1990: "Driving Cars"

This forum is for the individual discussion thread that goes with each new comic.

Moderators: Moderators General, Prelates, Magistrates

DavidSh
Posts: 129
Joined: Thu Feb 25, 2016 6:09 pm UTC

Re: 1990: "Driving Cars"

Postby DavidSh » Tue May 08, 2018 1:00 pm UTC

Clearly a randomized double-blind trial of the value of driving practice/instruction is called for. The control group would have to have some kind of placebo practice/instruction.

User avatar
Heimhenge
Posts: 231
Joined: Thu May 01, 2014 11:35 pm UTC

Re: 1990: "Driving Cars"

Postby Heimhenge » Tue May 08, 2018 3:29 pm UTC

Got my license in Wisconsin. We took a drivers ed course in high school ... I think sophomore year. Included this cool "box" you sat in for a car simulation with a rudimentary dashboard. Not a real simulator though. It was just for measuring reaction speed. When the red light on the dash came on, you were supposed to move your foot from the gas pedal to the break pedal, and it would tell you how many milliseconds it took for you to react. Not sure how the red light was randomized, but me and my buddies would go in there after class and try to "game" the thing, anticipating when the light would come on, and compete for the lowest time.

After that in-school course, you still needed to pass a written test and a driving test with a licensed examiner.

DanD
Posts: 304
Joined: Tue Oct 05, 2010 12:42 am UTC

Re: 1990: "Driving Cars"

Postby DanD » Tue May 08, 2018 3:57 pm UTC

DavidSh wrote:
petercooperjr wrote:I do suspect that if at the time cars were first invented everybody knew how many injuries and fatalities they would cause, they'd never have been allowed. Somehow it all came upon everybody slowly enough that it's considered acceptable risk to drive now, and of course cars have been getting safer and safer over the years, but it is kind of bizarre how we got here.


Does anybody know the fatality rate per passenger mile of horses, or horse-drawn vehicles? Not as fast as automobiles, typically, but control of a large animal isn't certain.


I don't know the numbers, but horses don't crash when their driver gets tired. Horses are fairly good at avoiding collisions with other horses, unless their operator is doing stupid things like charging at other horses while carrying a pointed stick. As a rule, I'm guessing horses are safer per passenger mile.

User avatar
zemerick
Posts: 25
Joined: Thu May 12, 2011 6:01 am UTC

Re: 1990: "Driving Cars"

Postby zemerick » Tue May 08, 2018 4:28 pm UTC

sonar1313 wrote:
somitomi wrote:Humans are in fact outstanding at driving - so good, that we've become very overconfident, which is the real cause of most human error behind the wheel (drinking/texting and driving, for example.)


I very strongly disagree with this. Humans are absolutely atrocious drivers.

We forget a huge amount of the rules.
Have slow reaction times.
Only manage 1 thing at a time. ( True multi-tasking is very rare, and even that might just be very fast task switching and good management. )
Vision is heavily designed to ignore/fake vast amounts of the information it takes in and has large blind spots.
Only look in a single direction at a time.
Have a strong tendency to overreact in stressful situations.
Only see in a very narrow range of light, which mostly disappears at night, and many things are similar colored.
Easily get relaxed, losing focus.
Easily distracted.
Are overconfident as you mentioned.

Consider this: All driving exams and tests are extremely thorough, require 100% to pass, and are required every year. If you are ever found at fault in an accident, your driving privilege is revoked. If you are ever caught breaking any law ( even just rolling through a stop sign, clipping a solid line, speeding etc. ), your privilege is revoked.

In just 5 years, how many people still have a license?

It isn't just over confidence. We are intrinsically bad at driving. This is why even quite early, simple self-driving technologies like the first iteration of Teslas Autopilot saw a 50% decrease in accident rates. Those technologies still aren't good at driving either..but they're already WAY better than we are. The real kicker though? We're getting WORSE at driving ( Despite all of the previously mentioned safety improvements to vehicles, including newer autodriving systems, fatality rates are on the rise. ), while auto systems are getting better at it.

I have never once, in my entire life, met a single person that can drive for any sort of reasonable distance without making a mistake. No matter how attentive and careful they are.

wumpus
Posts: 531
Joined: Thu Feb 21, 2008 12:16 am UTC

Re: 1990: "Driving Cars"

Postby wumpus » Tue May 08, 2018 4:49 pm UTC

Some thoughts:

Do ultralights still not require a license (and thus age)? The first death in my high school class (@16) died this way...

Pierre Currie died in a horse carriage accident. Presumably these were not unknown. Then there was chariot racing that was presumably entertainment similar to gladitorial games (don't accept chariot races, that's how you get Byzantine politics) [PS: don't expect chariots to be all that fast, the horse harness wouldn't be invented until the middle ages and they were using ox harnesses. A ox harness will strangle a horse leading to the dramatic depiction of chariot races that remain in Roman art. So maybe they weren't quite as dangerous as we think].

https://en.wikipedia.org/wiki/File:US_t ... deaths.png
this is an amazing chart of everything you need to know about highway safety in the US. I find the bottom red line to be the most useful for "safety while driving". Some notes:
That big drop around 1975 has to be the 55mph speed limit. While it was common belief then (and now) that 55 was certainly way too slow, don't think that 1975 brakes/tires could stop anywhere near as quickly as modern silicate tires with anti-lock brakes.
That big drop around 1980 has to be thanks to MADD (and possibly extends to 1990). Whatever you might think about them as neoprohibitionists now, they were needed in 1980.
The steady decline between 1985 and 1991 is a bit harder to figure. Seatbelt increase is hard to determine (an entire federal agency exists to promote it, they can't admit that people use seatbelts...), air bags came into play as well as antilock brakes. Safety also started creeping into advertising (you better have 4 stars or families won't buy your minivan).
The little drop right at 2009 coincides with cash for clunkers. Presumably "clunkers" weren't all that safe.

Leovan
Posts: 71
Joined: Mon Dec 05, 2016 9:31 pm UTC

Re: 1990: "Driving Cars"

Postby Leovan » Tue May 08, 2018 5:01 pm UTC

DanD wrote:I don't know the numbers, but horses don't crash when their driver gets tired. Horses are fairly good at avoiding collisions with other horses, unless their operator is doing stupid things like charging at other horses while carrying a pointed stick. As a rule, I'm guessing horses are safer per passenger mile.


