Uber Troubles For Uber

[YNM]

24 minutes ago, tater said:

Jaywalking is illegal in AZ, and Tempe.

I know, there are puny signs at the site of crash that restricts people from using those median paths and advising to try the crossings at the road junction next to these paths.

I'm just suggesting perhaps they could do more to lure people away from using such an illegal but feasible path, like fencing them, put some bollards, or just remove them altogether.

 

EDIT : Also, just throwing some fun :

 

Edited March 22, 2018 by YNM

[Bill Phil]

3 hours ago, LordFerret said:

This does not matter. Rule one driving across this nation last time I checked, was "pedestrians have the right of way".

And that reality is; As I said before, the pedestrian has the right of way in all circumstances. That is the law.

That's...not correct...

I assume you are referring to the USA. Pedestrians do not always have right of way. It varies by state, but as an example:

https://www.tn.gov/tdot/multimodal-transportation-resources/bicycle-and-pedestrian-program/tennesse-pedestrian-laws.html

That's Tennessee, not Arizona, so the laws will be different, but from what I can find pedestrians are supposed to yield to cars when outside of cross-walks... but don't trust me on that one.

[tater]

Yeah, right of way varies, and most pedestrians don't actually even know the laws for their own States and municipalities (which can also vary within States).

The critical issue is that the majority of fatal ped vs MVAs are outside of designated crossing areas, and at night, combined with the fact that in the US fully 80% of such deaths deem the pedestrian at fault.

Humans in all cases except this one have been the drivers.

By definition, there is nothing to do to solve this issue on the planning end. You can claim this one, isolated spot should have had a crosswalk (for people too lazy to walk 100m north of where she jaywalked), but then people will apparently ignore THAT crosswalk, and walk between the site of her death, and the old crosswalk. It becomes a Zeno's Paradox of crosswalks.

Since the meat drivers are also baked into these stats, it strikes me that the low-hanging fruit to mitigate pedestrian deaths due to cars---even in the majority case where the pedestrian is the one at fault---might be to make super safe, self-driving cars. Maybe this death, not the fault of Uber, could result in countless saved lives, as they decide to add alternate sensors to their vehicles.

There is a technical issue here, however, which will result in these deaths never being zero.

The "thinking time" (sensing a hazard, and acting on that data) can be reduced to tiny fractions of a second, but the vehicle still needs to interpret what's a threat, and what isn't. In some cases it's relatively easy, because the ped's vector is along the sidewalk, so it is reasonable to assume they will not leap left into traffic. None the less, they might, and within a certain range, that would be fatal. I have had school kids fake jumping into the street in front of cars to spook them before, which is pretty annoying, and should likely be criminal if it is not already. Doing this to self-driving cars will likely cause traffic jams as they do emergency stops, and a wave moves through traffic in reaction. Of course as all have cameras, the perps in this case could likely be charged, every car is also an array of spy cams.

[YNM]

3 hours ago, tater said:

By definition, there is nothing to do to solve this issue on the planning end. You can claim this one, isolated spot should have had a crosswalk (for people too lazy to walk 100m north of where she jaywalked), but then people will apparently ignore THAT crosswalk, and walk between the site of her death, and the old crosswalk. It becomes a Zeno's Paradox of crosswalks.

... Or remove every crosswalk features from the place you don't want people to walk onto.

They're smart enough to know to fence the one at the other side of the river.

It isn't rocket science (I'm definitely not learning one).

4 hours ago, Bill Phil said:

I assume you are referring to the USA. Pedestrians do not always have right of way.

Small wonder you drive down a 30sec walk distance.

Edited March 22, 2018 by YNM

[LordFerret]

7 hours ago, Bill Phil said:

That's...not correct...

I assume you are referring to the USA. Pedestrians do not always have right of way. It varies by state, but as an example:

https://www.tn.gov/tdot/multimodal-transportation-resources/bicycle-and-pedestrian-program/tennesse-pedestrian-laws.html

That's Tennessee, not Arizona, so the laws will be different, but from what I can find pedestrians are supposed to yield to cars when outside of cross-walks... but don't trust me on that one.

Having read though that, I understand the point. It's not so far from the laws defined here (NJ). However, here is the issue at hand...

Quote

TCA 55-8-134 - Pedestrian's Right-of-Way in Crosswalks
  ... snip ...
(b) No pedestrian shall suddenly leave a curb or other place of safety and walk or run into the path of a vehicle which is so close that it is impossible for the driver to yield.

