might as well grab the wheel if you are going to do that.
You're making the same mistake that TFA makes.
No one is saying that a car IN MOTION should cease autonomous operation.
What I said was that the car should STOP and then turn over control to a human when it encounters a problem it cannot handle.
oftware can make judgement calls. In the case of weapons, we just don't feel comfortable allowing that. For self-driving-cars, there is no moral issue.
That's not the way most people use the phrase "judgement call" - if there's a rule you can write down, it's not a judgment call. We do have weapons that will kill anything that looks hostile with millisecond response time, but only where there's no judgement call required in identifying targets.
You might want to read up on what autonomous vehicles can do. They have already driven millions of miles on public roads, and have a safety record better than average human drivers.
Most of the time. That's at least a decade from reliable. It's a world of difference. Nothing today is anywhere near safe if a human isn't ready to take over the controls. Sure, there are routes that are safe, in most weather, if there's no detour or policeman waving traffic down. Sure the cars are safe in normal conditions. That's the first half of the problem solved.
Socialism: a lie told by totalitarians and believed by fools.
Stopping suddenly because the computer goes stupid is only slightly less dangerous than handing over control to an unprepared driver. If you're in the middle of any kind of traffic, you're suddenly an obstacle that everyone else has to react to. The chain reaction crashes caused by stopping suddenly in the middle of the road are simply not acceptable. Sure, legally, you're not responsible for other people crashing into each other (or you) if you stop suddenly. But it's still the root cause of the accidents and not an acceptable behavior.
If your AI is not capable of handling driving and it has to stop suddenly to hand over control to a driver, it's not ready for production. Period.
Autonomous cars seem to be a product that is supply driven. Of the folks I know only a couple actually dislike driving, many hate bring a passenger, especially the two who easily get road sick. The whole thing strikes be as a "because we can" proposition.
I rather expect it to be a shiny feature that most folks will quickly tire of and shut off.
I agree with the possible part.
However, I don't like to make the mistake of underestimating the luddites and NIMBY crowds.
Right. Because stopping in the middle of traffic because the computer is stupid is a much better solution..
If your AI can't comprehend the situation, it's not ready for use in the real world. Period.
I have to admit.... I'd be more concerned about my automated vehicle defending itself from clever theft than I would be about it being in an accident with me in it.
Because reasons. Seriously this should be emberassing for MIT to come from one of their own.
"If robotics in extreme environments are any guide, Mindell says, self-driving cars should not be fully self-driving. That idea, he notes, is belied by decades of examples involving spacecraft, underwater exploration, air travel, and more. In each of those spheres, fully automated vehicles have frequently been promised, yet the most state-of-the-art products still have a driver or pilot somewhere in the network. "
Let's break down the reasons why this is an idiotic statement.
1) "If robotics in extreme environments are any guide." Why would extreme environments be a guide to how people commute. Previous Robotics were all about going to extremely hazardous environments where people couldn't live to learn things about places we know little to nothing about. Compare that to our transit system which by definition someone knows so well that they built a road there and the caveat makes no sense.
2) "That idea is belied by decades of examples" So because older inferior computers and analog systems were limiting we'll be limited for all time? Sorry but no. That's like saying "Because the horse and carriage have always traveled at 30mph or less, no good will come from traveling at speeds in excess of 100mph."
His entire argument comes down to fully autonomous cars (taxis) are bad because space exploration requires human direction and Apollo era hardware/software wasn't sufficiently capable of handling 100% of a mission.
When in doubt. Slow as quickly as possible. If you have a safe following distance (which you always should have) then unless a problem emerges somehow out of thin air the worst that will happen is that you can't proceed. But you should always understand the condition of the road ahead of you or stop before you reach it. That's true of autonomy and true of real human drivers. You drive only as fast as you can see. If you can't see around a bend, drive slow enough that you can stop should something be stopped around the corner. If you can't see far enough to complete the pass, don't pass.
If a autonomous car encounters something it doesn't understand, it can simply come to a stop because it should have entered its FOV beyond its stopping distance. Then you can (wake up) and give it the go-ahead or not.
Why aren't those other vehicles reacting to whatever caused the car to stop and turn over control?
