Lyft CEO: Self-Driving Cars Aren't the Future
Nerval's Lobster writes Google, Tesla, Mercedes and others are working hard to build the best self-driving car. But will anyone actually buy them? In a Q&A session at this year's South by Southwest, Lyft CEO Logan Green insisted the answer is "No." But does Green truly believe in this vision, or is he driven (so to speak) by other motivations? It's possible that Green's stance on self-driving cars has to do more with Uber's decision to aggressively fund research into that technology. Uber CEO Travis Kalanick announcing that self-driving cars were the future was something that greatly upset many Uber drivers, and Green may see that spasm of anger as an opportunity to differentiate Lyft in the hearts and minds of the drivers who work for his service. Whether or not Green's vision is genuine, we won't know the outcome for several more years, considering the probable timeframes before self-driving cars hit the road... if ever.
Buggy whip makers said automobiles aren't the future.
if the CEO of Uber and the CEO of Lyft both shut the fuck up. These two dimwits never have anything important or interesting to say. These two companies developed an easy way to find a ride share; they didn't invent the damn car. Like Mercedes, Google, or Tesla care what Logan Green thinks.
If Uber is right, and self driving cars are the future - the Uber drivers today can buy one, and send it off driving for Uber while they sit at home instead of driving. Or more likely they can sit inside reading, just to be present as a chaperone for the car in order to prevent their property from being part of a dramatic re-enactment of the Johnny Cab scene in Running Man...
Either way, less work.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
And what does he care about the outcome? A taxi service made up of drivers and a taxi service made up of self-driving cars would still do the same job, get people from points A to B (and any points A.1, A.2, A.3, etc inbetween) and charge them for it.
Uber was just raided for harboring Skynet, after all. Sure, at first it's just staying inside lanes and stopping before you rear-end something; but after the evil-bit flips, the roadways look like Carmageddon minus the reset button.
Corruption is convincing someone that the selfless ideal is the same as their selfish ideal.
self fighting wars. Hell we don't even need a country, just set out bots to kill people and call that a war.
A properly taught ant would probably be better at driving than about 80% of human drivers.
Have you been out on the road? With all the idiots making cellphone calls/driving drunk/texting/facebooking/putting up makeup/whatever? Have you seen the current state of affairs on the road?
My prophecy:
Self driving cars will be better than humans in 99,99% of the situations within 3 years. Sadly it will take another 5 to wide scale adoption and yet another 10 years for human driven cars to be banned to racetracks.
Well, I might have a way, but it only works on a semi spherical planet in a vacuum.
Well... I think what they'll probably do is continue testing - and they probably won't be widely deployed until they're as safe as human drivers (on average, they'll probably be safer in some ways and less safe in others). Soon after that, they'll be safer than humans (because they can share knowledge, are easy to upgrade, and once there's lots of them they'll be able to communicate in ways humans can't).. well, that is, if we keep going.
I say "if", because the more likely problem is Luddites who will want them banned after the first death, even if their overall safety record is better than humans. An enormous number of extra people will die because of how slowly we'll adopt self-driving cars. This is because people are dumb and ruled by emotional reactions: when people cause collisions (which they do thousands of times a day) it's just an accident, but the first time a self-driving car runs over a kid it's going to be pandemonium - and a good percentage of people will want to go back to the old higher death rates.
As to your argument, it's difficult to compare a computer to an ant or a person on some single scale of intelligence. An ant is very good at some things, but completely incapable at most everything else. Computers exceed humans at many tasks, while lagging behind in others. No computer today could learn how to drive well by itself, or have much conception of what driving is - but we've demonstrated that computers, designed and refined over time by people, can get very good at complex tasks. I think we're still a ways off from having safe computer drivers, but it's not in any way impossible or far distant; computers are already much closer to "humans" than "ants" on the "ability to drive" standard, and there's no reason they couldn't be better than humans at driving within the next 10-20 years.
Let's not stir that bag of worms...
Not a private person, no. Self-driving cars will be bought by public transportation companies to build up fleets of driverless cars.
Are self driving cars the future of transportation? Yes. The only question is how far in the future. Current technology can do many of the tasks of driving a car very well. Lane following is a good example of that. They also do many tasks very poorly. Differentiating someone waving hello from someone trying to warn you of danger is an example. To do the more complex tasks requires great leaps in AI. When will these advances happen? I think it will take at least a few decades. Ever heard of the 80/20 rule? In this case 80% of the challenges can be solved with 20% of the effort. That last 20% is going to take 80% of the effort.
self driving cars? You'll have to take the steering wheel away from my cold, dead palms...iceholes!
:) just wait till terrorist nerds are programming self driving cars to cause accidents on purpose. It will be fun like battlebots!
I'll tell you what the future is, and the CEOs of Lyft, Uber, etc. know it as well:
Self-driving car-sharing vehicles.
I'm a huge fan of the new car-sharing services that have popped up in recent years. The ones where you simply pick up a car wherever you find it (your iPhone App will show you the nearest ones if you are looking), drive to where you want, and leave it there for the next person to take.
You have a car when you need it, don't need to bother with it when you don't, you don't need to worry about fuel, inspections, washing it - nothing. And you can take the car you need for today. Good weather? Cabrio. Need to transport something? Bigger trunk. etc.
Main disadvantage? Sometimes there's no car nearby, and of course the usual parking space hunt in the city.
Solution: Self-driving cars. Tap a button on your smartphone, the nearest car comes and picks you up. Just exit it at destination and it'll go away by itself, either finding a parking space or going to the next person who called one. If it's an electric car, it can also go and find a charging station if it wants.
Who needs taxis? Who needs Lyft?
They know this, of course, and they know it's coming.