This is not what happened. Clearly the woman was more than half-way across the road when the Uber happened upon her. In any state in the union, it should be (and I think ultimately will be) on the Uber.

I will imagine that most states have a similar set of laws. Still, I can't see any of them putting more weight on 'rights' to the vehicle over the pedestrian. I'm tempted to draw a parallel here of boats out on the water, powered verses sails... if you know what I mean.

7 hours ago, tater said:

The critical issue is that the majority of fatal ped vs MVAs are outside of designated crossing areas, and at night, combined with the fact that in the US fully 80% of such deaths deem the pedestrian at fault.

While this might be true, it's my understanding that the liability is still placed on the car/driver. An examination of insurance payouts would emphasize that, and charges on record against the drivers. I don't recall anywhere that a pedestrian struck having their insurance to pay out for damages their body did to the vehicle, or face charges for same. I would imagine that if such has happened, the situation would be highly unusual - and probably an interesting read.

[tater]

1 hour ago, LordFerret said:

This is not what happened. Clearly the woman was more than half-way across the road when the Uber happened upon her. In any state in the union, it should be (and I think ultimately will be) on the Uber.

She put herself in front of a car. Doesn't matter when she left the curb, she timed it very, very poorly (self evidently).

She could see hundreds of feet in the direction the car was coming from.

I'll ask my friend who is a lawyer what he thinks about the liability.

Edited March 23, 2018 by tater

[micha]

12 hours ago, YNM said:

Also, jaywalking isn't a crime in the UK. Try to think about it.

Which also puts the onus on pedestrians to ensure they do so in a safe manner!  It doesn't mean pedestrians can just jump out into the road at a moments notice and expect all traffic to magically stop/avoid them.

Whether or not jaywalking is a crime is immaterial; the point is that IF you jaywalk, you do so safely.

Heck, even if not jaywalking, I always check to ensure approaching traffic will stop! There's plenty of cars which run red lights or ignore pedestrian crossings.

12 hours ago, YNM said:

2. You expected the driver to make mistakes. You don't expect the pedestrian to make mistakes. Both are human. What is that ?

I expect -everyone- (me included) to make mistakes.  I ride a motorbike, mostly for pleasure, long distance rides across Europe, but in the last year also for commuting in London.

As a motorcyclist I may have a lot of "rights" but I also know that in the event of an accident, I will always lose in a collision with another vehicle.  Ditto if I'm a cyclist or a pedestrian. As such, I will waive my "rights" in favour of avoiding such a collision. I don't want to be right and dead (or seriously injured).

The more vulnerable you are in a particular situation the more care you have to take. If you don't take care, you will get hurt, sooner or later.

 

So yes, I expect both to make mistakes, but I also expect the more vulnerable party to exercise higher vigilance.

 

Ultimately in this particular instance there was some sort of failure of the automated vehicles sensors / algorithm; it really ought to have detected something on the road ahead and reacted accordingly. Any alert human driver would have.

However, given the fact that this technology is still in its infancy and there was a human supervisor, I would place the blame of this particular incident primarily on the shoulders of the human supervisor who was not supervising but playing with her phone, and secondarily on the pedestrian who was crossing illegally without due care.  Let's see what the investigation will conclude.

 

[YNM]

1 hour ago, micha said:

Which also puts the onus on pedestrians to ensure they do so in a safe manner!  It doesn't mean pedestrians can just jump out into the road at a moments notice and expect all traffic to magically stop/avoid them.

Have you tried one in a 20 zone ? Would you try one on the A40 Westway, or the A13 proper, or the A102 or A12 Blackwall Tunnel Approach? (you've been to London, I presume you know.)

 

1 hour ago, micha said:

The more vulnerable you are in a particular situation the more care you have to take. If you don't take care, you will get hurt, sooner or later.

I thought it was "with great power, comes great responsibility".

Driving a car means having the power of blow from at least a tonne of steel.

Down here, if you're involved in an RTA with a road user of "lower class", and you report it to Police, the "higher class" always lost, no matter what, no matter who (alright, apart from if you have buckloads of money). Even if there was no direct hit involved, you have to help them. That is humane, that is sensible. (fortunately I haven't involved in one when I drive cars.)

I also drive motorcycles, I have to say sorry for every car which I might have upset.

I've been on both positions. I know my limits.

 

ADDEN. :

Now, you may ask, "how do I go long distances on cars without taking too long a time" ?

The answer is "get a special road". What might be termed "motorways" or "expressways". Those are for cars, you won't see a speck of cyclist on it.

