Why is everybody so hot for driverless cars in their current state?

Someone who claims to have worked on Uber’s self-driving technology lurks here.
This.

They don't have to be perfect, just better than humans which kill thousands a year in the US.

Humans. It is very obvious self-driving cars will rapidly be safer than humans.

Did they even have more accidents overall or are you just cherry-picking one accident category with a terribly low sample size?
And didn't Google's and other cars also rack up some million kilometers?

Either way, I don't much care about the statistics for experimental cars that aren't even allowed on European roads, what I think will be safer is the production models. Same as with planes, cars and almost every other powered vehicle before.

That 1 death isn't statistically relevant, and it was the pedestrians fault anyway.

>no regulatory body stifling innovation and market exploitation
>idiot pedestrian running out into the middle of traffic deserved to die, literally who cares. roads are for cars.
>car companies rightly blame operator error for any problems, freeing up more money to make better cars.
i can't see why anyone WOULDN'T be excited.

>That 1 death isn't statistically relevant
It is, actually.
There is only a 7.5% chance of it having happened if self driving cars are equally as safe as humans.
Through bayesian bullshit this turns into about an 80% chance that self driving cars are less safe than humans.

>Did they even have more accidents overall or are you just cherry-picking one accident category with a terribly low sample size?
I didn't cherry pick, it's just the one I know the statistics for. I also don't think it's unreasonable to use "deaths" rather than accidents as deaths are chiefly what people are concerned about, with "accidents" you will get an awful lot of mild fluff in there as well as deaths and severe injuries.

In the interest of fairness I've looked up the accident rates and they are 4.2 accidents per million miles for humans and 3.2 for self driving cars. They currently crash less (0.76 times of the accident rate of humans) but are vastly more lethal when they do (13.33 times the lethality rate of humans).

>And didn't Google's and other cars also rack up some million kilometers?
Yes, that's Waymo (a subsidiary of Alphabet/Google). They reported 4 million in November, which combined with Uber's 2 million makes 6.

>Either way, I don't much care about the statistics for experimental cars
I'm not saying you should, the statistics are very poor (one death is not a very good sample size). I acknowledged that. My point was that you shouldn't parrot this "Self driving cars are safer than humans" thing when it is quite obviously untrue at the moment. It's a prediction and I agree with you that it will most likely be true eventually but it is NOT true at the moment.

That assumes that all the mileage is equal when most driving is probably done in bumfuck nowhere where it's almost impossible to get in an accident, not in a city full of other people.

There was literally an operator in the car at the time who couldn't react. Some dumb cunt with a bike jumped out from behind a hedge and got creamed.

>Yea, but humans go wrong at a much increased rate and autopilots can be made near perfect.
That is why having both is good idea. There is only one problem to solve: how to make it not annoying. Because for a lot of people who learned how to drive manual first any automatic is annoying.
>Of course crappy airplanes still crash
This failure wasn't depended on airline, but rather on engine construction. And we are speaking about 737-700, that is pretty new.
>And they're trained to standards we will never train car drivers to.
Indeed.
>and autopilots for airplanes and ships can be trusted more than human pilots already, overall.
Autopilots there are really primitive, and airlines still trust more to humans.
>These work pretty good but no one pays for a decent deal of redundancy, unlike with streets.
Redundancy there is simple, train drives to the station and halts.
>. Even Switzerland and the Tokoyo area (which are both really incredible at public transit) have issues with various chokepoints and it's just hard to fix these.
They are. Actually this is why roads need to be rebuilt from scratch, but it is expensive.
There at night it is pretty bright. And even on that footage it is obvious that car could start braking and brake to 20-30 kph reducing risks.
They won't. Roads are not built for that. And as I said self-driving tram/metro>self-driving car.
For example, it would be difficult for pedestrians to understand what autopilot whats to do, while looking on real driver you can see where he looks, and what he does.
In my shithole it is ok not to give a pass for pedestrian on crosswalk, if there are a lot of cars and 1-2 pedestrians waiting. And in this case autopilot what will do? Every car will stop for every pedestrian and cause chaos and jam? Or not allow them at all? And adding traffic light in that area is an overkill.
If we will look in drive code - it just says that you should give way only if (s)he is on crosswalk...