Assorted stuff I do sometimes: Lemuria.org
One problem is who will bear the responsibility for the accident. As it is now, if I hit another car or a person, I am responsible for it and am punished for it. This is fair. Now, if a self-driving car hits someone and the car was maintained properly (no problems with brakes etc), then the accident was really caused by the software. Will the manufacturer pay for the damage caused? Because now I am not responsible for something I had no control over.
Another problem is that when a self-driving car causes an accident, it will be something that would have been obvious to a human driver, resulting in the general opinion that self driving cars are stupid.
As for me, I love my car made in 1982 with no computers in it (yes, it has a carburetor and mechanical controls) and would not want to replace it with a car that has software. Not as much for the software by itself, but the possibilities of lock-in, DRM, hacking and so on.
Odd, i was thinking the opposite way. Have some selfdriving cars be modified a bit extra by the police (with some proper calibration of everything, and some extra sensors), and you've got near permanent police presence on every big road. And it's no longer just people driving to fast past a speed trap that get caught, but all other antisocial behaviors too. It'll make the roads hell for human drivers who are used to being jerks on the road :).
My prophecy:
Self driving cars will be better than humans in 99,99% of the situations within 3 years. Sadly it will take another 5 to wide scale adoption and yet another 10 years for human driven cars to be banned to racetracks.
Irrelevant - the moment a self-driving car has an accident that causes loss of life, there'll be a public outcry against them, and a demand that they be banned for being too unsafe, statistics-be-damned.
Yes, Total Recall is right! Thanks for the correction.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
One problem is who will bear the responsibility for the accident.
We already know that. The insurance company is responsible.
Three states (California, Nevada, and Florida) already allow SDCs on their roads. The liability issue is already resolved.
Another problem is that when a self-driving car causes an accident, it will be something that would have been obvious to a human driver, resulting in the general opinion that self driving cars are stupid.
When a human causes an accident, it is almost always something that would have been obvious to another human driver, resulting in the general opinion that humans are stupid. Most people are intelligent enough to realize that SDCs will not be perfect, but they will prevent a lot more accidents that they will cause, and the tradeoff is worth it. Seat-belts and airbags also occasionally kill people, but they save ten lives for every one they take, and people accept the tradeoff.
Sorry, I am talking from a Dutch point of view. 18+ year old cars are rare here and most of them are oldtimers, only to be driven a few times a year. That argument will not stop it here.
Normal human idiocy like banning them because of one death (despite statistics) might, but I like having high hopes.
Well, I might have a way, but it only works on a semi spherical planet in a vacuum.
I think that autonomous vehicles will come and go, but they'll be around almost as long as cars with drivers. I'd bet that in the long-long term, urban planning will change such that cars become entirely unnecessary in all but the most remote places. I don't think that we'll ever become so densely populated that the world is one big city, but I'll bet that we'll see large high-rise condos become much much more common, and then it'll be a ride down an elevator to do your shopping and a walk or train ride to school.
It's not that suburbia isn't awesome. It's just the direction I kind of envision things going in. I could be wrong. This sort of radical shift in urban planning would take centuries, to take hold in the west.
Analogies aren't exact. If they were, they wouldn't be analogies, they'd be the things.
The pattern is "Person in business X says Y is shit", where Y is a competitor of X.
Confucius say, "Find worm in apple - bad. Find half a worm - worse."
I for instance would buy one as soon as they are affordable.
and software doesn't have it.
Really. I can't believe that all these nerds like to pretend that their toys are actually thinking. They're not. And "self driving cars" won't know that they're driving, won't know what a human is, won't know what a horse is, won't know what ANY OF THE THINGS IN THE ENVIRONMENT ARE. They won't recognize when trillion of possible conditions are strange.
You want something totally insentient DRIVING A CAR?
Are you all insentient yourselves?
Lyft CEO: Self-Driving Cars Aren't the Future
"...I hope."
systemd is Roko's Basilisk.
life and death situations. I'm a big fan of A.I. including "strong AI" (means actually sentient) a big enough fan that I know that we're no\where near having competent general A.I. let alone some incompetent version of strong A.I.
Only sentient beings should be allowed to drive. Period.
That's a fact about "self driving vehicles". Now what is supposed to make them worth the risk to every person in society who might be near a road sometime?
I'm really sorry to have to be so direct but that is the dumbest argument I've ever heard.
Driving is not a matter of intelligence. It's a matter of reaction. Sure, intelligence and experience help you anticipate when other drivers are being idiots, but there is very little involved in driving that can not be compensated by reaction time and adhering to proper distances.
The biggest hurdle to take is to correctly measure the surroundings. If you did that via image recognition, then yes, AI would be important. But there is laser, radar, GPS and so many other sensors involved that do nothing more than note distances to targets, location on road etc.
Autonomous driving certainly isn't trivial, but the other thing you have to keep in mind is that your oh so intelligent human drivers are actually driving like morons a lot of the time.
Please stop putting the bar for autonomous driving so high the systems have to practically be perfect the be viable. The moment they are twice as good as a human should be the moment we start switching. And we're not far from that.
Remember how badly the average driver actually drives. And then remember that half of them drive worse than that.
Add on top of that networked driving, where cars coordinate over several hundred meters and you'll see so much potential gain even from non-perfect systems it's staggering.
I doubt it. Because that would be poor software design. The software would be looking to avoid accidents in an order of severity. Your pushing off the road at an exist may work but your pushing to increase speed wouldn't.
In most vehicle brakes far exceed the power of the engine. Even in cases where it doesn't the brakes have twice as much contact patch with the road than the engine (unless you are in a 4wd with dual diff locks / limited slip diff). The most sensible and safest tactic is to not accelerate inside the box past a predefined safe speed. So at say 10% over the speed limit the car will refuse to go faster irrespective of how close the rear vehicle comes. Then if the 'tard behind does actually ram or nudge the vehicle the car would then begin to apply the brakes gradually until the vehicle is stopped.