As a rider, I disagree with your conclusion. Yes, the horse usually stays on the path when you don't pay attention, but it also suddenly accelerates when a bird flies by, suddenly stops and dips forward when a tuft of grass looks tasty, and I've seen two horses decide they really need to stop and roll while the rider is still on them. I've fallen off horses a lot more than I've had car or motorcycle accidents of any kind. A horse is an animal with a mind of its own and when you're riding you have to pay more attention to your surroundings than while driving, because you have to analyze everything not just for its trajectory and possible intentions, but also for the possible reactions of your horse.
I'd also recommend motorcycles for driver's ed. There's no better inspiration to learning how to judge traffic and practice defensive driving than knowing you're half their size, have no protection, and they might not see you. My sister had a moped when she was 14 and somebody ran her over because he didn't see her standing there waiting to turn left (despite headlight and blinker).
Plus if you don't see a rock on the road you're dead. Helps weed out bad drivers.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Tue May 08, 2018 5:03 pm UTC

zemerick wrote:
sonar1313 wrote:
somitomi wrote:Humans are in fact outstanding at driving - so good, that we've become very overconfident, which is the real cause of most human error behind the wheel (drinking/texting and driving, for example.)


I very strongly disagree with this. Humans are absolutely atrocious drivers.

We forget a huge amount of the rules.
Have slow reaction times.
Only manage 1 thing at a time. ( True multi-tasking is very rare, and even that might just be very fast task switching and good management. )
Vision is heavily designed to ignore/fake vast amounts of the information it takes in and has large blind spots.
Only look in a single direction at a time.
Have a strong tendency to overreact in stressful situations.
Only see in a very narrow range of light, which mostly disappears at night, and many things are similar colored.
Easily get relaxed, losing focus.
Easily distracted.
Are overconfident as you mentioned.

Consider this: All driving exams and tests are extremely thorough, require 100% to pass, and are required every year. If you are ever found at fault in an accident, your driving privilege is revoked. If you are ever caught breaking any law ( even just rolling through a stop sign, clipping a solid line, speeding etc. ), your privilege is revoked.

In just 5 years, how many people still have a license?

It isn't just over confidence. We are intrinsically bad at driving. This is why even quite early, simple self-driving technologies like the first iteration of Teslas Autopilot saw a 50% decrease in accident rates. Those technologies still aren't good at driving either..but they're already WAY better than we are. The real kicker though? We're getting WORSE at driving ( Despite all of the previously mentioned safety improvements to vehicles, including newer autodriving systems, fatality rates are on the rise. ), while auto systems are getting better at it.

I have never once, in my entire life, met a single person that can drive for any sort of reasonable distance without making a mistake. No matter how attentive and careful they are.


And yet, despite all that, we get from place to place every single day almost exclusively without running the big machine into stationary objects or other big machines.

Yes, we forget rules sometimes. On the other hand, we have an incredibly complex system of rules both written and unwritten that change between jurisdictions and sometimes between time of day and various situations, and we have the ability to adapt those rules as needed to make things work. If the functioning traffic light in front of us turns off suddenly, we know immediately what to do.

Yes, we also break rules sometimes. We have the judgment to decide whether we can do it safely. If I'm behind a slow-moving truck and can safely change lanes over a solid white line, I generally do it. So do we all.

Our vision and other physical traits hardly matter. We have set up the rules of the road, and the roads themselves, so that our vision can handle it. We are not trying to adapt to the jungle, we have made our own jungle, adapted to us.

"Better at driving" does not only mean "safer." If a self-driving car adds 15-20 minutes to my commute (30-40 per day) because it stubbornly follows every rule and puts up with nonsense that I don't, that's not better. If that's the case then I'm better than the self-driving car. I can make a split-second decision to change my route. I can pass a truck where a self-driving car might decide to just sit happily behind it. I'm familiar with the traffic light patterns. I know where the potholes are. And if my somewhat more aggressive driving (as compared to a self-driving car) doubles the very, very minuscule chance of getting in an accident, twice a very small number is still a very small number. I've been commuting for 14 years, so roughly 3,500 commuting days or 7,000 commutes, and I am 6,999-for-7,000 in accident avoidance. (One time I backed into a slow-moving car as I left my driveway, because I couldn't see around the parked cars on the street. A backup camera probably would've helped.) I'm betting a self-driving car, with its inflexible adherence to rules, inability to optimize the route, and lack of desire to optimize or prioritize traffic management to speed up the drive, would in fact add 15-20 minutes per commute and therefore it would not actually be better at driving.

Besides all that, it's hardly fair to say fatality rates are on the rise. In 2014 they bottomed out at 1.08 per 100 million miles traveled and have slightly risen to 1.18 in 2016. Over any kind of reasonable long-term view, they have been plummeting.

This thread actually illustrates how good at driving people really are. Given the minimal instruction and low standards for licensing, people still manage to get from place to place every single day, all the time, almost exclusively without hitting things. To say we are atrocious drivers implies we can barely get down the block. It ignores the insane amounts of miles traveled without incident.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Tue May 08, 2018 5:09 pm UTC

wumpus wrote:Some thoughts:

Do ultralights still not require a license (and thus age)? The first death in my high school class (@16) died this way...

Pierre Currie died in a horse carriage accident. Presumably these were not unknown. Then there was chariot racing that was presumably entertainment similar to gladitorial games (don't accept chariot races, that's how you get Byzantine politics) [PS: don't expect chariots to be all that fast, the horse harness wouldn't be invented until the middle ages and they were using ox harnesses. A ox harness will strangle a horse leading to the dramatic depiction of chariot races that remain in Roman art. So maybe they weren't quite as dangerous as we think].

https://en.wikipedia.org/wiki/File:US_t ... deaths.png
this is an amazing chart of everything you need to know about highway safety in the US. I find the bottom red line to be the most useful for "safety while driving". Some notes:
That big drop around 1975 has to be the 55mph speed limit. While it was common belief then (and now) that 55 was certainly way too slow, don't think that 1975 brakes/tires could stop anywhere near as quickly as modern silicate tires with anti-lock brakes.
That big drop around 1980 has to be thanks to MADD (and possibly extends to 1990). Whatever you might think about them as neoprohibitionists now, they were needed in 1980.
The steady decline between 1985 and 1991 is a bit harder to figure. Seatbelt increase is hard to determine (an entire federal agency exists to promote it, they can't admit that people use seatbelts...), air bags came into play as well as antilock brakes. Safety also started creeping into advertising (you better have 4 stars or families won't buy your minivan).
The little drop right at 2009 coincides with cash for clunkers. Presumably "clunkers" weren't all that safe.


A possibility: Cars got neutered starting in the late 70s and early 80s as manufacturers scrambled to meet fuel economy requirements the only way they knew how: by killing their horsepower. By 1985 through 1991 most of the big horsepower cars were off the road, replaced by Honda Civics and Plymouth Reliants that wouldn't have known what 85 mph was if you beat them over the head with it.

DanD
Posts: 304
Joined: Tue Oct 05, 2010 12:42 am UTC

Re: 1990: "Driving Cars"

Postby DanD » Tue May 08, 2018 5:28 pm UTC

sonar1313 wrote:
wumpus wrote:Some thoughts:

Do ultralights still not require a license (and thus age)? The first death in my high school class (@16) died this way...