In the US, this translate to Instertates, US Routes, State Routes with grade separation, etc. You can go at any speed on them (I know the limit is 65 in towns and 75 out town in AZ. Last I remember it's the same in NM. In CA I believe it's 70. But even then who bloody cares when you're in the middle of absolute nowhere and is guaranteed to be safe from mortals). Down here it's usually tollroads. We have a near-blanket 50 mph (80 kmh) limit on them, but even my father and a lot of other people don't mind to just go at 75 mph (120 kmh) on them - they're purpose-built for cars. I don't mind going as quick on those either.

Your congested town can't afford one ? Bad luck on dreaming cars.

Edited March 23, 2018 by YNM

[micha]

13 hours ago, LordFerret said:

Mathematics & sentience... DNA? Hummm. YOU, your entity, soul, what-have you, that energy flowing about in your head that is you and your thoughts, is not DNA. That DNA is just part of the physical construct which houses. That energy flowing about - that's the manifestation of the Mathematics I'm talking about... or, vice versa.

Hm. Ultimately in both cases it's just energy. In the case of biology, electromagnetic and chemical, in the case of AI, presumably, just electromagnetic.  So I'm not entirely sure about the concept of Mathematics manifesting, or at least, how it would differ from the current biological concepts of a "self".  You can ultimately also describe biology in terms of physics and hence Mathematics...

Also, Mathematics is just an artificial tool invented by humans to aid in describing the world around us. At best it's an approximation of the real physical world (* assuming we don't live in a simulation); a very good and very useful approximation which allows us to predict as-yet unobserved effects of reality, but ultimately, it's an artificial simplification of what's really going on.

1 hour ago, YNM said:

Have you tried one in a 20 zone ? Would you try one on the A40 Westway, or the A13 proper, or the A102 or A12 Blackwall Tunnel Approach? (you've been to London, I presume you know.)

I don't follow your question.

Yes, in a 20 zone it's safer to jaywalk than on a highway, but you're still dead if you walk out in front of a bus without giving it enough space to stop.

1 hour ago, YNM said:

I thought it was "with great power, comes great responsibility".

Driving a car means having the power of blow from at least a tonne of steel.

Sure; but we already established that accidents will happen. We're human. We make mistakes. Yes I have responsibility as a driver, but I ALSO have responsibility as a pedestrian.

Despite which I was talking from a personal perspective - it's my responsibility to avoid putting myself into situations which have a high risk of getting me injured or killed.

1 hour ago, YNM said:

Down here, if you're involved in an RTA with a road user of "lower class", and you report it to Police, the "higher class" always lost, no matter what, no matter who (alright, apart from if you have buckloads of money). Even if there was no direct hit involved, you have to help them. That is humane, that is sensible. (fortunately I haven't involved in one when I drive cars.)

Of course you're "right" and the "higher class" will lose a court case.  Not much help though if you're dead.  I meant losing in the sense of losing physically, not legally. It doesn't matter a pair of fetid dingoes kidneys whether you win a court-case, except maybe to the lawyer (your family will also get a payout, but that will never replace losing you).

[YNM]

1 hour ago, micha said:

Yes, in a 20 zone it's safer to jaywalk than on a highway, but you're still dead if you walk out in front of a bus without giving it enough space to stop.

... which is a really small gap for lethal effects in a 20 zone.

I should remind you that this was in a 40-legalese zone. That's twice the speed - four times the energy.

1 hour ago, micha said:

Not much help though if you're dead. 

And that is the goal - avoid getting people killed. I'm speaking from the angle of road design and safety/risk engineering here.

Sure, you have a few options :

- Make every pedestrian wear hi-viz (PPE)

- Enforce jaywalking laws legally (Admin controls)

- Enforce anti-jaywalking by fencing (engineering controls)

- Entirely separate paths and road for pedestrians + unmotorized vehicles and motorized vehicles (Nearly elimination)

As you can see on the list, ideally you'd want the top most option (near-elimination). But as you have mentioned, they also have something to do with cost.

Given that Administrative Controls are in place already, your option is to elevate one level - Engineering Controls.

I don't say that the entirety of it has to be implemented - this is entirely based on what levels you see appropriate. But if current measures are inadequate, elevating further is almost mandatory.

 

EDIT : And for the AI... They're still "programmed" by "gullible" humans.

Edited March 23, 2018 by YNM

[micha]

1 hour ago, YNM said:

And that is the goal - avoid getting people killed. I'm speaking from the angle of road design and safety/risk engineering here.