The computer would be able to counteract for rotation as it would have independent control of the brakes for each wheel allowing it to bias front / rear / left / right as required.
How about if we make sure that machines can do ordinary tasks before we put them in a position of endangering everyone who goes near a road?
Sure we'll have intelligent machines one day. So we're suppose to have machines driving vehicles some 80 years before they're smart? What idiot thought THAT was a good idea?
The moment something in reality changes, the map is wrong.
The moment there's radio interference, the GPS doesn't work.
The moment one of 10^20 things you know about but a machine wouldn't changes, you'd know that something was wrong but a machine wouldn't. Is it really that hard to see that having machine driving before machine intelligence is idiotic?
There's a lot of filtering to hide just how inaccurate GPSes are. That doesn't matter when a human is looking at a screen and ignoring it when it's obviously wrong. But it makes a huge difference when an idiotic machine is driving.
What's wrong with hiring drivers?
Self driving cars operated by the owner are a different situation from that of self driving taxis. The owner and rider interests diverge when those are different people. A self driving taxi has to protect itself from theft & abuse and protect its owner from lawsuits. That mean the person riding in the taxi won't be allowed to arbitrarily stop it, assume manual control, or exit in locations considered by the taxi owner to be unsafe. Putting my car in self driving mode with my average speed 10 mph in gridlock sounds attractive. Getting into a little vehicle capable of traveling to arbitrary locations and trusting it like I would a train takes the early adopter impulse right out of me. Maybe self driving buses would make the transition better.
I'd follow your own advice, and I'm be more courteous as well but that's mainly because I don't like looking like a keyboard warrior.
Nothing you said in any way highlighted a short coming of a automated car. You made a few unsubstantiated remarks about machines being 'moronic' etc. Personally when I look at the behaviour of many road users, and too many internet posters, it certainly seems like flesh-bag morons are pretty common already!
Do you really think so little of thought that it never occurs to you that it's important?
Oh, I think a lot about thought, but I also walk a lot and get to see human drivers from the view of a pedestrian.
I encounter situations on a weekly bases where I know that had it been a child or an animal in my position then it would have been dead.
It is small things. A driver looks the wrong way at an intersection, perhaps he just blinks at the wrong time, and quite recently he just checks his phone.
I don't get ran over because I realize that the driver isn't attentive or doesn't care about my right of way.
The reason the kid on the bike doesn't get ran over isn't because the drivers skill.
would make an awesome way to kidnap.
Irrelevant - the moment a self-driving car has an accident that causes loss of life, there'll be a public outcry against them, and a demand that they be banned for being too unsafe, statistics-be-damned.
What is sad is that you're probably right and it shows how ignorant most humans really are...
We are not, in general, as smart as we think we are...
Say there's an earthquake. Or a rockslide. Or an avalanche.
Humans can see if there's a gap in the road. They can see if the road has moved. They can judge unusual conditions.
If the roads are full of machine drivers will have to simply stop, because the machines won't have the judgement to handle the situations.
And who pays for the insurance? Why should I pay for it if the accident will not be caused by me, but by some programmer who forgot to take some variable into account when writing the software for my car?
I understand that I am responsible for the accident if:
1) I was actually driving the car (like it is now)
2) I neglected to maintain the car, therefore some part of it failed and resulted in the accident.
However, if the software causes the accident all by itself, why should I pay for it? Especially if I had no way to prevent the accident.
Naw, they're just Apple reality-distortion-field brainwashed to salivate whenever they hear about a new toy.
iCar asap. They can't wait to see a real car driven by a furby! It doesn't matter how dangerous it is.
My prophecy:
Self driving cars will be better than humans in 99,99% of the situations within 3 years. Sadly it will take another 5 to wide scale adoption and yet another 10 years for human driven cars to be banned to racetracks.
Your calculation implies that everyone replaces their car within a 5 year period.
That may be true in Beverly Hills, but it's certainly not true in the rest of the US.
The average age of cars on U.S. roads is 11.4 years, IHS Automotive reports. The average age of vehicles on U.S. roads has hit a plateau of about 11.4 years, according to an annual study by IHS Automotive, an auto industry research firm. Jun 9, 2014
In any case, I find your optimism unjustified.
In the Bay Area for instance, we could already make BART trains fully automated, and it's been studied they would become safer to boot, but attempting to change that system to get rid of the operators would be absolute political suicide.
We already know that. The insurance company is responsible.
Also, in many cases 'true accident' or 'Act of God'* is an actual category. IE 'Nobody did anything definably negligent, so the accident isn't really anybody's fault'.
A tiny piece of road debris was hit at just the wrong spot, causing a tire to blow, which the self-driving car wasn't able to adjust for fast enough, causing an accident. A sudden wind gust(in a wind storm, most likely), pushed the car too far over, etc...
When self-driving cars are developed enough to be usable by the average person, for something like 99% of trips, while being demonstrated to be substantially safer than the average driver for said trips, I'll support liability reform to get them on the road. What form that reform would take, I don't really know yet, other than a wild stab at 'limited liability per individual'. IE the insurance/auto company won't have to pay out more than $5M per individual, adjusted for inflation and increasing value of human life. Unless a 'coverup' is found, deliberate negligence, or whatever.
*A statement that I don't like because I don't believe he exists.
I don't read AC A human right
Keep in mind that the occupant of said self driving car will have both hands free to operate whatever firearms he might possess.
You never know, it might be me in that self-driving car, and I might take exception to your engine block in that case. If I'm feeling nice. If I'm not feeling so nice, I might go after the steering actuator.
