Slashdot Mirror
Google Unveils Self-Driving Car With No Steering Wheel
cartechboy writes: "We've already discussed and maybe even come to terms with the fact that autonomous cars are coming. In fact, many automakers including Mercedes-Benz and Tesla have committed to self-driving cars by 2017. Apparently that's not ambitious enough. Google has just unveiled an in-house-designed, self-driving car prototype with no steering wheel or pedals. In fact, it doesn't have any traditional controls, not even a stereo. The as-yet-nameless car is a testbed for Google's vision of the computerized future of transportation. Currently the prototype does little more than programmed parking lot rides at a maximum of 25 mph, but Google plans to build about 100 prototypes, with the first examples receiving manual controls (human-operated). Google then plans to roll out the pilot program in California in the next several years. So the technology is now there, but is there really a market for a car that drives you without your input other than the destination?"
These this will naturally become shuttles and taxi services almost immediately. Given the protests of Uber and Lyft, what will the outcry be for these?
"Don't teach a man to fish, feed yourself. He's a grown man. Fishing's not that hard." - Ron Swanson
Sorry. While I love technology, my not-so-humble opinion is that we're nowhere near the level of reliability needed for a car that's completely free of manual control.
Simply put, having seen the arc of technology advance over the last 30+ years, I still don't trust an automated driver system with my safety. PERIOD.
Chas - The one, the only.
THANK GOD!!!
No thanks, wouldn't want a car that I can't manually override when shit happens.
Surely Google already knows where it is best for you to go. It knows everything else about you...
Sorry. While I love technology, my not-so-humble opinion is that we're nowhere near the level of reliability needed for a car that's completely free of manual control.
Simply put, having seen the arc of technology advance over the last 30+ years, I still don't trust an automated driver system with my safety. PERIOD.
Millions of people fly in airplanes every day that rely on computer controls (since there is no mechanical linkage between the pilot and the control surfaces). And 30,000 people die each year at the hands of human drivers.
While the real time image recognition may not be quite ready for prime time, it will get there and when it does, computer drivers will be safer than human drivers. Google's driverless cars have already racked up 700,000 accident free miles in autonomous mode (albeit with a human ready to take over). Their car has already surpassed my own record, it's only been about 150,000 miles since my last accident (a car changed lanes into me, while the accident was not my fault, if I'd had computer-like reflexes and perfect awareness of my surroundings to know that the lane beside me was open, I may have been able to avoid the accident by sudden braking and/or making a quick lane change)
Sorry. While I love technology, my not-so-humble opinion is that we're nowhere near the level of reliability needed for a car that's completely free of manual control.
Simply put, having seen the arc of technology advance over the last 30+ years, I still don't trust an automated driver system with my safety. PERIOD.
Sorry. While I love humans, my not-so-humble opinion is that we're nowhere near the level of reliability needed for a car that's manual control.
Simply put, having seen the arc of traffic fatalities advance over the last 30+ years, I still don't trust a human driver system with my safety. PERIOD.
Cars are not safe: people will die. I'd rather have shitty AI that we can iterate on and improve every time it kills someone than having to start with fresh teenagers each time. An AI can learn from millions of cars, and not miss the learning opportunity of fatal crashes. Also, people have really bad sensors for driving compared to what an AI can use. Maybe its not better than good drivers yet, but I'd prefer a shitty AI that we can iterate on to people who barley manage to pass a driving test on the third try driving in the dark while distracted, and we let people do that... Compared to a person, such an AI could be a lot better at refusing to drive in unsafe conditions (it won't give into rage or peer pressure and do something stupid). That might be annoying, but having a car that can pick you up by itself might counter that out.
Millions of people fly in airplanes every day that rely on computer controls (since there is no mechanical linkage between the pilot and the control surfaces).
If that's what counts as 'computer control,' then we already have computer control today. There are plenty of computer systems in cars, and some won't even start without going through a computer system.