Pierre Currie died in a horse carriage accident. Presumably these were not unknown. Then there was chariot racing that was presumably entertainment similar to gladitorial games (don't accept chariot races, that's how you get Byzantine politics) [PS: don't expect chariots to be all that fast, the horse harness wouldn't be invented until the middle ages and they were using ox harnesses. A ox harness will strangle a horse leading to the dramatic depiction of chariot races that remain in Roman art. So maybe they weren't quite as dangerous as we think].

https://en.wikipedia.org/wiki/File:US_t ... deaths.png
this is an amazing chart of everything you need to know about highway safety in the US. I find the bottom red line to be the most useful for "safety while driving". Some notes:
That big drop around 1975 has to be the 55mph speed limit. While it was common belief then (and now) that 55 was certainly way too slow, don't think that 1975 brakes/tires could stop anywhere near as quickly as modern silicate tires with anti-lock brakes.
That big drop around 1980 has to be thanks to MADD (and possibly extends to 1990). Whatever you might think about them as neoprohibitionists now, they were needed in 1980.
The steady decline between 1985 and 1991 is a bit harder to figure. Seatbelt increase is hard to determine (an entire federal agency exists to promote it, they can't admit that people use seatbelts...), air bags came into play as well as antilock brakes. Safety also started creeping into advertising (you better have 4 stars or families won't buy your minivan).
The little drop right at 2009 coincides with cash for clunkers. Presumably "clunkers" weren't all that safe.


A possibility: Cars got neutered starting in the late 70s and early 80s as manufacturers scrambled to meet fuel economy requirements the only way they knew how: by killing their horsepower. By 1985 through 1991 most of the big horsepower cars were off the road, replaced by Honda Civics and Plymouth Reliants that wouldn't have known what 85 mph was if you beat them over the head with it.


My Toyota Prius, universally ridiculed as "underpowered", drifts above 90 if I don't pay attention. There are very few cars that are "underpowered" for highway use.

A big chunk of the drift down is ABS, Airbags, and, most importantly, crumple zones (to a lesser extent, stronger side impact and roll cage protection). If an 80's model hit something, the car stopped short, the passenger didn't. These days, the first five feet or so of the car crumples, and the passenger decelerates that much slower. Of course, other than ABS, none of those features prevent collisions. They just prevent severe injury or death.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Tue May 08, 2018 5:40 pm UTC

DanD wrote:
sonar1313 wrote:
wumpus wrote:Some thoughts:

Do ultralights still not require a license (and thus age)? The first death in my high school class (@16) died this way...

Pierre Currie died in a horse carriage accident. Presumably these were not unknown. Then there was chariot racing that was presumably entertainment similar to gladitorial games (don't accept chariot races, that's how you get Byzantine politics) [PS: don't expect chariots to be all that fast, the horse harness wouldn't be invented until the middle ages and they were using ox harnesses. A ox harness will strangle a horse leading to the dramatic depiction of chariot races that remain in Roman art. So maybe they weren't quite as dangerous as we think].

https://en.wikipedia.org/wiki/File:US_t ... deaths.png
this is an amazing chart of everything you need to know about highway safety in the US. I find the bottom red line to be the most useful for "safety while driving". Some notes:
That big drop around 1975 has to be the 55mph speed limit. While it was common belief then (and now) that 55 was certainly way too slow, don't think that 1975 brakes/tires could stop anywhere near as quickly as modern silicate tires with anti-lock brakes.
That big drop around 1980 has to be thanks to MADD (and possibly extends to 1990). Whatever you might think about them as neoprohibitionists now, they were needed in 1980.
The steady decline between 1985 and 1991 is a bit harder to figure. Seatbelt increase is hard to determine (an entire federal agency exists to promote it, they can't admit that people use seatbelts...), air bags came into play as well as antilock brakes. Safety also started creeping into advertising (you better have 4 stars or families won't buy your minivan).
The little drop right at 2009 coincides with cash for clunkers. Presumably "clunkers" weren't all that safe.


A possibility: Cars got neutered starting in the late 70s and early 80s as manufacturers scrambled to meet fuel economy requirements the only way they knew how: by killing their horsepower. By 1985 through 1991 most of the big horsepower cars were off the road, replaced by Honda Civics and Plymouth Reliants that wouldn't have known what 85 mph was if you beat them over the head with it.


My Toyota Prius, universally ridiculed as "underpowered", drifts above 90 if I don't pay attention. There are very few cars that are "underpowered" for highway use.

A big chunk of the drift down is ABS, Airbags, and, most importantly, crumple zones (to a lesser extent, stronger side impact and roll cage protection). If an 80's model hit something, the car stopped short, the passenger didn't. These days, the first five feet or so of the car crumples, and the passenger decelerates that much slower. Of course, other than ABS, none of those features prevent collisions. They just prevent severe injury or death.


Comparing cars across 30 years of development is going to get you some odd results. Your Prius almost certainly has more horsepower than what passed for a Dodge Charger in 1985. Those old '80s Civics gave you anywhere from 55 to around 80 hp. There are very few cars that are underpowered nowadays for highway use; the '80s were different.

User avatar
zemerick
Posts: 25
Joined: Thu May 12, 2011 6:01 am UTC

Re: 1990: "Driving Cars"

Postby zemerick » Tue May 08, 2018 6:07 pm UTC

sonar1313 wrote:
zemerick wrote:
sonar1313 wrote:
somitomi wrote:Humans are in fact outstanding at driving - so good, that we've become very overconfident, which is the real cause of most human error behind the wheel (drinking/texting and driving, for example.)


I very strongly disagree with this. Humans are absolutely atrocious drivers.

We forget a huge amount of the rules.
Have slow reaction times.
Only manage 1 thing at a time. ( True multi-tasking is very rare, and even that might just be very fast task switching and good management. )
Vision is heavily designed to ignore/fake vast amounts of the information it takes in and has large blind spots.
Only look in a single direction at a time.
Have a strong tendency to overreact in stressful situations.
Only see in a very narrow range of light, which mostly disappears at night, and many things are similar colored.
Easily get relaxed, losing focus.
Easily distracted.
Are overconfident as you mentioned.

Consider this: All driving exams and tests are extremely thorough, require 100% to pass, and are required every year. If you are ever found at fault in an accident, your driving privilege is revoked. If you are ever caught breaking any law ( even just rolling through a stop sign, clipping a solid line, speeding etc. ), your privilege is revoked.

In just 5 years, how many people still have a license?

It isn't just over confidence. We are intrinsically bad at driving. This is why even quite early, simple self-driving technologies like the first iteration of Teslas Autopilot saw a 50% decrease in accident rates. Those technologies still aren't good at driving either..but they're already WAY better than we are. The real kicker though? We're getting WORSE at driving ( Despite all of the previously mentioned safety improvements to vehicles, including newer autodriving systems, fatality rates are on the rise. ), while auto systems are getting better at it.