I don't read AC A human right
Drivers are expensive. Self-driving could be less than 1/4 the cost of a minimum wage employee for a year.
I don't read AC A human right
They might as well asked the CEO of Tim Hortons about the future of undersea exploration.
Lyft CEO doesn't know shit about cars or the automotive industry, why the hell is anyone asking his opinion as if he is an expert?
Do not look at laser with remaining good eye.
I don't know that. It should be perfectly possible to make a machine that can drive as well as, or better than, a human can. Have we managed to make that already? I don't know, but from the info Google have been publishing, it actually looks like we have, or are pretty damn close.
Just because it's a machine doesn't automatically mean that it sucks at making decisions. Humans are machines too, and we let them drive.
In St. Petersburg, Russia subway does actually already have a computer-driven vehicles. The problem there is much simple: there is a single path. All you need is essentially only start/stop commands. Subway carts driving is completely automated. But the person still sits there. Probably for legal and emergency reason. For example if somebody is holding a door, preventing it from closing - initially automated voice would make an announcement: "do not hold the door", if no effect - machinist would yell on PA. I think if you want to have a self-driving car you need to modify the road, not the car, making it more like subway/railroad with all its advantages and disadvantages.
You've never seen a machine attempt any sort of image recognition then. Spoiler: They suck at it. Google make a car that behaves according to a set of known and understood circumstances in a controlled environment. I've never driven on a public road that exhibited those characteristics.
How do you define better? I'd love to see what maths you'd use to tell a computer the difference between a plastic cone and a small dog. Image recognition is the elephant in the room for self-controlling robots. Until that is solved then they'll never be allowed on the streets.
Your calculation implies that everyone replaces their car within a 5 year period.
That may be true in Beverly Hills, but it's certainly not true in the rest of the US.
It also implies these wonders of technology will cost the same as a regular car. A cheap aircraft cost $300k. An automated drone costs a couple of million. If those costs translate to cars then they're guaranteed to never be mainstream.
There things that can happen with auto drive cars that fall under criminal liability and criminal courts.
When a human causes an accident, it is almost always something that would have been obvious to another human driver, resulting in the general opinion that humans are stupid. Most people are intelligent enough to realize that SDCs will not be perfect, but they will prevent a lot more accidents that they will cause, and the tradeoff is worth it. Seat-belts and airbags also occasionally kill people, but they save ten lives for every one they take, and people accept the tradeoff.
Your logic is not quite right there. Firstly a seat belt or airbag is not a decision making piece of machinery. A self-drvie car actually has to make decisions. Is that a plastic bag, or a cat? Each decision is a risk of a crash. You're also assuming that if human crash rate is x, then as long as automaton crash rate is less than x then everything is fine. But it doesn't work like that.
People believe it is only other stupid people that crash, and as long as they're in control they have a say in whether they crash or not. Handing over control to an automaton takes away that control. So your perceived risk goes from zero to something greater than zero.
Put this in with the flying car. Even if it did actually work, it's too expensive/impractical/risky to ever get approved anywhere public.
I have a friend who is epileptic who would really benefit from this, there are times that she simply can't drive. Her roommate has fibromyalgia, and there are times she's almost immobilized by pain. Me, I'd want one as the two out of state places that I go to the most are 500 and 620 miles away, I'd LOVE the ability to pilot the car to the interstate then sit back with a book.
The problem is idiot American drivers. I've been accident-free for over 20 years now, but having worked for a police department and helped develop a database for tracking car accidents, I definitely appreciate the infinite diversity of the idiots out there on the road.
When you sympathize with stupidity, you start thinking like an idiot.
Road rage. Won't happen with self-driving cars. Also, gang-bangers will have a harder time doing drive-by shootings, car jackings, etc.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
The biggest hurdle to take is to correctly measure the surroundings. If you did that via image recognition, then yes, AI would be important. But there is laser, radar, GPS and so many other sensors involved that do nothing more than note distances to targets, location on road etc.
And this is impossible IMO. On a busy road you are only centimetres from other vehicles doing 60km/h, 120km/h for oncoming traffic. A human accepts that the car next to him probably won't change lanes suddenly and crash, so accepts the risk knowing that there is a chance it could happen. A computer needs AI to decide which of those risks are acceptable. When you come around a slight bend and a car is coming straight at you, you know they are likely to miss (still might hit, but that is the risk you accept) because of the bend in the road means the driver will veer past you. How does a computer work whether that is an acceptable risk, or an actual impending collision which requires emergency braking? The public road has too many variables. Too many risks that a human accepts that a computer won't be allowed to.
Man with stake in business driving people places claims new technology that would obviate said business aren't the future! News at 11.
Therefore you're just fine with putting machines on the road that you KNOW won't have any judgement? Just because some human drivers aren't good? You want fleets of more insentient drivers? How is that acceptable?
Better than putting alcoholics and other impaired drivers on the road - and there's a lot of them. And people who drive, text, and use their laptop at the same time. Or stuff their face while driving - that Big Gulp completely blocks your view of the road. Or drivers that are too tired. Or too timid to do a merge onto a highway in heavy traffic but they try it, and their nerve fails them halfway through, blocking everyone behind them. Or the pizza delivery guy (or soccer mom) who thinks that they can do 45 mph in a school zone. Or can't park worth a darn and end up 4 feet from the sidewalk.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
the intelligence to avoid situations better than say an actually intelligent being?