Which has been discussed ad nauseum here and you just won't accept the fact that increasing engineering controls to reduce risk is just not possible everywhere.  You have to do a cost/risk analysis - the highest risk areas get fixed, and the lowest-risk areas don't.  And I strongly suspect that this particular road section just doesn't factor into the high-risk bracket.

It was an accident, and we'll have to wait where the blame will be assigned by the investigators.

 

In the meantime, there -is- an effort to implement an engineering solution which will NOT require fixing millions of miles of roadways, but instead by removing or reducing one of the fallible factors - humans!

1 hour ago, YNM said:

EDIT : And for the AI... They're still "programmed" by "gullible" humans.

Which is true. HOWEVER, humans will program the AI as an "expert system", that is, they take the BEST human drivers as input into the system. Once the system becomes an "expert" driver (and the bugs are ironed out and the hardware has sufficient redundancy), it will be BETTER than a human because, unlike a human, it will NEVER get distracted or sleepy or otherwise run under impaired conditions. It can have better sensors, it WILL have much better reaction times, etc etc.

Will accidents still happen? Of course - whenever unpredictable behaviour is such that it exceeds the physical capabilities (eg, a human stepping out in front of it where physics dictates that it cannot stop in time).

But it SHOULD reduce the number of accidents significantly. The #1 factor in accidents is humans.

[YNM]

2 hours ago, micha said:

Which has been discussed ad nauseum here and you just won't accept the fact that increasing engineering controls to reduce risk is just not possible everywhere.

I digress. I won't have a future job otherwise. XD

As you said, it could start from where accidents are common.

2 hours ago, micha said:

HOWEVER, humans will program the AI as an "expert system", that is, they take the BEST human drivers as input into the system.

They had someone just like that from long ago XD

2 hours ago, micha said:

Once the system becomes an "expert" driver (and the bugs are ironed out and the hardware has sufficient redundancy)...

When can we expect that with human programmers ? XD

[LordFerret]

10 hours ago, micha said:

You can ultimately also describe biology in terms of physics and hence Mathematics...

At best it's an approximation of the real physical world (* assuming we don't live in a simulation);

Everything in existence is describable as a waveform. Yay Fourier.

That's a very interesting concept which I've heard before, and seen many a video about on YouTube. It makes a compelling argument.

[tater]

A few things I found out. Uber uses lidar, so unsure why it didn’t see bike. 

Also, talked to friend (attorney I mentioned above) about Uber’s exposure. He said they likely get reamed, as juries don’t particularly care what the law is when they have sympathy. He focused on the chick (I thought it was a dude) who was supposed to be ready to take over. Had she been doing her job, much better for Uber to claim it was unavoidable since even with lidar it doesn’t notice her. Since she was explicitly being irresponsible, huge problem. He said he’d ask for other self driving vids to show that Uber knows that their human driver backups are never doing their jobs, so they knew.

There are already car sensors that can ID a driver closing their eyes, for example, so Uber could easily have the car prompt the safety person to get off their phone.

He said here in NM, the jury would assign relative responsibility for the accident. The woman hit is clearly partially at fault, so they’d likely say most fault was hers, but a human might have avoided the crash anyway, hence XX% Uber’s fault. Since it was a death, it might be X% of a very large number of dollars.

Edited March 23, 2018 by tater

[Shpaget]

You don't need an expert human driver to make an expert computer driver.

You can take a team of above average programmers and let them loose. The first iteration of the software will not be perfect, but after each error in judgement while doing simulations you can analyze it and find not only a fix, but the best fix. That is something that a human can never accomplish. They only ever get one chance to make a decision in a real life situation. Simulations give the opportunity to go back and try another approach, and another, and another until you find the best one. Then you propagate the update to all instances of your test fleet and simulate more. At some point the software becomes better at driving than the humans programming it.

Sure, there will always be a possibility of a catastrophic bug that results in a crash, but that happens even with "stupid" cars that are standard today, and happened with entirely mechanical cars before them, and happened with people walking where their shoe straps broke on their sandals and they fell down a hill and broke their neck.

All that matters is that AI provides the possibility of reducing the consequences of human error, which is the most common cause of accidents (drink driving, speeding, texting...).

[kerbiloid]

Run a GTA (or much better - Saints' Row) championship for bots. Objective - to hit as many NPC as possible.
Select the worst of them. Let them write AI.
 

[YNM]

9 hours ago, LordFerret said:

Everything in existence is describable as a waveform. Yay Fourier.