But anywhere that's populated enough for a taxi service is populated enough for an autonomous taxi service.
A taxi service isn't more useful just because it doesn't have a driver. A taxi service IS available where I live and do you know how much I use it? Never! Because it is economically inefficient for me except in very rare circumstances. I drive over 30,000 miles a year and that's normal where I live. Eliminating a driver from a taxi will not change that. Owning a car is far cheaper given my transportation needs. Furthermore how do you propose I get a taxi to help me bring home a load of dirt for the garden? Or 2x4s for construction? You really going to take a taxi to the grocery store? How do you plan to store the car seats for the little ones after taking a trip to the mall in a taxi?
Seriously, you haven't really thought this through...
What is begin ignored is that some of the biggest real-world use cases are driver-less or have a driver-less component.
Drop me off at work, then return home for my family to use. Come pick me up at the end of the day.
Drive my elderly grandmother somewhere. Or me when I'm sick/tired/impared/got a good book.
Distribution, getting loaded at warehouses by their staff and dropping of at final locations, unloaded by their staff.
LITTLE GIRL: But which cookie will you eat FIRST? C. MONSTER: Me think you have misconception of cookie-eating process.
Stopping in the middle of traffic because the computer can't figure it out means the computer is not ready to drive autonomously.
Because people can tell the difference between a crumpled up newspaper in the road and a baby in the road. They won't need to stop for the first one. The current autonomous cars have trouble telling the difference and do have to stop.
You have to solve a ton of very difficult technical problems to get an autonomous car to be able to comprehend as much or more as a human. Those problems are nowhere near being solved. I'm not saying that we can't solve them but it's going to take a ton of effort by a bunch of very smart people who aren't guaranteed to be successful before an autonomous car is really ready for the mass market. And it's good that people are doing the work and I wish them well but when it comes to the safety of not just the occupants of the cars but the occupants of the cars around the autonomous car, I think we should hold the AI to a very high standard.
I seriously doubt anyone would ever be able to sell an "autonomous" car that might, at literally any moment, at any speed, in any conditions hand control back to the human, for exactly the reasons you state. Who would want to buy such a thing?
A more likely outcome is that when something that can't be handed occurs, the car comes to a controlled stop (maybe quickly?), and then hands over control. FOr example, if the road is significantly different than expected (maybe a fallen tree, or subsidence or something). Or, perhaps just it meanders up to the farm track that's overgrown with weeds and says "the GPS says to go this way, but I can't figure it out - over to you buddy".
Of course, once the technology matures, then the frequency of human interventions will be so small that cars will be made without any human controls in them. The worst thing those cars will do is say "sorry, I can't get you to your destination - wanna go back home?".
If you're a passenger, how would you tell the human driver that there's an open parking spot 4 rows over, 3 spots in? I imagine you'd be able to tell an automated car the same way.
The passenger is a human who can understand language at a human level and even then we get it wrong a lot. We currently have no computers capable of even close to that sort of level of capability and are in no danger of getting it soon.
It's probably even easier, as the automated car will likely have something like a touch screen for input, which can show you on a map and/or on a camera view exactly which spot it's taking you to. If it's the wrong one, you can correct the car verbally, or even just tap on the correct space on the touch screen.
If you are bothering to use a touch screen then you are controlling the vehicle and you may as well just grab the wheel yourself. The car can override you if you look like you will his something. Seriously, do you REALLY want to navigate a car via iPad? "Computer... wait you passed the spot. No I didn't mean that one I meant the one over there. Tap, tap, tap.... Stupid computer..." Seriously, I think you really haven't thought this through at all. You're thinking it'll be some Jetson's technology that will magically infer your intentions in fine grained detail and I think you have no idea how difficult that human interface problem really is.
Ah, I think the problem is that you're thinking of automated cars like airline flights, where you buy a ticket to a destination before you head out, and once you board you can't change anything.
Not at all. I'm pointing out that communicating anything much more complicated than an address or an intersection is going to be a REALLY hard problem to solve. I'm not saying it's impossible but it is going to be super hard to do well. Basically I don't think we are going to be able to strip out driving controls from most vehicles for a very long time.
Military drones *can* be fully autonomous when it comes to movement. They are *also* able to be controlled by a pilot. The judgement calls are for when weapons are being used.