It's not the intelligence. It's probably going to be a long while before the computer is capable of modeling the mental state of other drivers sufficiently to tell whether they're likely to cut you off. It's the patience. It won't execute a passing move at such a speed differential that it can't get out of the situation if the other driver goes all asshole on it. If that means some dick cuts the computer off and then rides alongside a semi for miles, the computer will be patient and wait for the road to clear. You'll get to your destination in most cases within 5% or so of the time you would have got there if you drove like an asshole, and you'll get there more often. So will everyone else.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
The fact you are comparing the processing power of a self-driving car to a SunPlus SPC81A shows you're not even bothered in having a rational, adult discussion. You've made your mind up and no amount of evidence will dissuade you from vomiting up paranoid, incorrect nonsense for all to see, as if somehow everyone else won't instantly realise you are a grade-A muppet.
Anti: blah, blah, blah, self-driving cars will never be good enough, they're unsafe, they don't think!!!!!1111 Panic!!!1111
Pro: Yes, they're wonderful and never distracted and will be better than puppies and rainbows!!!!!11111
It's really simple. People who want it to work are trying to make it work. If you antis are right, they won't be good enough and they'll never be much more than a curiosity. If you pros are right, they'll be provably better and the anti argument will be simply refuted by saying "Look at the data."
The LIDAR in a Google car would see you before you saw it. The rest is just interface work - developing a more human-friendly interface than blinking lights or the ambiguity of a driver's face and gesticulations (is he waving at you or the car trying to turn behind you, etc.).
Foolish, IMO. My ex has a friend who was about to cross the street, made eye contact with an older gentleman who motioned to her that it was ok to cross, then promptly hit her when she tried. Turns out he never saw her and the gesture she thought was telling her it was ok to cross had nothing to do with her at all.
IF self-driving cars work as planned, they'll always notice you. They'll never (ok, nearly never, aka less than a comparable human driver) hit you. If they aren't provably better than human drivers, they shouldn't be allowed on the roads, and I daresay in this litigious society, never would be.
The very same controlled environment it's being designed for use on: public roads. And bear in mind that the current version is an early attempt. Clearly they will get better - much better. Computer image recognition has progressed a lot in recent years, and continues to improve every year. Human image recognition isn't perfect, and can not be improved. Human image recognition is ridiculously bad when the image to be recognised is in the road, yet the human's eyes are on their phone/radio/attractive person on the sidewalk/clouds/daydreaming/whatever. Computers aren't distracted by any of those.
You've got to try harder! If you are emotionally distressed by the idea of automatic cars, just say so. Pretending you have logical concerns isn't helping anyone.
We already know that. The insurance company is responsible.
Three states (California, Nevada, and Florida) already allow SDCs on their roads. The liability issue is already resolved.
This has never been tested in court. It's one thing to quietly pass a bill and another when CNN, MSNBC, FOX, etc all run "first human baby murdered by rampaging machine" as a headline
Another problem is that when a self-driving car causes an accident, it will be something that would have been obvious to a human driver, resulting in the general opinion that self driving cars are stupid.
When a human causes an accident, it is almost always something that would have been obvious to another human driver, resulting in the general opinion that humans are stupid. Most people are intelligent enough to realize that SDCs will not be perfect, but they will prevent a lot more accidents that they will cause, and the tradeoff is worth it. Seat-belts and airbags also occasionally kill people, but they save ten lives for every one they take, and people accept the tradeoff.
Are you seriously busting out "people are logical and rational AND will know better" as an argument? Seatbelts have no sensors they are purely mechanical. Airbags need some extremely basic smarts but there have been injuries and fatalities due to design problems. But very few. SDC need extreme smarts bordering on strong AI do do what even the worst humans can. Humans stem from creatures that have been navigating the world for billions of years - a strong ability to navigate is built into all animals. People wont tolerate machines failing to miss simple human avoidable accidents even if safer. SDC, if implemented widely, will cause tens of thousands of deaths a year or more the the political backlash will have more to do with stupid people and politics than sense.
Is the object conical and orange? It's a cone. Is it fluffy, small, and even slightly moving? It's a small animal. You appear to be arguing from ignorance and emotion. It's not very becoming. The concerns you have are already solved.
Airbags do make decisions. The airbag decides whether to deploy, and, based on the impact, how fast to inflate.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
This argument reminds me of the arguments I heard about seat belts (e.g. I want to be thrown clear in a crash, Don't want to drown trying to unbuckle, etc.) Children are run down pretty regularly, so are dogs. In many, if not most, of those cases the car could have been stopped safely if 1) the human driver reacted faster / wasn't distracted and 2) the driver was traveling at a speed safe for context and conditions.
I assume the statistic that 90% of automobile accidents are caused by human error is correct. The automated cars of the near future won't be perfect -- but they will outperform humans as drivers by a large margin.
The sensors and programming used in the self driving cars will continue to improve (at a much faster rate than human drivers improve -- if they do at all). Black ice is "invisible" to most drivers. Is it invisible in the infrared? Could wheel sensors detect minute slippage and compensate faster than a human? Could networked self driving cars warn each other of such conditions as they are encountered?
It is true that specially designed roads would also improve self driving cars, but even on today's roads they will be superior to human drivers.
Millions of business owners do this already everyday when they insure their workers against doing something stupid. By choosing to own and operate a SDC you have to accept the minimal level of risk that it could be the cause of, or faulted in an accident. You pay the probably incredibly small liability insurance and move on. When most cars are SDC accident rates will likely fall drastically, and so in turn with the cost of insuring a vehicle. Even with the high accident numbers we have today many people pay for full coverage insurance, including clauses protecting them from uninsured drivers. I'm rather confident that insurance will be a non issue with SDC's.
I'm really sorry to have to be so direct but that is the dumbest argument I've ever heard.