... Or Fourier just approximates the real world. This is at least true in engineering - everything is just an approximation and you need to prepare for the absolute perceivable worst.

9 hours ago, tater said:

Uber uses lidar, so unsure why it didn’t see bike. 

Maybe it can't detect stick figures ? XD

9 hours ago, tater said:

There are already car sensors that can ID a driver closing their eyes, for example, so Uber could easily have the car prompt the safety person to get off their phone.

That would be a better stuff...

8 hours ago, Shpaget said:

The first iteration of the software will not be perfect, but after each error in judgement while doing simulations you can analyze it and find not only a fix, but the best fix. That is something that a human can never accomplish. They only ever get one chance to make a decision in a real life situation.

... Unless simulations doesn't exactly equal out the real world. And whoever that made the simulation must be aware of how intricate real world can be.

And given our roads for the foreseeable future features human-vehicle interactions, how do you simulate that ? (example) (more example)

 

This is why, at least as someone who aspires to be an engineer in the future, I choose to protect people no matter what's the condition, no matter who drives, no matter where's the location. Choosing a safety precaution that works for everyone is better than having a safety precaution that not only works for one part of the equation, but also is a new unknown variable.

Edited March 24, 2018 by YNM

[tater]

It will never be about protecting everyone. There will always be times when there is a trade off. This is true of the tech, as well as for broader planning issues. Ultimate safety might require vehicles that are slower than pedestrians, for example. Sucks when you need to get to the hospital in less time than walking. Everything is always a trade off.

[YNM]

1 hour ago, tater said:

There will always be times when there is a trade off.

B0llocks. Absolute b0llocks.

If something much, much heavier and faster can leave less fatalities, why not something that tends to be slow and less heavy be more dangerous ?

[tater]

Rail? LOL.

Rail is never cost effective. The only rail that doesn't lose money is cargo, and engineering requires making things cost effective. Or you trade off that, and everyone gets to live in tiny houses, because all their money goes to taxes to subsidize the rail, etc. Trade offs.

I like my large house, relatively far away from other people, for example. I pay enough taxes, I don;t want to pay more for rail that I can't use, anyway (unless they erect a cog railway up to my house, the grade is too steep otherwise).

[LordFerret]

13 hours ago, tater said:

A few things I found out. Uber uses lidar, so unsure why it didn’t see bike. 

Also, talked to friend (attorney I mentioned above) about Uber’s exposure. He said they likely get reamed, as juries don’t particularly care what the law is when they have sympathy. He focused on the chick (I thought it was a dude) who was supposed to be ready to take over. Had she been doing her job, much better for Uber to claim it was unavoidable since even with lidar it doesn’t notice her. Since she was explicitly being irresponsible, huge problem. He said he’d ask for other self driving vids to show that Uber knows that their human driver backups are never doing their jobs, so they knew.

There are already car sensors that can ID a driver closing their eyes, for example, so Uber could easily have the car prompt the safety person to get off their phone.

He said here in NM, the jury would assign relative responsibility for the accident. The woman hit is clearly partially at fault, so they’d likely say most fault was hers, but a human might have avoided the crash anyway, hence XX% Uber’s fault. Since it was a death, it might be X% of a very large number of dollars.

The missed by Lidar issue will compound this, for sure. Not good for Uber.

I kind of figured your friend would take that position. Others I know personally here, all in various levels of law practice, have stated pretty much the same. Not good for Uber. I was equally surprised to find the driver was female, I thought it a male... insert rude comment about appearances here. I'd also heard (but did not want to bring up) about their monitor drivers being lax in their duties. I'm not really surprised by this, looking at the caliber of employee involved. I'm not sure how this case will play out. I can imagine Uber will use it's financial might to try and worm its way out of as much liability as possible. For sure, what does happen next will likely set precedent... this is just the beginning.

 

12 hours ago, Shpaget said:

You don't need an expert human driver to make an expert computer driver.

You can take a team of above average programmers and let them loose. The first iteration of the software will not be perfect, but after each error in judgement while doing simulations you can analyze it and find not only a fix, but the best fix. That is something that a human can never accomplish. They only ever get one chance to make a decision in a real life situation. Simulations give the opportunity to go back and try another approach, and another, and another until you find the best one. Then you propagate the update to all instances of your test fleet and simulate more. At some point the software becomes better at driving than the humans programming it.