I have never once, in my entire life, met a single person that can drive for any sort of reasonable distance without making a mistake. No matter how attentive and careful they are.


And yet, despite all that, we get from place to place every single day almost exclusively without running the big machine into stationary objects or other big machines.

Yes, we forget rules sometimes. On the other hand, we have an incredibly complex system of rules both written and unwritten that change between jurisdictions and sometimes between time of day and various situations, and we have the ability to adapt those rules as needed to make things work. If the functioning traffic light in front of us turns off suddenly, we know immediately what to do.

Yes, we also break rules sometimes. We have the judgment to decide whether we can do it safely. If I'm behind a slow-moving truck and can safely change lanes over a solid white line, I generally do it. So do we all.

Our vision and other physical traits hardly matter. We have set up the rules of the road, and the roads themselves, so that our vision can handle it. We are not trying to adapt to the jungle, we have made our own jungle, adapted to us.

"Better at driving" does not only mean "safer." If a self-driving car adds 15-20 minutes to my commute (30-40 per day) because it stubbornly follows every rule and puts up with nonsense that I don't, that's not better. If that's the case then I'm better than the self-driving car. I can make a split-second decision to change my route. I can pass a truck where a self-driving car might decide to just sit happily behind it. I'm familiar with the traffic light patterns. I know where the potholes are. And if my somewhat more aggressive driving (as compared to a self-driving car) doubles the very, very minuscule chance of getting in an accident, twice a very small number is still a very small number. I've been commuting for 14 years, so roughly 3,500 commuting days or 7,000 commutes, and I am 6,999-for-7,000 in accident avoidance. (One time I backed into a slow-moving car as I left my driveway, because I couldn't see around the parked cars on the street. A backup camera probably would've helped.) I'm betting a self-driving car, with its inflexible adherence to rules, inability to optimize the route, and lack of desire to optimize or prioritize traffic management to speed up the drive, would in fact add 15-20 minutes per commute and therefore it would not actually be better at driving.

Besides all that, it's hardly fair to say fatality rates are on the rise. In 2014 they bottomed out at 1.08 per 100 million miles traveled and have slightly risen to 1.18 in 2016. Over any kind of reasonable long-term view, they have been plummeting.

This thread actually illustrates how good at driving people really are. Given the minimal instruction and low standards for licensing, people still manage to get from place to place every single day, all the time, almost exclusively without hitting things. To say we are atrocious drivers implies we can barely get down the block. It ignores the insane amounts of miles traveled without incident.


Us having built the rules and "the jungle" around our abilities and shortcomings, yet still having any accidents, shows that we are not excellent at driving. We have tried very hard to make it so we don't kill each other all the time, and still manage to. You have about a 1% chance of dying to a car accident, despite the plethora of safety features, regulations, laws, restrictions and requirements.

Lets also not forget, that's JUST fatalities. Injuries is about 1% every single year.

These numbers are also just the US. Many other places are significantly worse, because they rely more on the humans ability to drive, rather than placing heavier restrictions to limit the damage.

Better evidence that people are getting worse, is the number of reported accidents overall, not just fatalities:
2011: 5,338,000
2012: 5,615,000
2013: 5,687,000
2014: 6,064,000
2015: 6,296,000

During this time, miles driven stayed pretty much even. So, the accident rate over just 5 years, jumped by nearly 18 percent.

Of further note, those are just the reported collisions, and are not per vehicle either. ( A collision between 2 vehicles is just 1 collision. 55 car pileup? Still 1 collision. ) It's very hard to get the exact numbers of these incidents. With lots of accidents involving multiple cars, and lots of accidents not reported, it's easy to get to even a 5% chance for an incident every single year.

These aren't insignificant numbers, and halving them is a big deal. Dropping them to 0.1% would be pretty gargantuan. Consider how much money a crash costs. Repairs, lawsuits, insurance, medical, disability, death.

You also forgot something really big about self driving cars. They could easily make it so you get to work faster, not slower. Yes, right now with very early and simple systems, with very few active, they often are a little bit slower. ( Not 20 minutes slower, unless you have a very long commute. Don't forget you can tell these automated systems to do things like change lanes, pass, etc. ) However, imagine how much faster you could go without traffic. Nearly all traffic is caused either by poor human driving habits, or by accidents. We could handle 10 times the number of cars if they were driven by computer systems. Traffic light patterns? Say goodbye to traffic lights. Learning where potholes are in your area? Try knowing where all potholes are everywhere. Traffic management? Try knowing where all traffic is at all times. We could also raise many road speeds to quite significant numbers. Most any road could be raised at least 5mph from where they are if we switched to automated drivers. Those limits aren't because of limits of the vehicles: It's the people that are the problem.

There's also nothing to say you can't take what strengths humans DO have, and add them to automated systems, which as I said we already have. Would the right lane be better for the next 6 miles? Turn on your right blinker, and the car moves itself over. Is that back road actually faster than the main road the GPS chooses automatically? Change the route then. Scared of heights, so you want to avoid that Cliff? Go for it. Speed up? Can do. All of these already exist on Teslas automated systems here and now. You're now combining humans decision making, with the advanced capabilities of the automated system. Let it handle the tedious, mundane, etc. You can focus on the big picture. Everyone wins.

So, yea. We are really bad at driving. Automated IS better than we are at driving, especially when you add in what few pros we have to it.

There is no reason at all to not be pushing automated advancements right now. The sooner we can get switched over to full auto, the better.

chrisjwmartin
Posts: 51
Joined: Wed Jul 30, 2014 8:01 am UTC

Re: 1990: "Driving Cars"

Postby chrisjwmartin » Tue May 08, 2018 6:32 pm UTC

Kilader wrote:Is no one going to bring up the fact that it seems Randall accidentally left a sketch of Cueball turned on? There's a bigger and rougher Cueball slightly transparent, and it's distracting...

Thank god it's not just me - I was worried I had a problem with my monitor!

rabidmuskrat
Posts: 38
Joined: Wed Apr 20, 2011 11:37 am UTC

Re: 1990: "Driving Cars"

Postby rabidmuskrat » Tue May 08, 2018 7:10 pm UTC

chrisjwmartin wrote:
Kilader wrote:Is no one going to bring up the fact that it seems Randall accidentally left a sketch of Cueball turned on? There's a bigger and rougher Cueball slightly transparent, and it's distracting...

Thank god it's not just me - I was worried I had a problem with my monitor!


I can't help but wonder how something like that ends up on there. Does Randall do these comics purely digitally, or does he go pencil/ink on paper to a scanner or w/e?