Driving is not a matter of intelligence. It's a matter of reaction. Sure, intelligence and experience help you anticipate when other drivers are being idiots, but there is very little involved in driving that can not be compensated by reaction time and adhering to proper distances.
actually extracting raw enviornment data is fairly trivial compared to the interpretation, fusion, and implementation problems. When sensor data conflicts, or sensors become dirty, its things get hard. But how do you know if that bicycle is going to stop when its trajectory intersects with your vehicle? People make eye contact to make sure the cyclist sees you - simply running them over for a minor infraction is not a solution, nor is applying the brakes every time any pedestrian or cyclist is near
The biggest hurdle to take is to correctly measure the surroundings. If you did that via image recognition, then yes, AI would be important. But there is laser, radar, GPS and so many other sensors involved that do nothing more than note distances to targets, location on road etc.
as above this is the most trivial thing - if that was all that it took it would be already solved
Autonomous driving certainly isn't trivial, but the other thing you have to keep in mind is that your oh so intelligent human drivers are actually driving like morons a lot of the time.
Please stop putting the bar for autonomous driving so high the systems have to practically be perfect the be viable. The moment they are twice as good as a human should be the moment we start switching. And we're not far from that.
Remember how badly the average driver actually drives. And then remember that half of them drive worse than that.
Add on top of that networked driving, where cars coordinate over several hundred meters and you'll see so much potential gain even from non-perfect systems it's staggering.
With 50 different brands, car nuts screaming the government is taking away their freedom, etc. We will never have cooperative piloting on general roads. It's pretty obvious you have never worked with this technology.
You say that as if this isn't already a solved problem. There are loads of autonomous vehicles already successfully navigating the public roads in general traffic. This technology is so far along that companies like Nvidia have off-the-shelf autonomous car kits on the market, ready for carmakers to integrate into their vehicles.
Your rant sounds like the guy in 1906 saying that travelling faster than 35 mph was impossible as the Stanley Steamer screams past at 120 mph in the background.
What would be a lot of fun would be the rednecks encountering many state troopers a mile or so ahead. That's because these self-driving cars are made by Google so it will have all types of sensors up the wazoo. The second the car detects it's being "attacked," it will send a distress call over wifi, along with the license plates of the attackers.
This though, is the reason I don't like self-driving cars. The car manufacturer and the govt can track your every move. It's a ridiculous invasion of privacy.
We will see businesses insist upon self driving cars. The local pizza joint has huge issues with delivery. The liability as well as the expense create too much demand for self driving vehicles. Imagine a machine making your pizza and another machine boxing that pizza and loading it into an autonomous vehicle for delivery. A human might need to come in a few minutes a day to be certain that all is well but the entire business can be automated with zero labor expenses. Then imagine insurance companies pushing for self driving cars. After all the great risk to their profits is not sheet metal but wounded mortals. So the local grocery store can pull your order, slam it to your debit card and your autonomous vehicle can pull into the loading bay and bring it home to you. Reduced exposure to driver and passenger injuries will cause insurance companies to push for self driving vehicles. The grocery stores will love it as they will need far less cashiers and they won't even need top display the products to the public. Frankly businesses could care less about the public or what the public wants or needs. Businesses care about money and money says that cars and trucks without humans are far more profitable. Imagine Fed X or UPS doing their thing without human drivers.
You can get a fully automated drone for hundreds of dollars, not millions. Not big enough to put a person in, but still fully automated. Google and Amazon are looking to deploy huge fleets of fully automated drone delivery aircraft costing a few thousand each.
We already know that autonomous vehicle technology isn't prohibitively expensive. Nvidia already has their system on the market for carmakers to integrate. It adds thousands, not millions. And depending on how things pan out, you might recoup all of that initial outlay in insurance savings pretty quickly.
The end date on his calculation was 18 years from now, not ten. By then, most cars on the road will either have self driving as a standard feature, or can be readily retrofitted for that functionality. Assuming the GP is right about self driving taking off, of course.
Luckily, these cars won't be built by no bid, bloated military contractors.
Because if a cabbie can't keep their taxi clean, what makes you think the average person will too?
Maybe that's a thing in your area? With a single exception, all the car-sharing cars I've used for the past few years have been fine and on the level of taxis except for a little more dirt on the floor (and only the floor).
more commodity like cars
Thanks to the used car market, the price of buying a car is not the problem. The cost and hassle of maintaining one is. If you don't need a car every day, it's simply not worth it.
Assorted stuff I do sometimes: Lemuria.org
Why should you pay for insurance on your house?
So you think millions of humans with millions of differing attention levels, skill levels, alertness levels, differing visual abilities, physical abilities, inability to communicate with each other, etc. is better than millions of computers with all equivalent abilities and which can communicate with each other? So that 16 year-old and that 89 year-old on either side of you are better than strictly computer controlled vehicles?
Thought is wonderful. It doesn't matter one whit.
Reaction and reaction speed is what is critical in any non-standard driving situation, not thought.
You are somewhat the rarity with your vintage car. For most people, drive by wire is already a thing. The throttle has been drive by wire for years on most cars, and some of today's cars are steer by wire. (Yes, there is manual reversion if it fails, but in normal driving you have no mechanical connection between the steering wheel and the wheels of the car). Many cars can brake independently of the driver. Even my 2007 Civic has traction control and ABS fitted as standard.
Oolite: Elite-like game. For Mac, Linux and Windows
As someone who worked in a robotics lab for nearly 8 years and who knows the technology - no we are no where near being able to do as well as the worst humans comparing apples to apples. In 30-50 years we may have them until then maybe dedicated lanes on highways or something may be possible in 10-20 years.
I like things I understand. My car is simple enough that I can understand it. When I took apart and cleaned the carburetor , not only my car started running better (the intended result), but I got more insight into how it works. For example, when the engine is cold and the RPM is increased, when the engine warms up, the RPM does not decrease until I briefly press the accelerator pedal (which means that this situation does not really happen when I am driving, just when I am warming up the engine in a very cold winter day). This is due to the way the cam that adjusts the throttle works.