Sure, there will always be a possibility of a catastrophic bug that results in a crash, but that happens even with "stupid" cars that are standard today, and happened with entirely mechanical cars before them, and happened with people walking where their shoe straps broke on their sandals and they fell down a hill and broke their neck.

All that matters is that AI provides the possibility of reducing the consequences of human error, which is the most common cause of accidents (drink driving, speeding, texting...).

I really have to jump on this. THIS is exactly the problem with nearly all software development today... namely, it doesn't happen. You can use KSP as an example, you can use any Microsoft product as example, you can use just about anything there is out there save for maybe military / financial / medical mission critical systems... and that is - testing and debugging and trials are thrown onto the unsuspecting unknowing public. It's as bad as or worse than the pharmaceutical industry... of which I spent over 12 years in as a consultant**. People simply do not shoot for 100% anymore, not like was done back in the early days (before there was a PC). Today, stuff gets shoved out the door at 75% to 80%, and the rest is worried about in light of 'updates'... the Microsoft school of product development. If Uber, or any of them, were to do it right, they'd have a simulation village/town, complete with every manner of road and highway, with a very dedicated team of researchers and volunteers setting up and playing out all manner of real life scenarios while putting their AI cars through the paces. To the best of my knowledge, no such place exists in the Uber design / manufacturing arena.

**
Alpha, Beta, and Clinical Trials. Today, it's all a 'fast track' through FDA approval (here in the USA anyway, used to be 12 years, now, I think it's 7 years). Next time you get yourself a prescription drug, read the PDR insert and make note about the statistics pertaining to Clinical Trials and Post Clinical Trials... check the numbers and see just how many have been tested. You're the testee.

 

Back on topic...
While not entirely pertaining to AI-driven cars, Uber and Lyft (and other's I'm sure) are not without their fair share of issues. We touched briefly on this previously (@YNM), the idea of thugs purposely stopping AI cars to rob the passengers. This is something we're going to see more and more of, not to mention the potential of hacked / stolen (car-jacked) AI vehicles.

http://www.whosdrivingyou.org/rideshare-incidents

 

[YNM]

7 hours ago, tater said:

Rail? LOL.

Then why are rails safer than highways then ? What can we learn from rail operations ? Is there absolutely none ?

 

6 hours ago, LordFerret said:

... The idea of thugs purposely stopping AI cars to rob the passengers. ...

Keep the doors locked, when car windows broken call 911 and immediately leave.

 

TBH, you just need to "hack" one. Putting something on any form of digital network means exposing it to threats from around the world, 24/7. "Digital/cyber" means time and space barriers becomes meaningless, capital barriers also goes massively down. Given the  "deadly" results, imagine who will be interested in hacking these stuff.

After all, we have better thugs down here (bajing loncat / "flying squirrels") that would puposefully jump down tree branches over the road onto moving vehicles to rob them. (I know you don't have one there because you don't have dense covers and you have tunnels.)

 

 

EDIT :

As Randall Munroe pointed, today's physical environment involves a very important key aspect - (mostly) nobody's any murderous. I'm not sure this is also the case in cyber world...

Edited March 24, 2018 by YNM

[kerbiloid]

9 hours ago, tater said:

Rail is never cost effective.

Really?
You (USA) just have a lot of roads for cars and several private railways differing even with standards. So, definitely it's cheaper to rent a caravan of trucks rather than get a headache with rail exotics.
 

Spoiler

Optimal kind of transport depending on distance.

авиа = aviation
авто = truck
ж/д = railway

Left  column - "distance, km"
Header row - "mass of cargo, kg"

 

Edited March 24, 2018 by kerbiloid

[tater]

24 minutes ago, kerbiloid said:

Really?
You (USA) just have a lot of roads for cars and several private railways differing even with standards. So, definitely it's cheaper to rent a caravan of trucks rather than get a headache with rail exotics.

Read what I said.

"Rail is never cost effective. The only rail that doesn't lose money is cargo, "

Cargo rail is great. Passenger rail is almost always subsidized. A few commuter lines here and there (places like NYC commuter rail) do OK with huge ridership.

5 hours ago, YNM said:

Then why are rails safer than highways then ? What can we learn from rail operations ? Is there absolutely none ?

 

For people doing their day to day stuff that they use cars for in the US? Rail teaches us absolutely nothing at all.

Edited March 24, 2018 by tater

[kerbiloid]

15 minutes ago, tater said:

Passenger rail is almost always subsidized

When you already have a huge cargo railway network, it's cheaper to subsidize passenger trains rather than make a parallel network of roads.