>car companies will just blame you (as seen in with the recent tesla crash) for not knowing how to operate the device DESPITE advertising their lane assist features using phrases like "Full Self-Driving Hardware"

Tesla might have been the first to mass market it, but it feels like its falling behind now.
>call it autopilot
>need to be 100% attentive like you're driving anyway
>build up a false sense of security
>die

That uber death was clearly a software/hardware failure, it was near perfect conditions for LIDAR and yet something went wrong.

Attached: 1514944161414.gif (570x498, 851K)

I did not say it's much safer now, I said
>Sticking with regular drivers is the biggest risk

Sensors got better perception, computers have better reaction times and more systematic reactions and so on and so forth. They'll also drive more ecologically - better braking and acceleration behaviour in cities and so on.

Humans will never get that good, even if you accident-proof cars more and more they'll still crash the cars all the fucking time and no realistic training will ever (nearly) fix this.

Self-driving cars will be out of the experimental stage soon and just work better and better rapidly.

Wtf stats are you using when the cause of the accident isn't even public knowledge? When we have a sample size of one when it comes to the conditions that led to the accident? How, exactly, did you come up with 7.5%? Do you have statistical data for accidents where pedestrians were struck on that stretch of road in those lighting conditions and similar?

>if there are a lot of cars and 1-2 pedestrians waiting. And in this case autopilot what will do? Every car will stop for every pedestrian and cause chaos and jam? Or not allow them at all? And adding traffic light in that area is an overkill.
>If we will look in drive code - it just says that you should give way only if (s)he is on crosswalk...

You understand that once you're not in control of driving that kind of shit annoys you a lot less.
Think about the last time you got stuck in some jam in a car VS riding the bus (or some other shared mode of transportation)

i just hope we don't reach a point where normal cars are banned entirely, even on tracks (probably for muh emissions or some bullshit). as a car enthusiast we're basically getting buttfucked and if they have a full-on ban i'll prob kms

>>no regulatory body stifling innovation and market exploitation
>making it sound like a good thing
leave, demon

Attached: 1338579934272.png (851x392, 26K)

>>uber death from a couple months ago already puts them more orders of magnitude more dangerous than drunk drivers
not this shit again
thousands of people died in traffic accidents all over the world in the time it took me to write this post, yet you choose to shit yourself over 1 death

What a complete load of shit.

>who couldn't react
Even in Uber's video there is almost a full second where the person is in view, never mind that videos taken by other people in the same area suggest Uber's footage was artificially darkened. Even a human (who was paying attention) could have reacted to that, let alone a machine. The car does not even attempt to brake.

>jumped out
No. They were walking across the road at a steady pace.

>a hedge
No. Go back and actually watch the video. They had already crossed the center line of the road. There is unambiguously nothing in the middle of the road for them to jump out from behind, just a line of paint.

I don't know if you're an Uber shill or what but you really sound like one.

No, the camera footage is very deceptive. If you look at other videos of that exact part of the road at night, it's lit well enough that a human driver even with poor vision should be able to see a pedestrian at either edge of the road and anywhere in between without the aid of headlights. As in, even if the headlights on that Uber car we're off, the driver would have been able to see that woman. The camera, on the other hand, was shit.

a ban is unlikely to happen, what will likely happen is that having a non-driverless car will just become more and more expensive, eg insurance costs will go up significantly.

Also maybe consumers won't even have a choice because manufacturers will decide that few people are buying regular cars (one day) and just stop making them.

Who knows, in the future a car might turn into another commodity instead , something people just use on demand but don't want the hassle of ownership.

>>no regulatory body of any kind overseeing e.g. code quality, minimum standards for equipment used, testing guidelines and frameworks
regulatory bodies just add additional burden of producing shit no one will ever read.

en.wikipedia.org/wiki/TAM_Transportes_Aéreos_Regionais_Flight_402
everything was ok with the plane, software actually detected thrust reverser malfunction, everything would be ok if "pilots" wouldn't react and just let computer do the job.

what i remember from episode, probability of switch failure was 10e-9, but they did not account from reliability of shitty contact itself.