Pretty much anything in the car can be understood easily - this pushes that, which rotates this, which moves that etc... Even if I got access to the source code of a modern car, it would be really difficult for me to understand it, and the mechanical systems are more complex too.
Also, with my car it is quite easy to repair or replace parts - here is a lot of empty space, no need to remove a wheel to change a lightbulb. In the time that it takes to change the lightbulb in a modern car I could probably change the whole headlamp assembly (which I have removed, replaced the reflectors (old ones were rusted) and put back in).
You should have stayed at the lab for a few more years :) Clearly we can't compare apples to apples, as we are comparing a computer system to a human - perhaps if you elucidated your point you'd get a better answer. The self-driving cars are doing fantastically, and humans are still, on average, shitty drivers. The worst human drivers are fucking abysmal, so making a system better than them has already happened (just look at the Google car stats).
Thank you!
How do you test? All of the "ran someone over" bugs will have to be labeled "can not reproduce!"
Thanks, APK! Off your meds again?
modeling? How about if we wait until we all have experience that computers are good at the basics before we assume that they can safely drive cars?
And APK = guy with serious mental health issues who needs help but doesn't realise it.
And who pays for the insurance?
You do.
Why should I pay for it if the accident will not be caused by me, but by some programmer who forgot to take some variable into account when writing the software for my car?
Your car already contains millions of lines of code, so this is not a new issue.
And one of these days, I'll take a picture of a road I use going to or from my home on the occasions when I drive: 3 lanes wide, parked cars in one lane, no center stripe, oh, and did I mention that buses use this street? No, a self-driving car will have smoke coming out of its ears....
On the other hand, self-driving. They're going to be *really* expensive. For less money than that, couldn't you use existing methods of getting somewhere without having to drive? There is this thing called "public transportation", as well as this business called "taxis", and all your fares, over the estimated lifetime of a self-driving car are probably a *lot* less than you'd spend on the car.
Oh, and how safe will you feel knowing there are 5 and 10 year old self-driving cars, some years from now, many bought used, whose owners may, or may not have had the recall upgrades applied?
mark
A modern fuel-injected car already runs well without having to dick around with a carburetor. We don't need to mess around with our cars just to keep them running decently like we used to.
Lots of open space under the hood is nice if you have to work on a car, but you wouldn't need to work on a better car very often, and would get much better mileage, and a better chance of surviving a wreck instead of being crushed by a steering wheel. I used to drive a gas-guzzling 1964 Cadillac. My 2001 Accord is superior in all ways.
"And who pays for the insurance? Why should I pay for it if the accident will not be caused by me, but by some programmer who forgot to take some variable into account when writing the software for my car?"
What if insurance for a SDC is 1/10th the cost of insurance for you driving the exact same car? Will you still complain, or just drive yourself (and pay more) because you don't trust the software?
Ninjas don't carry tic tacs
95% of driving might be mundane, but an alarming number of people where I live seem incapable of doing that 95%. The number of times I've seen someone on the side of the road that have run into a telephone pole. Daily I see at least 1 car that can't stay in a lane. Definitely multiple times a day that people just doesn't use their turn signals at all, and I know some of them. People driving through stop signs and red lights.
According to local statistics most accidents are caused by drugs and failures to follow traffic rules. Our police keep on going on about how speed kills and there are articles in the paper quite frequently about how these people were racing, or just driving really fast. The thing I find funny about these articles is that usually they will have the smallest line about them being way over on their blood alcohol level, high on heroin, or stoned on marijana, but they really try to downplay these parts.
Right now self driving cars would stop the majority of accidents.
How about that 5%? Well as more self driving cars would get on the road, that 5% would probably shrink to less than 1%, depending on the area. Plus there are already real time algorithms for detecting moving objects, coupled with multiple cameras the car can see things like deer and children much sooner than humans and react to them.
The main problems will be in the transition where there are still human drivers doing really stupid things and the self driving cars. We humans are affected by lack of sleep, by drugs, but not getting enough vitamins, distractions, tunnel vision, bad driving habits, missing signage, having two eyes facing one direction at a time, and sometimes just pure stupidity.
Microsoft, Apple, Google, Amazon what's the difference? All steal money from devs and control with walled gardens.
same could be said about pedestrians (I say that as both, a driver and a pedestrian).
One as, I was trying to cross a street a a swarm of said pedestrians (for which the light was RED) prevented me and a bunch of other people (cars included) from doing just that, because they wen't paying any attention (and I'm fairly certain that it is not an exception at that particular intersection).
Another time, a drone forced me and a whoever was behind, to come to a COMPLETE STOP in an intersection when the light was GREEN because whatever he was doing on his phone was more important that his or others's people life.
Your car already contains millions of lines of code, so this is not a new issue.
My car doesn't. Last time I checked, there were no CPUs in my car (other than in the tape deck), didn't find one when cleaning the carburetor too...
When they built my car in 1982, it looks like computers were not as common as they became later.
Cleaning the carb once in 30 years doesn't seem so bad. Also, as if fuel injection does not fail or just wear out...
I would still need to work on a newer car too, unless modern cars are now failure proof, something will wear out or just fail over time (if I bought a newer car I would be expecting to drive it for many years). The problem will be that when that happens (inevitably before or during some trip), it will be more difficult to repair.
As for the fuel consumption, well, one, it's not that bad (and LPG is quite cheap) and two, as I really dislike environmentalists, I now at least have the option of turning the mixture screw just a little bit when they annoy me enough. I have never done this, mind you, yet, just that I have the option.