"First they came for the slanderers and i said nothing."
what kind of hells is that???
Donald 'Duck' Dunn: We had a band powerful enough to turn goat piss into gasoline.
A quick call to google's helpdesk is all that's needed to stop the car in an emergency.
If you drive on the same streets that I do, you trust me with your safety. As my driving skills are below median, this should be a lot more worrying to you then the prospect of being in a computer-driven car. (Fortunately for you, surveys show that below-median drivers are rare.)
Quattuor res in hoc mundo sanctae sunt: libri, liberi, libertas et liberalitas.
And this is an excellent argument against the "it will always need manual controls in case of failure" argument. Modern vehicles have fly-by-wire accelerator, brakes, gears, etc.
The driver isn't in direct physical control of the vehicle and hasn't been for some time. Progress towards fully autonomous vehicles is a matter of degree, not of kind.
Rampant carbon sequestration destroyed the Dinosaurs' tropical paradise. I'm here to help repair the damage.
Sorry, but there's a big difference between flying around in a plane in a pre-planned course that's been cleared of other traffic and driving around on the ground on an expressway or city street.
Chas - The one, the only.
THANK GOD!!!
10/10, would buy.
Automated cars are already better than people. The trains in Canada have been automated for decades and they're fine. The Google fleet drove across the US several times, something most human drivers would probably screw up at some point.
The only thing I dislike is the fact that I love my car and I can't think of a way to convert it economically. Otherwise I would, without hesitation. Including removing the steering wheel and pedals.
I don't want to drive it. I want auto-driving cars and I want them now.
Check out my sci-fi book "Lacuna" at http://goo.gl/MVxX8
A car which automatically takes me places I don't want to go, based on my browsing history.
Worst. Signature. Ever.
The Google car has done something like 700,000 miles and crashed twice. Both times this occurred, it was under control of the human occupant.
I drive to work every morning and the number of times I see people not paying attention is extraordinary. Women doing their makeup, people texting, trying to argue with their children etc.
Honestly, in my view, removing the steering wheel is a safety feature.
And if you think your judgement and perception is better than this computer system, you are full of hubris and a menace to other road users. It works both ways.
In one sense, we've been a fully autonomous vehicle ever since we started walking. ;)
Rampant carbon sequestration destroyed the Dinosaurs' tropical paradise. I'm here to help repair the damage.
The problem is that you confuse manual controls and manual control inputs. What type of control a person or computer uses is irrelevant. What is relevant is the mind making the decisions about what inputs to make. Computers are yet to be sophisticated to handle many situations as well has humans do.
The driver isn't in direct physical control of the vehicle and hasn't been for some time. Progress towards fully autonomous vehicles is a matter of degree,
Completely false. Whether it is fly by wire or cables the inputs are still made by humans and that is the important part.
The brain is the important part of the machine and not the nerves. Computer brains are not up to the task yet.
It's not about judgement. It's about abilities.
A self driving car can simultaneously look in every direction around the car and never have to blink. If an object is detected and the car needs to stop, it takes a person time to physically lift their foot from one pedal and press the other(s). Not much time, sure, but in a sudden stop scenario, every little bit helps.
Humans have much better non-linear thinking. We can navigate dirt roads, or unmapped territory. But for day to day commuting on established roads, automation is the way to go. Computers never get sleepy, they don't get distracted and they can be programmed to obey speed limits. Google's test vehicle is already well above the safety record of an average driver, with nearly half a million miles, safe and sound.
And that's just the prototype.
This signature is false.
computers are also:
1. hackable.. one bored 16yo with a laptop on an overpass + 20000 wirelessly networked cars on a highway = fun.. oh and state mandated kill switches are only there for the children, right?