>You understand that once you're not in control of driving that kind of shit annoys you a lot less.
Actually not. Chaos is even in the bus chaos.
>Think about the last time you got stuck in some jam in a car VS riding the bus (or some other shared mode of transportation)
Literally always, when it is a rush-hour. Metro doesn't have traffic jams, but it does have jamming people in. Bus - even though it has lanes, that doesn't work. Tram - it kinda works, but it still can stuck in traffic in certain spaces. And it is also full. Car actually is better since you are pressing brake/accel(/clutch) pedals and accelerations are not sudden, thus you are not becoming sick as in bus.

>spaces
places. Sorry my esl

>not this shit again
It's not shit, it's statistical fact.

>thousands of people died in traffic accidents all over the world in the time it took me to write this post
The world death rate is one every 25 seconds, so unless it took you seven hours to write that post that's hyperbolic bull.

>yet you choose to shit yourself over 1 death
I don't shit myself, I suggest that people should stop saying "self driving cars are safer than humans" until it actually starts being true.

>That is why having both is good idea.
No. Humans make the system worse.

> This failure wasn't depended on airline, but rather on engine construction. And we are speaking about 737-700, that is pretty new.
1984 isn't pretty new.

> Redundancy there is simple, train drives to the station and halts.
Congrats on getting some trains, but it's not quite as simple regardless. Train platforms and rails ways grow fairly physically big and expensive once you need redundant tracks and no chance of people running across tracks and so on.

> They are. Actually this is why roads need to be rebuilt from scratch, but it is expensive.
Extremely expensive, to the point where few but Switzerland had any will to actually go far and then expand further.

Sure, China is putting down a lot of rail, but will they care to invest 40+ times again in the same place, and keep it all in good enough shape for fairly fast rail vehicles? [Road vehicles -even autonomous ones- can go a varying speeds over crappy roads, rail technology doesn't really do this yet].

considering you're using a "pure statistical" approach here, why don't you go ahead and tell us how many people have died in vehicles driven by humans and compare it to the ones driven by computers
i'll wait

Woops, that was the -300. I guess the -700 was 1993. Still not new.

with so much insecurity in basic pieces of code in most parts of our lives, why the fuck has it been a force memed muh autonomous car.. banks are hacked, personal info sold, countries going after each other, and now you want to trust networked autonomious vehicles... cmooooonnnnnnn guy

You missed my entire point.

Which was when you're not in control of driving you get stressed out a lot less over that shit.

is it more stressful to be in traffic as the driver or as a passenger ?

When you don't need to pay attention to the road and driving you can just fuck around on your phone or close your eyes and if your trip takes a few minutes longer because of some traffic jam it had little to no effect on you

> thus you are not becoming sick as in bus.
lol just who does that happen to ? out of all the complaints ive heard about public transit that would be a first.
your personal ease at which you get motion sickness is not a fault of public transportation

because people are really really really really really fucking lazy.

t. unabomber

>The world death rate is one every 25 seconds
About 55 million people die every year and there are less than 32 million seconds in a year, so I'm gonna you a few days to do that calculation again, user.

I don't generally want them controlled over a network.

I want autonomous vehicles working off a cryptographically signed and checksummed openstreetmap, with lots of sensors ensuring distances and speeds are safe, and insanely good reactions even if you do something silly and shoot the wheel off or something.

Also, software is probably already in your car controlling just about everything from the fuel injection to your motor to the servo steering, and somehow it works fine.

Here's some fun statistics for that line of thinking: About 120 billion people have died in Earth's atmosphere. Approximately 18 people have died in space. Space is the safest place you can be.

Your comparison isn't any better because it doesn't account for statistical relevance. Even now, more people have been passengers in autonomous vehicles than have been to space.

It's not supposed to be any better, it's supposed to echo your own words back to you with a slightly more absurd but functionally identical example so you can hear how stupid it sounds.

Not my words user. But I do happen to think that the stats of deaths from autonomous vehicles vs. human operated ones are much more relevant than what you brought up. To the point where your comparison doesn't make sense. Just as clarification for the point you ARE trying to make, since I honestly don't know what it is, how would you propose measuring the relative safety of autonomous vehicles?

Just like horses are banned entirely, right?