1964 Cadillac, eh. I always though that if I had the money to buy a second car it would be soething lie a 1974 Cadillac, seems great, well, other than the fact that it only has an automatic transmission (I am not used to it and like the control that a manual transmission offers me).
Unless you or your family has been affected, then you tend to think a little differently. I've had the misfortune of having ABS fail thinking I was in a skid, so during a low speed stop I almost rearended the car in front of me. I did have the reflexes and quick thinking to go to the shoulder to miss an accident. However, as a result, I don't completely trust the ABS on any car now.
Moores law pretty much guarantees that anything that can be done at all today will be much better faster and cheaper in a few years.
Those things are hard, yes. However, Google has quite awesome programmers. Note: to me it only has to be better (defined as in less lethal accidents) than the average driver. That is a high bar, but I assume Google isn't doing it because it's easy.
Well, I might have a way, but it only works on a semi spherical planet in a vacuum.
Wait, you run propane through a standard carburetor? How does that work? It isn't like LPG can settle in the float bowl. I thought you had to replace the carburetor with a regulator and mixer assembly.
I'd like to see how your setup works.
Oh, and to the point.... you don't really work on modern cars. They rarely need anything beyond fluids and wear items like tires and brakes, but when they do, you end up letting the professionals handle it. I used to do all the work on my cars back in the carburetor and timing light days. Not any more. They have everything packed in there so tight it takes hours just to change the spark plugs on some cars (BMW 328i, I'm looking at you). The computers and sensors are so good that you never really adjust anything anymore. While you give up some of the fun of detuning your Chevelle SS to get that 100rpm idle lub-dub-lub-dub sound, what you get in return is 100k miles with nothing more serious than an oil change out of a little turbo charged 4 cylinder that puts out more horsepower than the biggest muscle cars of the 70's. Not a bad exchange.
I'd like to see how your setup works.
You are right, I cannot just pipe LPG to a carburetor. The LPG goes to a regulator that reduces the pressure to around 1 bar and a pipe from the regulator goes to the air intake of the engine (between the air filter and carburetor).
LPG cools down when it expands, so the engine has problems starting nad running when its cold. It is recommended to use LPG only when the engine temperature is above 40C. The regulator is heated by the hot antifreeze from the engine during normal operation.
There are valves tat control gasoline and LPG flows. I can switch one on or switch both off. To start the engine when its cold I have to switch on gasoline, start the engine (can take a few turns for the mechanical fuel pump to get the gas to the float bowl), wait for it to warm up (can drive during that time), switch off gasoline, drive until the float bowl is empty and turn on the LPG.
When I come home from work, I switch the fuel to gasoline for a few seconds and then turn off the engine. This way the float bowl is full and the engine starts quickly the next day.
As for the modern cars - going to a mechanic takes time (most of them work at the same time that I do, so I have to excuse myself from work) while I can repair the faults that I am able to during the weekend (if the problem is not big) or on the side of the road (if it's something that prevents the car from working). Modern cars do not require frequent adjustments, sure, but maybe that's because they are still relatively new. My car is 33 years old.(it has been in our family for 19 years now). Maybe a car made in 2010 will have as much or more problems than my car in 2043 (with less ability to repair)...
Riding in a vehicle, where the one in control is not themselves in the vehicle with you, is probably not the best decision you could make. 8-)
Thanks for sharing.... your ride sounds amazing. Back when I was in high school around 1980 a buddy and I rebuilt and customized a VW Super Beetle. Nothing as far-out as your LPG setup, but still a lot of fun. We had the cylinders bored out for new pistons, put in high-ration rockers and a new crankshaft - we even put dual Webbers on the thing. 8 barrels of carb for a four cylinder, pretty funny.
What a great project. We ended up getting it to dyno out at 135hp, which ain't too shabby for a car designed for more like 45 hp.
I bet your project has given you no end of topics for conversation.
The LPG setup was done by professionals about 10 years ago. I actually have problems with fixing the body etc. Electric stuff is easy, so are the parts where I can remove said part, repair and replace it. OTOH, patching a rust hole in the body is a big problem. Converting a gasoline car to LPG saves a lot of money if you drive more, because, at least in my country, LPG price is about 41% of gasoline (A-95) price, but the car does not burn twice as much of it.
As there are not a lot of cars like mine left, I do sometimes get nice reactions from people at a gas station or wherever.
A public road is not a controlled environment.
Computer image recognition is not at the early attempt stage, there are decades of research and is still shit. There has been zero success in getting a computer "eye" to function like an animal equivalent.
Until either of those two things change dramatically, it won't happen.
Just wishing it will get better won't make it so. That is emotion, not logic.
Citation? A cone that has been run over is no longer a cone shape. How about a plastic bag? Of a small branch? These problems have not been solved and no amount of wishing they were will change this. If they have then I await you link to the relevant research.
If you think simple trigger logic and advanced visual recognition AI are the same problem then I can't help you.
The problem hasn't been solved because I can't buy one and use it today. Marketing hype and vapourware, is different from a solution.
Did you even read your own link? Nvivdia is selling a chip. That is so far from a "solution" that it's laughable. And as for drones, it the carrying the human and satisfying the safety requirements that is the expensive part. I feel like I'm explaining this to a child...
Google does indeed possess some fine talent, but there are some things that no amount of skill will solve (within a reasonable time frame). Time machines, invisibility, flying cars, and maybe self-driving cars. I'm well aware of the Google situation, however still can't see how it could work in this world.
Take away the technological challenges, even if they produced something the equivalent of a human today, would it be politically feasible? What would the costs be (vehicle, infrastructure, administrative/enforcement)?
It's also worth noting that for all Google's talent, they have a long list of shit products to show for it. It says something when your best product was your first product, and 15 years later, other than a couple of clever acquisitions, most of everything else has been rubbish.