DanD
Posts: 304
Joined: Tue Oct 05, 2010 12:42 am UTC

Re: 1990: "Driving Cars"

Postby DanD » Tue May 08, 2018 7:12 pm UTC

zemerick wrote: We could also raise many road speeds to quite significant numbers. Most any road could be raised at least 5mph from where they are if we switched to automated drivers. Those limits aren't because of limits of the vehicles: It's the people that are the problem.


It's also possible to take the vehicle into account. A tractor trailer is not safe going 120 on a highway. A low slung car, in good weather conditions, might be. A good automated system can be trusted to make that judgement call realistically, including adjusting for traffic, etc. Therefore automation allows the possibility of eliminating one size fits all speed limits.

rabidmuskrat
Posts: 38
Joined: Wed Apr 20, 2011 11:37 am UTC

Re: 1990: "Driving Cars"

Postby rabidmuskrat » Tue May 08, 2018 7:20 pm UTC

DanD wrote:
zemerick wrote: We could also raise many road speeds to quite significant numbers. Most any road could be raised at least 5mph from where they are if we switched to automated drivers. Those limits aren't because of limits of the vehicles: It's the people that are the problem.


It's also possible to take the vehicle into account. A tractor trailer is not safe going 120 on a highway. A low slung car, in good weather conditions, might be. A good automated system can be trusted to make that judgement call realistically, including adjusting for traffic, etc. Therefore automation allows the possibility of eliminating one size fits all speed limits.


The quirk is you need all vehicles fully automated for that to work out, unless you have a separate speed limit for automated vehicles and manned vehicles.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Tue May 08, 2018 7:22 pm UTC

zemerick wrote:
sonar1313 wrote:
zemerick wrote:
sonar1313 wrote:
somitomi wrote:Humans are in fact outstanding at driving - so good, that we've become very overconfident, which is the real cause of most human error behind the wheel (drinking/texting and driving, for example.)


I very strongly disagree with this. Humans are absolutely atrocious drivers.

We forget a huge amount of the rules.
Have slow reaction times.
Only manage 1 thing at a time. ( True multi-tasking is very rare, and even that might just be very fast task switching and good management. )
Vision is heavily designed to ignore/fake vast amounts of the information it takes in and has large blind spots.
Only look in a single direction at a time.
Have a strong tendency to overreact in stressful situations.
Only see in a very narrow range of light, which mostly disappears at night, and many things are similar colored.
Easily get relaxed, losing focus.
Easily distracted.
Are overconfident as you mentioned.

Consider this: All driving exams and tests are extremely thorough, require 100% to pass, and are required every year. If you are ever found at fault in an accident, your driving privilege is revoked. If you are ever caught breaking any law ( even just rolling through a stop sign, clipping a solid line, speeding etc. ), your privilege is revoked.

In just 5 years, how many people still have a license?

It isn't just over confidence. We are intrinsically bad at driving. This is why even quite early, simple self-driving technologies like the first iteration of Teslas Autopilot saw a 50% decrease in accident rates. Those technologies still aren't good at driving either..but they're already WAY better than we are. The real kicker though? We're getting WORSE at driving ( Despite all of the previously mentioned safety improvements to vehicles, including newer autodriving systems, fatality rates are on the rise. ), while auto systems are getting better at it.

I have never once, in my entire life, met a single person that can drive for any sort of reasonable distance without making a mistake. No matter how attentive and careful they are.


And yet, despite all that, we get from place to place every single day almost exclusively without running the big machine into stationary objects or other big machines.

Yes, we forget rules sometimes. On the other hand, we have an incredibly complex system of rules both written and unwritten that change between jurisdictions and sometimes between time of day and various situations, and we have the ability to adapt those rules as needed to make things work. If the functioning traffic light in front of us turns off suddenly, we know immediately what to do.

Yes, we also break rules sometimes. We have the judgment to decide whether we can do it safely. If I'm behind a slow-moving truck and can safely change lanes over a solid white line, I generally do it. So do we all.

Our vision and other physical traits hardly matter. We have set up the rules of the road, and the roads themselves, so that our vision can handle it. We are not trying to adapt to the jungle, we have made our own jungle, adapted to us.

"Better at driving" does not only mean "safer." If a self-driving car adds 15-20 minutes to my commute (30-40 per day) because it stubbornly follows every rule and puts up with nonsense that I don't, that's not better. If that's the case then I'm better than the self-driving car. I can make a split-second decision to change my route. I can pass a truck where a self-driving car might decide to just sit happily behind it. I'm familiar with the traffic light patterns. I know where the potholes are. And if my somewhat more aggressive driving (as compared to a self-driving car) doubles the very, very minuscule chance of getting in an accident, twice a very small number is still a very small number. I've been commuting for 14 years, so roughly 3,500 commuting days or 7,000 commutes, and I am 6,999-for-7,000 in accident avoidance. (One time I backed into a slow-moving car as I left my driveway, because I couldn't see around the parked cars on the street. A backup camera probably would've helped.) I'm betting a self-driving car, with its inflexible adherence to rules, inability to optimize the route, and lack of desire to optimize or prioritize traffic management to speed up the drive, would in fact add 15-20 minutes per commute and therefore it would not actually be better at driving.

Besides all that, it's hardly fair to say fatality rates are on the rise. In 2014 they bottomed out at 1.08 per 100 million miles traveled and have slightly risen to 1.18 in 2016. Over any kind of reasonable long-term view, they have been plummeting.

This thread actually illustrates how good at driving people really are. Given the minimal instruction and low standards for licensing, people still manage to get from place to place every single day, all the time, almost exclusively without hitting things. To say we are atrocious drivers implies we can barely get down the block. It ignores the insane amounts of miles traveled without incident.


Us having built the rules and "the jungle" around our abilities and shortcomings, yet still having any accidents, shows that we are not excellent at driving. We have tried very hard to make it so we don't kill each other all the time, and still manage to. You have about a 1% chance of dying to a car accident, despite the plethora of safety features, regulations, laws, restrictions and requirements.

Lets also not forget, that's JUST fatalities. Injuries is about 1% every single year.

These numbers are also just the US. Many other places are significantly worse, because they rely more on the humans ability to drive, rather than placing heavier restrictions to limit the damage.

Better evidence that people are getting worse, is the number of reported accidents overall, not just fatalities:
2011: 5,338,000
2012: 5,615,000
2013: 5,687,000
2014: 6,064,000
2015: 6,296,000

During this time, miles driven stayed pretty much even. So, the accident rate over just 5 years, jumped by nearly 18 percent.

Of further note, those are just the reported collisions, and are not per vehicle either. ( A collision between 2 vehicles is just 1 collision. 55 car pileup? Still 1 collision. ) It's very hard to get the exact numbers of these incidents. With lots of accidents involving multiple cars, and lots of accidents not reported, it's easy to get to even a 5% chance for an incident every single year.