The original post only sounds incredibly stupid to you because your scenario is incredibly stupid.
If the autonomous car with driver fall-back runs into a situation that it can't handle it will start an audible/visual alarm while bringing the vehicle to a controlled stop. The human driver can take stock of the situation and if they decide to, they can take over once then have the situation figured out, or they can simply wait for the autonomous system to sort out what's going on and continue. The autonomous system wouldn't just stop controlling the car, it would stop the car.
This idea that the car would just stop functioning with the vehicle in motion and require the human occupant to take control in a split-second is ridiculous.
"Grab them by the pussy" -- President of the United States of America
If your AI is not capable of handling driving and it has to stop suddenly to hand over control to a driver, it's not ready for production. Period.
If you are not capable of handling driving without having to stop suddenly to avoid an unexpected situation, you're not ready for a driver's license. Period.
"Grab them by the pussy" -- President of the United States of America
I maintain a safe following distance. I can't necessarily say the same for the guy behind me. Neither of the rear-end accidents I was in was my (or my wife's) fault, but it didn't stop my back from giving me serious pain and some incapacity for months.
"When you have eliminated the unacceptable, whatever is left, however improbable, must be the truthiness" - Holmes
A lot of old people lose their licenses when they can no longer drive without being a danger to others around them.
The only control I need in an autonomous car is a stop/abort button. Why? Because when the car starts to swerve dangerously I'd probably be too panicked to decide whether to hit the pedestrian or hit the back of the truck. If I press abort, the car should deactivate itself gracefully. As for manual controls, the steering wheel is itself an abstraction of the actual state of the car wheels. The steering wheel is effectively a primitive computer that automates the direction the car wheels point.
ABS = I don't need to learn basic car control techniques.
I agree with most of your post but not this one. In fact I think it contradicts your main point that people can't do what they don't practice. Expecting people to remember and apply correct emergency braking techniques in the few seconds they have to react during a panic-inducing situation, despite never having the opportunity to practice them never worked all that well. It is unrealistic for the same reason that expecting people to be able to manually control of a normally autonomous car in an emergency situation is unrealistic.
The Japanese have had that handled for years. Not a problem.
Fuzzy logic. It's what's for dinner. :)
I've fallen off your lawn, and I can't get up.
It'll be bigger than a house, but you are also wrong. The amount of waste from US nuclear power plants for the last 40 years fit in one football field. What is your source?
http://www.nei.org/Knowledge-C...
"Over the past four decades, the entire industry has produced 74,258 metric tons of used nuclear fuel. If used fuel assemblies were stacked end-to-end and side-by-side, this would cover a football field about eight yards deep."
I don't think there's an validity for an argument that applies examples from history at a time when we are going through unprecedented technological changes. We are still at a relatively early stage of the digital error, and we are barely at the start of a new era in robotics.
His whole argument seems to extend from an age old egotism that nothing can replace the human mind.
Well I think he's totally wrong.
In the case of driving cars, the process can be replaced with a relatively simple autonoma i.e. don't run into anything.
The human mind is often so bored with this simple task that it finds other things to distract itself with, and that's how accidents occur most of the time i.e. whilst driving, people simply forget to drive the car.
If anything Googles program has proven that automated systems can make driving safer than human controlled systems
http://www.pcworld.com/article...
To me this guy seems to be making a statement which is purely opinion based, with no real technological or scientific methodology behind it and he is making this statement purely to gain publicity for himself
I have had a femur destroyed by a car driver. Of course this is not considered an act of violence -- simply that I (as a pedestrian or cyclist) got in the way of a car. I was in the "right" -- the driver received a charge of "careless driving". Nearly killed me and I still use a cane after three years.
I don't operate a motor vehicle; I don't even have a driving license.
My take on this? Cars *must* be completely automated. I do not trust drivers. My other "attitudes" to the car society? Inner-city speed limits set to STRICTLY 30 to 40 kilometers/hour.
No seatbelts or airbags allowed as "safety devices" -- these certainly do not help with cars killing pedestrians and allow the car operators to be mre reckless.
On car-only highways, I would certainly allow unlimited speeds -- I don't even care about "drinking and driving" on those highways.