2. careless about self preservation. A computer will happily cause an accident due to a programming bug or sensor fail. Was that a rock or a plastic bag? A human can tell, but your computer? doubtful. How about that truck carrying those huge metal pipes? Is that top pipe about to fall off the back and flatten your car? A human can tell, a computer? hell no. A kid decides to step into the road at the last moment, you saw him running down the driveway from behind a line of parked cars on the side of the road. Did the computer? of course not. Did it see that huge pot hole? nope.
3. Only as good as the programmer that programmed it. Yes, a half million carefully selected miles with a person who knows the details of the system being tested in the vehicle..oh and he has an override capability that the rest of us will lack. Speed is 'a' factor, but not always 'the' factor in accidents, although the cops love to say it is (to help justify keeping the limits lower so they can rake in more cash). If that aforementioned truck is about to jackknife, it might actually be more prudent to get out from around it which will probably mean violating the speed limit. It just depends on what your options are at that moment. It would also take a human to see the fact the truck is in trouble. All the computer can do is react by hitting the brakes after it detects the truck blocking the way... gl with that, even if it was following a 'safe' distance.
As chaotic and inconsistent as humans can be, I think we're better off fixing the newfound inattentiveness while driving than trying to shoehorn more complacency and dependence into his life. Owning your mobility is an important distinction between those who are free and those who only go where they are told...and you can't tell me that governments won't want backdoor access to this.
Just plug your gamepad into the USB port.
You can ask the same for normal cars, as they're basically driving computers these days.
And if you think your judgement and perception is better than this computer system, you are full of hubris and a menace to other road users. It works both ways.
Whatever. My driving skills (or lack thereof) are a known quantity to me. I have some grasp of what I can and cannot do in a vehicle.
I think that's unlikely, at least for most drivers. How many times have you experienced an emergency stop from 70mph? Or practiced regaining control from a skid? Or when sliding on ice, or aquaplaning? Most drivers will have no idea how their car behaves in those situations and have no idea how good their skills are because they've never been tested in those circumstances, or have only tested them once or twice. One would expect that a self-driving car's abilities will have been tested much more.
And we know that the Google cars' safety record is that of an exemplary driver. And no - your driving skills are *not* a known quantity to you - you are assuming they are, but as you are human that is patently not the case. You even admit it yourself - you have "some grasp" - not an entire, thorough understanding of everything you can and can not do (unlike the Google cars). Infinitely more people have driven themselves off cliffs and into lakes than Google cars have - why you'd assume they'd do such a thing is beyond me. Oh, no - wait - arrogance.
So you have it entirely backwards - you are the unknown quantity, and the Google cars are the known quantities. Your hubris seems to have swapped those round for you. Again: you are the problem, and every complaint you raise against being labelled as such, so far, has shown it for the truth it is.
You don't suppose, do you, that the "Stop button" you mentioned might not be a way to tell the car to, well, "stop moving at any point", do you?
"I do not agree with what you say, but I will defend to the death your right to say it"
This kid can break military-grade encryption? If that's the case we've got far bigger problems on our hands.
So you'd run over a rock in a plastic bag because you thought it was a plastic bag, whereas the radar on the driverless cars would have seen through the plastic bag and seen the rock. The pipe on the back of the truck? Well, the car would keep a safe distance, enough for it to avoid any falling object in front of it. That's what humans should be doing anyway. The cars' LIDAR scans for objects approaching the road, and can do so far better than any human can, so your kid-running-into-the-road situation would work out worse with a human behind the wheel. The LIDAR can see farther, with more accuracy, and in 360-degrees. You can't.
The rest of your post is ill-though-out guesswork ascribing idiocy and incompetence to the development team. They are experts in this field - you are not. You spend a lot of your time on Slashdot, being racist and sexist. I wonder who's more trustworthy when it comes to logical appraisal? You've demonstrated you are a slave to gut instincts and untrusting of data which might change your world-view, so no-one in their right minds should be listening to you.