These aren't insignificant numbers, and halving them is a big deal. Dropping them to 0.1% would be pretty gargantuan. Consider how much money a crash costs. Repairs, lawsuits, insurance, medical, disability, death.

You also forgot something really big about self driving cars. They could easily make it so you get to work faster, not slower. Yes, right now with very early and simple systems, with very few active, they often are a little bit slower. ( Not 20 minutes slower, unless you have a very long commute. Don't forget you can tell these automated systems to do things like change lanes, pass, etc. ) However, imagine how much faster you could go without traffic. Nearly all traffic is caused either by poor human driving habits, or by accidents. We could handle 10 times the number of cars if they were driven by computer systems. Traffic light patterns? Say goodbye to traffic lights. Learning where potholes are in your area? Try knowing where all potholes are everywhere. Traffic management? Try knowing where all traffic is at all times. We could also raise many road speeds to quite significant numbers. Most any road could be raised at least 5mph from where they are if we switched to automated drivers. Those limits aren't because of limits of the vehicles: It's the people that are the problem.

There's also nothing to say you can't take what strengths humans DO have, and add them to automated systems, which as I said we already have. Would the right lane be better for the next 6 miles? Turn on your right blinker, and the car moves itself over. Is that back road actually faster than the main road the GPS chooses automatically? Change the route then. Scared of heights, so you want to avoid that Cliff? Go for it. Speed up? Can do. All of these already exist on Teslas automated systems here and now. You're now combining humans decision making, with the advanced capabilities of the automated system. Let it handle the tedious, mundane, etc. You can focus on the big picture. Everyone wins.

So, yea. We are really bad at driving. Automated IS better than we are at driving, especially when you add in what few pros we have to it.

There is no reason at all to not be pushing automated advancements right now. The sooner we can get switched over to full auto, the better.

If you're hoping to switch over to full auto any time soon, you'll likely be very disappointed. If I see it in my lifetime I'll be surprised. If it never comes, I wouldn't be surprised. I believe what you're describing is a utopian experience requiring complete and total connectivity between cars, who all tell each other every piece of information they come across and automatically update their maps and information accordingly. How else would they all have the completely accurate and up-to-date pothole database you describe?

We are over ten years into the smartphone revolution and still only 2/3 to 3/4 of the US population has one. That's for a device that costs a few hundred dollars, is typically replaced every couple years, and isn't all that complicated to engineer. Cars cost $25,000 and up, are insanely difficult to engineer, and are often replaced no more than once a decade or more. When they are, they're very often replaced by used cars. If every automaker came out right now with complete and total self-driving capacity and perfect total connectivity and put it on every single model tomorrow, it'd be 20-30 years before they became ubiquitous enough for the benefits of the full-autonomy world you describe. But the technology to really do that is still years off. And I don't think they come with all those benefits, either. Two six-lane roads intersecting, with all the traffic that entails, cannot move that traffic through the intersection at speed, both ways, collision-free, without some kind of major slowdown or management system. And autonomous cars will always be programmed to err on the side of caution.

I would add that it's not quite true that vehicle-miles-traveled (VMT) has been flat in the time frame you describe. Going from 2.95 trillion to 3.095 trillion is a 5%-ish jump. 145 billion is no small thing. Combined with the accident stats you quote, our crash rate is one accident per 491,582 miles traveled. Yes, it's up from one every 552,641 miles traveled. Still, that kind of failure rate would be considered absurdly outstanding in any other area, except maybe aviation. The medical industry should be so damn good at what it does, as humans are at driving. If that's the definition of being bad at something, then what are humans good at?

User avatar
zemerick
Posts: 25
Joined: Thu May 12, 2011 6:01 am UTC

Re: 1990: "Driving Cars"

Postby zemerick » Tue May 08, 2018 8:35 pm UTC

sonar1313 wrote:If you're hoping to switch over to full auto any time soon, you'll likely be very disappointed. If I see it in my lifetime I'll be surprised. If it never comes, I wouldn't be surprised. I believe what you're describing is a utopian experience requiring complete and total connectivity between cars, who all tell each other every piece of information they come across and automatically update their maps and information accordingly. How else would they all have the completely accurate and up-to-date pothole database you describe?

We are over ten years into the smartphone revolution and still only 2/3 to 3/4 of the US population has one. That's for a device that costs a few hundred dollars, is typically replaced every couple years, and isn't all that complicated to engineer. Cars cost $25,000 and up, are insanely difficult to engineer, and are often replaced no more than once a decade or more. When they are, they're very often replaced by used cars. If every automaker came out right now with complete and total self-driving capacity and perfect total connectivity and put it on every single model tomorrow, it'd be 20-30 years before they became ubiquitous enough for the benefits of the full-autonomy world you describe. But the technology to really do that is still years off. And I don't think they come with all those benefits, either. Two six-lane roads intersecting, with all the traffic that entails, cannot move that traffic through the intersection at speed, both ways, collision-free, without some kind of major slowdown or management system. And autonomous cars will always be programmed to err on the side of caution.

I would add that it's not quite true that vehicle-miles-traveled (VMT) has been flat in the time frame you describe. Going from 2.95 trillion to 3.095 trillion is a 5%-ish jump. 145 billion is no small thing. Combined with the accident stats you quote, our crash rate is one accident per 491,582 miles traveled. Yes, it's up from one every 552,641 miles traveled. Still, that kind of failure rate would be considered absurdly outstanding in any other area, except maybe aviation. The medical industry should be so damn good at what it does, as humans are at driving. If that's the definition of being bad at something, then what are humans good at?


It indeed is a little ways away, but it doesn't have to be an all or nothing. As I mentioned, Tesla already has drastically improved the accident rate for their cars, and their system is incredibly simple. Nvidia recently unveiled a series of chips, cards, etc. that will drastically improve the capabilities of such systems. Some of those examples I gave would indeed require a whole networked system, but those were just extreme examples of what is capable, that humans can not even remotely compete with, ever.

Self-driving systems exist now, and more and more new cars are adding them. With an average age of cars under 12 years, this means we can expect a majority of cars to be automated in around 15 years, give or take. At around this point, the transitions can start. Things like only automated cars on highways. Automated fast lanes. Only automated vehicles in congestion zones.

None of that really matters though. My point was simply that we are NOT any good at all. We built a massive buffer system designed to try and allow us to make constant mistakes, but not have have an accident. I personally consider making more than 1 mistake per trip to be atrocious. To be adequate, it should be less than 1 a month or something. Excellent? 1 year, at most.

Further, my point on the self driving was more that the systems are actually very flexible, improving every day, and could easily meet your qualifications for "better driver".

Most of the time you see self-driving cars going slower than normal traffic isn't because that's how the system works. It's because they have the speed turned down while testing, to give the human inside more time to react when the car makes a mistake ( since it's being tested, IE isn't entirely ready for prime time. And before anyone tries to complain about testing on an open road...that's exactly how we teach and test humans too, because a simulation just can not compete with the real world. )

I'm also not saying we should force self driving on everyone here and now. I'm just saying that people should admit that we are not as capable as we like to think in driving cars, and that we should help not hinder efforts to develop automation.