>No. Humans make the system worse.
As I said, shit does always go wrong. What will autopilot do it rear wheel unscrews? What autopilot will do it something covers lidar or whatever (for example you've washed the car and it attracted birbs)? Or in case of just wiring malfunction that will cause fire. You might say that this is impossible, and sure, possibility is small.
>1984 isn't pretty new.
en.wikipedia.org/wiki/Southwest_Airlines_Flight_1380
Manufactured in 2000. And these engines were used in newer aircraft. And the point here is not in the shitty engine, but in that shit goes wrong in spectacular manner. And automatics can't just decide what is more important: depressurization or asynchronous thrust.
>Congrats on getting some trains, but it's not quite as simple regardless. Train platforms and rails ways grow fairly physically big and expensive once you need redundant tracks and no chance of people running across tracks and so on.
They work for 50+ years with minimal service. You can still find rails from 1970 still used today, even on high-speed routes. Sure, gravel were replaced, bad sections of rails were replaced... But they don't wear as bad as roads (especially in soviet russia, but there it is not roads, it is 4 cm of blacktop on shit).
>no chance of people running across tracks and so on.
Rail roads have fences, trams on 'high-speed' (30+ kph) zones have some sort of idiot protection as well.
Dunno, for me it is better to be jammed and be angry, or be in jam at wheel and be angry, rather being in bus sick, but calm.

They are banned only in highways tho. But people just found that cars are more reliable and produce zero shit.

It's being pushed by industry "PR" (shills)

That's how much people hate traffic, OP.

its all the citi faggots too who meme shit like this into existance... people who live in bumblefuck dont have congestion issues far from the shit tier conformity factories

Oh look a tiny sample size distorts the stats.
Township of 2 people, one kills the other, OMG the murder rate is 50,000 per 100,000 most dangerous place Earth!

I would measure them in deaths/serious injuries per distance traveled, not just pile up the corpses on each side and see which one's bigger. Of course the pile for human drivers will be bigger, there's been nearly a full century of human drivers and maybe ten years of AI drivers in one or two states of one country (and even within those states they're a vast minority of all road traffic).

i dont trust fly by wire in newish cars either, servos go bad, resistance can be fooled, ecu's have been hacked, everything with a wifi chip can be hacked, and be sure there gonna have gps in these systems... someone will always get in a break the systems for their personal gains/abuse/psyop

> What will autopilot do it rear wheel unscrews?
Generally a better reaction than a human pilot since it can observe each millisecond on multiple sensors whether a steering reaction has a desired result.

You can even see this on hobbyist quadcopters that NEVER would stabilize as well on sudden collisions and prop damage with human pilots & you can make this a lot better on cars with more sensors and more failure-handling-oriented software.

> Manufactured in 2000.
Ancient shit that can't fly autonomously - and either way, I don't get the argument here.

No, human pilots don't manage to catch you when something happens and you get ejected from the airplane, either?

cyclists are fucking cancer anyway so nothing of value was lost

Can you even read? It says "Now obviously that's really limited shitty data" right there. I wasn't saying we should damn all self driving cars on that basis, just that people should stop saying things that are objectively untrue.

You can't sensors for everything. Or you will put fart sensor in seat for climate control?

Why saying something misleading in any context?

Because you want to implant the idea that your technology is super safe in people's minds so they'll buy lots of it when it's available.

>You can even see this on hobbyist quadcopters that NEVER would stabilize as well on sudden collisions and prop damage with human pilots & you can make this a lot better on cars with more sensors and more failure-handling-oriented software.
Sure, there are some thing that electronics do better. ABS, ESP, ESC and crash prevention systems are indeed working in cars. But removing human input as dangerous and unsafe, as letting human to it alone.

Yes, you can. High resolution laser, RADAR, ultrasonic and visual (camera, both normal and IR) sensor arrays on all sides of the car [well as-necessary by engineering goals]. Human eyeballs never match this.

And sure, thermal sensors on each seat if you want automatic per-seat climate regulation or fart dissipation or whatever else the fuck. Wouldn't tie that one in with the driving sensor control, but computers with software that can handle this cost like $5-10 these days and use very low power, so why the fuck not.