If you take the set of people who might be willing to buy a self-driving car (a set underrepresented on /.), few of them are going to want to do it if they're on the hook for whatever the car does. If that's the case, you might as well drive yourself. Google doesn't want that, either, and just put out a statement to that effect. My guess is that they're going to try to get the relevant laws changed, but, in the meantime, what better way to protect your users from liability than to make it impossible for them to have had any control of the vehicle?
Stop learning! Only you can prevent esoterrorism.
Just look at this: https://www.youtube.com/watch?... (look ma no hands)
While the A380 is capable of autolands on properly equipped runways (which by the way
takes quite a bit of work by the pilots to set up), this video doesn't show one.
Final approach and landing are in fact flown in full manual mode. It's Lufthansa policy
to land manually whenever possible, as to not lose manual flying proficiency.
Other airlines do have other policies, but I doubt any use autolands routinely -
as I said, they are more work.
The next article on this car will likely be on how it can be hacked (because everything gets such an article).
Then that manual stop button might come in handy.
Joke ---> Whooosh
You
Carefully crafted sig.
Human level AI? I would rather not, most human drivers are bad. Try to use the cruise control on a crowded highway and you know what I mean. Just maintaining uniform speed is an impossible task for most drivers. Oh there is a minor bend in the road, I need to slow down, there is a minor hill, I need to slow down, look an accident on the other side of the road, I need to slow down.
In almost all cases a computer will be able to react way faster and with more precision than a human can. Yes there will be some minor flaws in the first systems, they will be systemic and patched out. With human drivers it is a gamble each time they get on the road.
Speak for yourself! The public transport in the city in which I live is wonderful, as it frequently is across the country where I live (Germany). You just need to be relatively near to a tram stop (which is the vast, vast majority of people, even those in the styx), and you can get wherever you want quickly, efficiently, and cheap. Great connections to the high-speed rail means you can be anywhere very quickly with minimal fuss, including neighbouring countries. It's awesome to leave your house, walk 20m to the tram stop, take a tram to the train station, then ride at 186mp/h+ to your destination while sat in the diner car drinking awesome beer and eating suspiciously-delicious food. Don't assume all public transport is broken just because your locality can't manage it.
So the technology is now there, but is there really a market for a car that drives you without your input other than the destination?
I think the summary has this backwards. Of course there is a market for a vehicle (let's not call it a car for the moment) that drives you around without your input, think of buses, trains, planes, taxis. If the price is right, it will definitely be a success - it doesn't really need to compete with cars to be useful, although it seems likely that many of those who think of their car as an expensive annoyance they have to have to get around would be interested.
But the thing is that this is still a prototype. The technologi is in fact not there yet - it may be in a couple of years, but we don't know yet.
IMHO the prototype makes sense as a statement and as a challenge. With no steering wheel, there's no 99% self-driving non-sense - they have to have a plan for all corner cases, even if that's something like car stops and is remote-controlled around obstacle.
I'm afraid your assertion is quite false - about 90% of all landings done daily by large civil aircraft (737 upward) is done by the autoland system, with the only requirement for a manual landing being to retain certification for the pilot.
> "I don't believe..."
That's not really a counter argument.
- Chuq
I called google support for my nexus 4. The result was a very kind lady who solved my problem quickly.
Google's ultimate goal isn't to build self-driving cars. Their goal is to get the technology off the ground and convince other auto manufacturers that it is possible so that people can browse the web on their way to work, and look at ads on the Google search engine. They didn't build a phone operating system because they thought they could do a better job than Apple or because Apple was doing a bad job. They did it to built up market share and mind share around their search engine. Google wants us to have self driving cars so that we have more leisure time in order to see more ads, or possibly to buy apps, games, books, and movies off their app store. This is where the real money is. Not in selling physical objects that take money to produce, but in generating revenue from a product that costs nothing to produce, and can generate obscene amounts of money.
Anthropic principle: We see the universe the way it is because if it were different we would not be here to see it.