User avatar
somitomi
Posts: 619
Joined: Fri Nov 06, 2015 11:21 pm UTC
Location: can be found in Hungary
Contact:

Re: 1990: "Driving Cars"

Postby somitomi » Tue May 08, 2018 8:58 pm UTC

zemerick wrote:Nearly all traffic is caused either by poor human driving habits, or by accidents. We could handle 10 times the number of cars if they were driven by computer systems.

This isn't entirely true, there is an intrinsic limit to how many cars can pass through any given cross-section of the road network. The simplest upper limit is that a car of length l traveling at v speed takes l/v time to pass through a line drawn across the road. So even with cars driving bumper-to-bumper, there's a limit to how many cars can cross this line in a given amount of time and in reality cars need to keep some distance from the car ahead of them. Theoretically self driving cars could drive faster while keeping shorter distances, but this limit will still be there.
—◯-◯

wumpus
Posts: 531
Joined: Thu Feb 21, 2008 12:16 am UTC

Re: 1990: "Driving Cars"

Postby wumpus » Tue May 08, 2018 9:20 pm UTC

sonar1313 wrote:Comparing cars across 30 years of development is going to get you some odd results. Your Prius almost certainly has more horsepower than what passed for a Dodge Charger in 1985. Those old '80s Civics gave you anywhere from 55 to around 80 hp. There are very few cars that are underpowered nowadays for highway use; the '80s were different.
[context:]
A possibility: Cars got neutered starting in the late 70s and early 80s as manufacturers scrambled to meet fuel economy requirements the only way they knew how: by killing their horsepower. By 1985 through 1991 most of the big horsepower cars were off the road, replaced by Honda Civics and Plymouth Reliants that wouldn't have known what 85 mph was if you beat them over the head with it.
[/context]


The 1980s cars were also lightweight deathtraps compared to modern cars (60hp and lightweight works a lot like a Prius. Until you get hit), although I'm pretty sure getting >60mpg requires fine tuning and hypermiling, again like the Prius.

I'm convinced it was safer new cars, and younger drivers who assumed that you need to wear a seatbelt to drive.

This is also *way* too late for the shear lack of power department. Detroit basically stopped building muscle in 1973-1974 [opec oil crisis] (just look at the corvette engines shipped in 1974 vs. 1975, 45hp lost and no big block option), not 1980. In 1977 the "high output model" for a Chevy Malibu Classic (305 cubic inch, 4 barrel carb) was about 160hp (and mated to a 3 speed automatic without lockup). 4 cylinder imports produced quite a bit more hp/l, but still not significant power. Also there is simply no way to expect any of these cars to last more than 10 years, which would mean "steady decline" should be between 1974 and 1984, not 1984 through 1991. I'm guessing anything but power loss. Remember, these were 1970s and 1980s cars we are talking about: US cars fell apart, Japanese cars simply rusted if you looked at them funny. They didn't last for decades the way modern cars do (areas blessed with zero humidity and salt might be an exception. But even well inland in Maryland they rusted to death).

Notes from an old driver (my parents' cars): [Note: the first two cars listed had shoulder belts in front and lapbelts [only] in the back. Cars before this had seatbelts [only] in the front and no seatbelts in the back (at least until my dad added one for me)].

1976 Datsun B210 "Honeybee" 1.4l 4cylinder engine, 4 speed manual gearbox. Acceleration involved downshifting, flooring it, and waiting for a downhill. Life expectancy is simple: the car would only die due to rust. At 4 years it is obviously rusting out and by 10 years it will likely come apart on its own. 65mph was living dangerously, but the car didn't appear to complain too badly.
1977 Chevy Malibu 5.0l V-8, 4-barel carb, 3 speed auto (lockup? What's that?). Once convinced to drop down to second, the car will take off (and drink the rest of the tank). Don't expect it to turn, and nobody attempted to take it to 80mph.
1984 Toyota Corolla 1.5l, 5 speed manual (basically an AE-86 with pushrods), quite a bit more pep, but the car felt like it was going to shake itself to death at 85mph (no idea actual top speed, for some reason I never realized the relationship between the tach, the transmission and the wheels).
1986 Acura Legend 2.5l V-6, automatic, 150hp [listed]. I didn't drive this car very often, but was driving my grandmother back from my parents' place when I looked down and saw an indicated 85mph (back in the 55mph days). Yes, you could hit 85 while driving sufficiently calmly for a grandmother. Sufficiently fast and well crafted 80s cars could speed like modern cars.

Generally speaking, these problems gradually went away in the 1990s. Power wildly increased along with rigidity (don't expect to read a car review from 1990-2010 that doesn't describe the car's "rigidity"). My guess is that cars that could go on an adventure to 100mph became very hard to find after 1973 (and not long after were forbidden from even showing >85 on the speedometer) until roughly 1990, and then only slowly becoming common (I lived on the road and rented cars 1994-1995. Typically horsepower was ~150, probably double what it might have been in the 1980s (hint: get the V6 unless you want to learn all about lack of torque [and gears. Might have been tolerable with a 5 speed manual, not so much with a 3 speed auto]).

User avatar
zemerick
Posts: 25
Joined: Thu May 12, 2011 6:01 am UTC

Re: 1990: "Driving Cars"

Postby zemerick » Tue May 08, 2018 9:44 pm UTC

somitomi wrote:
zemerick wrote:Nearly all traffic is caused either by poor human driving habits, or by accidents. We could handle 10 times the number of cars if they were driven by computer systems.

This isn't entirely true, there is an intrinsic limit to how many cars can pass through any given cross-section of the road network. The simplest upper limit is that a car of length l traveling at v speed takes l/v time to pass through a line drawn across the road. So even with cars driving bumper-to-bumper, there's a limit to how many cars can cross this line in a given amount of time and in reality cars need to keep some distance from the car ahead of them. Theoretically self driving cars could drive faster while keeping shorter distances, but this limit will still be there.


Absolutely true. I meant that current traffic slowdowns are almost all caused by people. Pull them out, and you can greatly increase the flow rate.

Mikeski
Posts: 1018
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1990: "Driving Cars"

Postby Mikeski » Tue May 08, 2018 10:47 pm UTC

sonar1313 wrote:If every automaker came out right now with complete and total self-driving capacity and perfect total connectivity and put it on every single model tomorrow, it'd be 20-30 years before they became ubiquitous enough for the benefits of the full-autonomy world you describe.

You'll also need fully-autonomous motorcycles.

And autonomous bicycles.

And pedestrians.

And deer. And dogs. And geese. And anything else big enough to affect a vehicle's travel.