In cities and shared paths? I really want automated cars.
Just another "Cubible(sic) Joe" 2 17 3061
just as dumb.
Look- I understand the need for an emergency- hard wired- kill switch.
I.e. a hard switch you can flip to turn off the vehicle.
I can even understand "not autonomous for now/the next few years" but anything else is just silly.
She was like chocolate when she drank... semi-sweet at first and then increasingly bitter.
For the same reason the military does not use (at at least doesn't admit to using) fully autonomous armed drones: software can't make judgement calls.
Yes, it does. They are called cruise missiles.
That's the stupidest thing I've every heard. Have you every heard of commercial airline service? Do you know how they fly in cloudy weather? Instruments!
What you meant to say, if you're not an idiot, is that "Google cars should be able to work on a reduced set of input to allow for redundancy."
Sorry to be so harsh, but - well, if you propose foolish things...
Is it just my observation, or are there way too many stupid people in the world?
Here's the thing - there's a finite chance that your seatbelt will trap you in your car as you burn to death or drown in a lake. In fact, that was the exact argument made against mandatory seatbelt laws.
And yet today pretty much everyone wears their seatbelts and we don't hear about people burning alive or drowning because they were trapped by a seatbelt and unable to escape. Instead, the survivability of crashes is greatly increased and vehicular deaths are on the decline even as road miles and number of cars in continuing to climb.
By the time autonomous cars are allowed to operate on their own, the safety benefits will far outweigh the complications.
Is it just my observation, or are there way too many stupid people in the world?
That's silly - the iPhone will never work properly with Google car.
Is it just my observation, or are there way too many stupid people in the world?
My source is the fact that all the waste in the US is stored in dozen different places where each one is far bigger than a foorball field:
e.g. https://en.wikipedia.org/wiki/...
or read this: http://www.nei.org/Issues-Poli... particular this: http://www.nei.org/CorporateSi...
The only thing giving your a small edge about your claim is that the above waste (first link) includes waste from weapon production and decommissioning.
FYI: http://www.wired.com/2015/07/p...
"All told, the nuclear reactors in the U.S. produce more than 2,000 metric tons of radioactive waste a year, according to the DoE"
From: http://www.scientificamerican....
I really wonder how pro nuclear advocates can be so uneducated that they not even know the basic facts.
In which desert do the USA store the biggest amount of nuclear waste in the word?
http://articles.latimes.com/ke...
Cost free eBook I read (by iBook/Kobo/Amazon/ObookO/Gutenberg etc.): "The Green Odyssey" by Philip Jose Farmer.
Robots driving on roads on which humans and OTHER robots are driving IS extreme robotics. Yes, if there were a transit "system", I'd agree that it is well known situation and not extreme robotics. But roads are transit chaos, not transit system.
Bingo Dictionary - Pragmatist, n. A myopic idealist.
If you are safely hurtling along an Interstate at 70 mph, in a car whose systems required literally millions of hours of engineering, on a road surface that was also heavily engineered and painstakingly built by hundreds of workers using millions of dollars worth of large equipment... then congratulations, you've already ceded control of life-critical operations to numerous big, opaque corporations.
You're trusting your life to GM, Toyota, and the manufacturers of all the cars around you, that they won't spin out of control and kill you. You're trusting the road designers and government regulators that the surface won't suddenly buckle or turn. You're also trusting all your fellow bozos on the road not to be drunk, sleepy, texting or spilling coffee in their laps. Looking at the odds of what causes road deaths, I'm not that unhappy about extending further trust of my life to the relatively capable hands of auto company engineers.
You shall see a cow on the roof of a cotton house.
if everyone on the road stopped when they didn't know wtf they were doing, we'd have a lot fewer dead people. the real problem is that people are asshats, so they speed up and drive worse when they're confused or upset. it may not be as emotionally satisfying to stop and think, but it's actually the most sensible thing to do.
usually not until after they've killed someone
well, hopefully we'll get to a point where the idiot who didn't stop will pay for the treatment and never be allowed to drive a car again
This whole argument is a fallacy based on absolute ignorance of *actual* human behavior, history, and even basic linguistics.
Ok. Let's see point by point.