1. That goes for manually driven cars too. If it has a kill switch your hacker could just as easily kill the power at a "fun" time, say when you've just pulled round to overtake a truck with what you thought was enough time to pull back in in time to avoid the oncoming. And what kind of idiot would expose the AI to the internet, least of all via wireless? Designed properly it would be a little black box receiving carefully filtered inputs from the sensors, providing carefully filtered outputs to the actuators.
2. Citation needed. People get squashed by falling loads which suggests that they aren't good at spotting them reliably. An AI? Don't know. But they would spot the child you referred to, and be tracking its movement and capable of stopping in time. (The radar is located higher up than your eyes, has all-round coverage and is specifically feeding algorithms looking for child- (or person-)sized moving objects, among other things. And while I don't know for sure about the pothole if the radar is any good it ought to detect an anomaly - which it can be trained to avoid. Best of all it has a better idea of where its wheels are running than most drivers I've seen, so the avoiding action could be less drastic (ride over the pothole not around it into the oncoming traffic).
3. Citation needed. The whole point of machine learning is that it can improve on itself. The codebase can be relatively small and very rugged, the learned behaviour can be more complex. Also, with the kind of high resolution radar google is using there should be no problem detecting pre-jackknife behaviour in your hypothetical truck.
Pretty much all of your post is unsubstantiated rhetoric. If Google turned up at my house with a prototype driverless car and a legal indemnity would I get in, let it drive me to work and have a snooze as it did so? Probably not today. But it's not like they're offering the public these today. It's another step along the testing roadmap. My guess, having followed developments in this area since working at a company doing early lane guidance and collision avoidance work in the mid 90s, is we'll see these cars, or their successors, on general sale in around 2020, in high end vehicle lines initially (think executive class motors like Merc E class, BMW 7 series, Jag XJ) and then in the consumer market around 5 years later, with the technology becoming pretty much universal in new cars by around 2030. It's not a question of "no steering wheel, no deal" - if you don't make the deal, others will so you'll still be relying on the safety of these computer controlled vehicles for your safety on the roads, just like you rely on the skill of other drivers at the moment.
To be provocative, I wouldn't be surprised if human driving of cars on public roads was outlawed by 2050, with old manual vehicles either requiring an AI / servocontrol retrofit or being restricted to track use. So eventually, if you live that long, you'll be required to make the deal or stay off the roads.
I've called Google more than once due to errors with listed phone numbers. (The phone number listed when you searched for one of my company's departments was a different department. You can imagine how annoying the flood of wrong "I looked it up online" numbers was.) The people I spoke with were very helpful and knowledgeable. I'll admit that finding a support number for Google isn't easy, but it can be done.
My sci-fi novel, Ghost Thief, is now available from Amazon.com.
I'm surprised people are talking about whether this should or shouldn't happen. It will. It's a given. It's only a matter of years.
US Politics is often driven by the reliable senior vote, and legal self-driving cars will quickly become legally available in Florida and other states with large retirement populations, and eventually in all states. Google knows this and is already working that angle, which is why the video featured the elderly and the blind.
Entire job categories will soon be wiped out. Taxi drivers will go first, followed by truckers. Other industries are less obvious, but will be impacted - some positively, some negatively. This will affect suburban moms driving their kids everywhere, pharmacies (prescription delivery), grocery stores (more delivery options), and even Home Depot (contractors spend hours driving to and from Home Depot and waiting in line).
Restaurants will be changed. More will offer delivery, and those that don't will let you send your car for pickup. When people do go to bars and restaurants, they will drink more alcohol than before, and may head out more, because they no longer have to worry about driving drunk.
The average Joe will have the transportation resources of what used to be available only to someone so rich as to have a pool of chauffeurs working 24/7.
Police forces will lose speeding ticket revenue, towns and cities will lose parking revenue. Parking garages in medium size towns will empty, and be replaced with other real estate. Commutes to work may lengthen, as people are more willing to endure a long commute if they can work in their car, or sleep.