Even Chinese early model robot cleaners for a few hundred can have dozens of sensors on them, it's no financially unrealistic plan to stick a LOT of them on cars and make much better software for them.

Attached: Xiaomi-Robot-Cleaner.jpg (1920x1080, 152K)

It will be a fucking macbook, that breaks all the time.
You just can't design all sensors. Too much things are going on in car.

Not the idiot you are talking to, but cars equipped with LIDAR and all that other shit can barely run in perfect conditions, however if the sun is in the wrong place, it is raining, its snowing or heaven forbid there is snow on the ground, or the car begins to lose traction everything goes to shit.

You have no idea how much shit like that costs to implement and program, it is obvious you will just keep responding with 'w.. w.. well they will just fix that!' when anyone with a brain knows that will take decades to bring to maturity.

Decades? The cars are already on the road. The training algorithms are already there. At this point it's just a matter of collecting data to train the software with. I don't think you fully understand how the process works. The software just has to be trained how to react correctly to various inputs. And you're right; that will take time, but not decades. Maybe you think that the engineers have to manually program how each sensor is incorporated, but that's not the case. Check out those simulations where you build a simple robot and it tries to learn to walk based on your design. It's the same concept. With self driving cars, the fitness algorithm is much more complex, but the fundamentals are the same.

And as far as the range of sensors these cars have, once the software know how to use them correctly, no human driver will ever match the software.

That is why tram will work. It doesn't have a steering wheel, this sensors can be as simple as ultrasonic parking sensors, infrared/laser distance measuring, and couple of cameras.
They don't need GPS since you can embed tags into rails, which will tell tram its precise position on rails. Since it is rails - there is signalling is possible.
And in driving code (at least in my shithole) it is comparable to ambulance and cops car, untouchable thing.

Just stop, we are nowhere near any of this working on a wet road much less in bad driving conditions. You seem to be living on a completely different planet or are just a fucking kid reading press releases from tesla.

If video game developers can have semi-competent driving AI on some background thread, I'm sure we'll be fine with people dedicated to replicating the real world as data via several sensors and interpreting that data in a competent manner with which to make decisions on.

It's not like we haven't been simulating this kind of stuff for decades, and sensor technology just keeps getting better and cheaper.

Right, which is where the fitness algorithms come in. You run the car on wet roads enough times for it to learn what not to do by fucking it up.

I'm not reading Tesla press releases, I'm basing this on how machine learning works, retard. This has more to do with what Google showed with AlphaGo than anything else. Again, the problem is that you have a fundamental gap in knowledge when it comes to how these things work.

you gotta crack a few eggs if you want to make an omelet.

Vehicles are pretty expensive eggs.

On your wallet, maybe? Not very much for those that managed to produce the estimated 1bn passenger cars in use today.

Plus you can start with pure simulations. Introduce external forces simulating wheels loosing traction, winds, cars bumping into the car and whatever else comes to mind & let the car react. It'll be pretty good before real tests produce a few wrecks from which you iron out remaining bugs.

THE AVERAGE COMMUTER SPENDS ~15 DAYS EACH YEAR SITTING IN TRAFFIC GOING TO WORK.

GOOGLE/AMAZON/APPLE/ETC ONLY WANT DRIVERLESS CARS BECAUSE THAT'S 15 MORE DAYS FOR THEM TO BOMBARD YOU WITH ADS AND TIME FOR YOU TO SPEND BUYING MORE SHIT

>just feed it more data XDDDD
just stop it man. Take a class on it or something.

>b-but censors are better than 'humans'
Regardless of the correctness of this statement this is Jow Forums you guys should understand the compromise between safety and freedom. I will never trust a software and internet based vehicle regardless if safety. Between botnet and potential for Cia niggers to execute you there is to much risk

Woke

Normalfags are drooling retards, are you surprised they'd throw away freedoms for mild convenience?

It's hard to care about public transit when it becomes a cesspool of blacks using it to aid in crimes or just as rentals.

Taking public transit really sucks sometimes

because people are retarded and when in control of something with the amount of power a car gives them, they somehow manage to be even dumber than dumb.
also what better way to store peoples lives than to have 50 gorillian cameras on their cars, and a dozen on their spy devic- i mean cellphones?