>Children have had calculators for *decades*, and still learn how to do math in their heads. (In fact, students today are actually being expressly taught the methods that work best for doing math in your head, instead of having to stumble across the methods on their own. That accounts for about 90% of the 'Common Core suxxors' arguments you see out there.)
Common core teaches children a method to understand what a math problem means. Only problem is that it is basically a "let's count on our fingers" method and it has not been shown to produce an adult with a good working ability with math. It probably won't make a difference seeing that they will all have a calculating device in their pocket. (Odds are they are one EMP or lack of a charge station from helplessness.)
California has been allowing children to use calculators in class as early as 4th grade which misses the whole point of math classes in school. The point IS NOT to get the correct answer. The point is to pound basic mathematics (including the time tables) into their heads so they have a chance of remembering how to do the math when they are 40.
People have had contact lists for *centuries*, and still remember names, addresses, phone numbers, birthdays, etc. Storing the list electronically is no different than storing it on paper, except that if you lose your electronic device, you can probably recover all that information instead of having to painstakingly replace it over the course of the next few years.
Yeah, this argument is so obviously wrong. Twenty plus years ago the average adult had a minimum of 20 phone numbers in their head. Now you have "functioning adults" who don't remember their home phone number. If you have a smartphone, put it to the side and write down all the phone numbers that you can remember. You will prove my point.
Apparently you think that manipulating a clutch with your foot and a lever with your hand is what 'driving' means. Strangely enough, people have driven vehicles that lacked *both* of those for *centuries* before they were invented, and are *still* driving vehicles in which manual access to those controls isn't required.
You can poo-poo using a clutch but for starters: Name one single attribute of driving that is improved by using an automatic transmission.
[crickets]
Driving a clutch will teach a driver to manage the engine power which is required for any level of true car control. It allows you to shift the weight and the balance around so the driver is controlling it instead of simply reacting to it. If you don't understand what this all means I have to say, "Exactly. You are proving my point."
Driving like all arts is all about the subtle things. Probably the most subtle thing you can do driving is drive on ice or snow. If you have a chance (with your automatic) pull up to a complete stop where the surface is glazed by sliding tires and drips coming off cars. Now go back and as you get near put the car in neutral and see how much quicker it stops without skidding a tire. The difference at really low speeds is huge. (Now you can try to figure out WHY?)
Automatic transmissions reinforce a bunch of bad habits.
- Not coming to a complete halt at a stop sign is a common one (nope a little bit of roll isn't the same thing).
- Delayed acceleration of a second or two comes from the delay you get when you step on the gas. That "second or two" is a surprisingly big contributor to traffic congestion on large freeways.
Take a backup of 50 cars and add a 1 or more second delay to each one's acceleration. Yeah, it is a problem. The other bit is that it can influence a person's driving mindset to delay other seemingly unrelated driving behaviors. You will tend to see a correspo
Completely irrelevant and ignorant post. In the late 60s early 70s, we had over 50 000 accident related deaths. Now we have around 30 000. Per capita numbers are staggering, from around 25 deaths per 100 000 people, to about 10 deaths per 100 000. From 5 deaths per 100 million miles traveled to 1 death per 100 million miles traveled. Seems like all that new tech is paying off.
Sounds like David Mindell doesn't really understand the difference between 'Weak AI' and 'Strong AI'. If this was the basic technology of cars we were talking about here that would be like not understanding the difference between cars powered by engines and cars pulled by horses...
(Weak AI = Non-self-aware / non-sentient, Strong AI = Self-Aware / sentient.)
There are different problems between weak and strong AI. Strong is expected to be vastly superior in potential capabilities and driving ability and almost certainly much better at basic safety than weak. However Strong is still at least ten years away and certifying a Strong AI for cars is at least 15 years away.
Weak AI has the basic problem that it doesn't really understand the world or what people are. Weak AI also has the severe problem that as it becomes more sophisticated it reaches a point where the possibility of machines becoming spontaneously self-aware begins to rise exponentially. (At least 9 times out of 10 such a machine will immediately fail and crash - not good in a car.)
True Strong AI does not encounter this risk because it is always sentient right from the beginning. The big problem with Strong AI of course is that its no longer merely a machine, and as it requires a kind of homeostasis it can be argued that it is alive or even has a 'soul'. This creates problems with 'ownership', and how machines are treated. It also creates the problem that a strong AI can potentially choose to kill malevolently.