Every car maker is going to start losing market share to Google unless they offer this. Google's head start combined with their better engineers means that Google might become the dominant software provider to cars. Unless car manufacturers come up with software that is equally safe, Google will soon start making more money per car sale than the manufacturers.
Security and policing will change. Suicide bombers can be replaced by software hackers. Drug smuggling within countries will change. Security guards will be replaced, in part, by cars with video cameras. Warfare will change, with drone cars becoming as ubiquitous as drone planes.
It might take 5 years for these cars to be road legal for non-licensed adults (like the elderly who can't drive). But once that happens, within 10 years all new cars sold will have this ability.
How do autopilots in airplanes work? They work on the assumption that they have a clear path along their assigned course based on the flight plan and when that assumption is incorrect (assuming the potential obstruction has TCAS), alarms start going off to prompt pilots to do something to avoid the probable crash. As long as the pilots and ATC do their jobs right, most collision avoidance is taken care of before the plane even lifts off.
This is very different from driving on the streets where there are no "flight plans", no authority managing space reservations along streets, tons of other cars, cyclists, pedestrians, intersections, etc. all over the place that require continuous on-the-spot adjustments to immediate surroundings instead of only worrying about probable obstacles 50km or 2-3mins ahead. Driving a car is a lot more involved than just setting a speed, direction and altitude to the next way-point and let the computer worry about maintaining those parameters between way-points.
Autopilots on planes are basically cruise-control on steroids: about equivalent to slapping lane-sensing cameras and front/rear radar/sonar to measure distances on a car and letting cruise-control handle speed, holding lane position and maintaining safe distances within that lane; basically, take most of the mundane part of flights or highway driving off the pilots' or drivers' hands - if they choose to.
Designing a car autopilot to handle staying in a lane for 300km of highway is a much simpler challenge than designing a car autopilot that can cope with urban and downtown neighborhoods where obstacles can pop out from anywhere with little to no warning.
I'm afraid your assertion is quite false - about 90% of all landings done daily by large civil aircraft (737 upward) is done by the autoland system, with the only requirement for a manual landing being to retain certification for the pilot.
Using ILS, I totally believe. Full autoland, i.e. flare, touchdown, rollout: I'd like to see a very good source.
Considering that autoland requires that the runway be equipped with ILS CAT III(b), this seems unlikely: China has one FAA approved CAT III runway, Hong Kong 25R.
There are none in Singapore, none in Thailand, there's one in Australia (Melbourne 16), three in India (all the same airport though, Delhi).
Of the 1369 ILS-equipped runways in the US (Excel warning), just 113 have CAT III (no idea whether those are level a or b).
Sure, most of them are at the biggest an busiest airports, but considering that an autolanding
plane severely limits a runway's capacity due to increased spacing requirements, I doubt ATC
would be too happy to accomodate lots of autolands especially on those.
They just don't have the timeslots.
[citation needed]
IAAPilot, though not of the big ones. My understanding is that this is totally false. Autoland (ILS CATIII) requires a specially equipped runway, airplane, and crew (training) and each of these must be kept certified to do it as well. By no means all runways, airports, and crews are certified to do this - in the case of runways, most are not, even of those that the commercial operators fly to.
It's true that many approaches are done automatically, but an approach is most definitely NOT a landing. A dinky little Cessna may have an autopilot sophisticated enough to follow a glideslope down, but it'll disconnect at the DH just like a 787 will (without a CATIII ILS). The last several hundred feet (of altitude) are almost always flown manually, and to be frank, that's the hard part.
Where did this meme come from? Airline pilots most definitely do still fly...
I have developed a truly marvelous proof of this comment, which this signature is too narrow to contain.
A quick post on google's support group is all that's needed to stop the car in an emergency.
FTFY
The final production version will feature a single "I'm Feeling Lucky" button.
At least put in a CD / CD changer.
People still use CDs?
Eagles may soar, but weasels don't get sucked into jet engines.