This is the real debate at the heart of future autonomous machines. Even with todays weak AI machines vision capabilities have improved hugely and have come a long way since say the Spirit or Opportunity Mars rovers - or since Apollo or the Ford model T. Computers today are about 10,000 times faster than those used 1n Apollo, about 10 billion times faster than the mechanical calculators used at the time of the first Ford.
Below the speed of light Special Relativity is one of the most accurate theories in physics - above the speed of light..
A situation where
A) there is not enough time for the car to stop completely or give adequete warning before releasing control, or
B) a "reasonably capable" system--one that most people feel comfortable giving full control to as they sleep--could not handle itself
is not going to be a situation a human driver can properly handle, even one who is above average and has been paying close attention and constantly prepared to take control. These are bleeding edge cases, like a falling boulder or earthquake.
Annnnd I think I'm actually agreeing with you, but I stopped reading too soon. So my reply is moot.
So driver automation caused the decrease in accident deaths? Really? Exactly what driver automations were instituted in that time period? (other then automated distractions)
I guess the massive improvements in energy absorbing crush structures didn't have that much to do with it. The pervasive requirement that occupants use seat belts didn't make much difference either. Passive restraints such as airbags, drastically improved seats, and car interiors specifically designed around energy absorption for occupant safety probably didn't make any difference. The MASSIVE improvement in tire safety and grip doesn't add up to much either then...
Your argument fails ridiculously and if you take a look at information from the NHTSA or the Insurance Institute for Highway Safety it points that out VERY clearly.
An interesting little auto safety tidbit feeding back to the original point that "automation of any human ability leads to humans losing that ability for themselves."
An independent analysis of the data gathered after 10 years of the 55 MPH national speed limit did not find anywhere near the number of lives saved that should have been expected from the change. (Especially when you start adding in the leaps and bounds that were made in crush structures that happened in that time period.)
One of the more accepted interpretations was that the reduction of speed had increased driver inattention and had basically bred a new generation of drivers that were less competent in general.
Use it or lose it.
More important than self-driving cars is to reduce the number of cars on the road. Invest in public transit of all kinds (bus, tram, light rail, regional rail, long distance rail, and yes, air travel as well unless we get serious about high speed rail) and a tremendously better Internet infrastructure. I could do my work from home without having to drive the (comparably measly) 8 miles to work...if I was offered a decent VPN with split tunneling and acceptable performance...something AT&T is incapable to deliver. I'd go for that because I hate driving. Until hell freezes over and the above mentioned happens, how about drastically improving driver education? Make it mandatory across the nation that drivers have to attend at least 20 hours of driving training including driving at night and on the highway plus a much better theoretical education that currently is practically non-existent. In other words, if we keep uneducated morons from driving around in their pick up trucks or doing motorcycle stunts during rush hour we might not only be safer overall, but autonomous vehicles would perform much better. The big problem for AV is other drivers.
The current ideas of what computers and automation can or should do, are bases on hundreds of years of human experiance with horses and slaves. That is not good because computers are quite different! 8-)
Many people's idea of what a car should do, came down from stories told by family. Tales of taking the carriage to the local pub, gettintg drunk, and climing back into the carriage so that the horse could take you home. Since the horse had lived there for most of it's life and knew quite well how to get home! 8-)
Computers just don't work that way. But people keep looking for a horse with no upkeep...
They'll never (in my lifetime at least) be fully autonomous like I am, because then they would be responsible for accidents and by that time we'd have to have courts and prisons for robots. They'll be autonomous at the "do what my owner says at a supervisory level". I expect what many cars will do when they're fine-detail autonomous is "go out and earn me money". It'll be the next BuyToLet. Wealthy people will buy self-driving cars that less wealthy people will rent. The cars won't need parking space - it'll be cheaper to drive around or 'hide' during quiet times. They'll book their own servicing and valet visits.
The cars won't be our overlords; the wealthy people who own our transport and accommodation will be. 'Carlords' competing with each other for our transport coin can only be a good thing - I can't see how they'll be able to monopolise supply like they have with accommodation.