Or you'll need some Jet-Age-sci-fi-looking gerbil tubes over all your roadways. Which would, admittedly, look pretty rad. They'd be hard to add to downtown NYC, though.

Combined with the accident stats you quote, our crash rate is one accident per 491,582 miles traveled. Yes, it's up from one every 552,641 miles traveled. Still, that kind of failure rate would be considered absurdly outstanding in any other area, except maybe aviation. The medical industry should be so damn good at what it does, as humans are at driving. If that's the definition of being bad at something, then what are humans good at?

Current data on computer-driven cars is available, at least in California.

Picking on the Google/Waymo project, and also looking back to some previous reports, their "disengagement rate" (i.e. the human driver had to take over when the computer gave up) was:

2015: ~0.80/1000 mi (1 failure per 1244 miles)
2016: ~0.20/1000 mi (1 failure per 5127 miles)
2017: ~0.18/1000 mi (1 failure per 5596 miles)

The 90/90 rule of engineering is lumbering out of hibernation, as the computer-driven cars are reaching... 1% of human ability.

Granted, not all "disengagements" would have caused crashes, but if the idea is to "eliminate human drivers", you have to eliminate the need for human drivers to take over due to hardware/software failures.

If you think human drivers are bad now, imagine how they'll look when they only drive 1/5000th as much as they currently do, because their car drives itself the rest of the time. I suppose, like for airline pilots, we could have all drivers log time in a simulator to keep their skills up. And we could combine that with known downtime by putting the simulators in the driver's seats of autonomous cars. Genius!

I'm a technophile myself (I design microchips for a living, for Moore's sake), but I don't see autonomous cars taking over in my lifetime.
Last edited by Mikeski on Wed May 09, 2018 7:45 am UTC, edited 1 time in total.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 3483
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1990: "Driving Cars"

Postby Soupspoon » Tue May 08, 2018 10:51 pm UTC

sonar1313 wrote:Cars cost $25,000 and up, are insanely difficult to engineer, and are often replaced no more than once a decade or more.

Personal experience of a succession of family cars (replaced with new models, and in at least one case continued to serve a similar time 'second-hand' as a hand-me-down within the famly) I can work out in hindsight by recalling their registration numbers, with the year letter. Mostly 4, 5 or 6 years. Seemed longer, I was surprised when I worked it out just now.

And other people I know who go for new cars tend to do 'new model every other year', except for the flash one that wouldn't be seen with 'last year's' numberplate (but too stingy to invest in a more timeless and transferable 'personalised' vanity-plate).

The used-car buyers I know run them into the ground, from a baseline of often there being something (maybe minor) already wrong with it (perhaps just a niggle like a rattling glove compartment hatch, but <insert anecdote about the clutch that went on the car within days of being bought from a major reseller, tried to blame the new owner>), IME. This and the above estimation may arise from observation bias (who I know, what more stands out in my mind and what less), so really I should look at stats.

Something that seems to be well corroborated (or at least good enough to be repeated or inspiring the 'facts' to be re-used elsewhere, not necessarily the same thing I know!) is the to be found in the report https://www.racfoundation.org/wp-conten ... report.pdf which talks about this, as well as featuring (horribly rendered!) graphs. Tge most relevant ones may be 8,9, 12 and 15 if I've remembered that list correctly. Anyway, we're told that there's an average 16 years of life in a car, across four owners, the first of which (private owners) has slightly less than equal share in the total time-span, if I read it correctly. (Company cars spend much less time 'new' before probably entering the general used-car population.) The trend may be shortening a little, but no indication that the instantaneous 'life expectancy' was much more even a couple of decades ago.

So maybe we were typical, back then. For the UK market, at least. But still a replacement rate (per ownership) of less than a decade I think. Maybe back in the '60s or before there could have been longer possessions, but from '70s onwards (especially the sometimes booming '80s, as the all-but-the-last-stubborn-trickle cars of the 1997 scrappage figures were being manufactured) the rate seems to be fairly high.

Not sure about what I' m arguing here, though, I just thought the quote quoted (in the midst of a hilking big wall of text) didn't quite ring true and needed investigation. Obviously you'd get different results in Cuba to those in Monte Carlo or Bahrain, and I don't know where the US strictly lies upon that spectrum (I suspect quite a bit of both, in extremis, with a large fuzzy middle-ground dominating the graphics).


But, most recently, there has been the trend towards ultra-short 'permanent leasing' arrangements, on a 'buy in on ownership or move on to another car' end-of-contract basis. I think the trend might be to supercharge the very-nearly-new resale end of the car market, satisfying bleeding-edge customers with new vehicles and just-off-the-bleeding-edge buyers of barely-used cars getting a higher proportion of the (re)sales, perhaps mostly at the expense of the middle-ground inhabited by medium-age 'clean diesels' that have since been shown to have been erroneously/ill-advisably marketed and may be swallowed up by manufacturer buy-back/scrappage schemes (while post-catalytic/pre-big-diesel-lie era vehicles scrappage or enter 'classic' status in similar proportions as before.

But that's just a gut feeling, mostly.

User avatar
LaserGuy
Posts: 4540
Joined: Thu Jan 15, 2009 5:33 pm UTC

Re: 1990: "Driving Cars"

Postby LaserGuy » Wed May 09, 2018 12:00 am UTC

somitomi wrote:In my experience driving is one of those things practice makes perfect, so I fail to see how the time spent on practicing driving with a trained instructor before getting your license could be spent on more useful things.


The data seems to suggest that driver's education programs have a fairly negligible effect on collisions. Occasionally, they may even increase the number of collisions because some of these sorts of programs allow people to get their licenses at younger ages if they take the training, which pretty much offsets any potential benefits the training might give. The most straightforward mechanism to reduce collisions appears to be just increasing the age that people can get their licenses. Pushing the age up to 21 or 25, even if there's no training requirement, would probably do more to cut collisions than any mandatory training program.

tl;dr: Teenages are stupid.
Last edited by LaserGuy on Wed May 09, 2018 12:02 am UTC, edited 1 time in total.

User avatar
GlassHouses
Posts: 122
Joined: Thu Nov 24, 2016 12:41 pm UTC

Re: 1990: "Driving Cars"

Postby GlassHouses » Wed May 09, 2018 12:02 am UTC

Soupspoon wrote:
Not sure about what I' m arguing here, though, I just thought the quote quoted (in the midst of a hilking big wall of text) didn't quite ring true and needed investigation. Obviously you'd get different results in Cuba to those in Monte Carlo or Bahrain, and I don't know where the US strictly lies upon that spectrum (I suspect quite a bit of both, in extremis, with a large fuzzy middle-ground dominating the graphics).

I've seen the number 12 bandied about a few times recently, but as in "the average age of cars in the U.S. is 12 years." I find that hard to believe, I'd say an average lifespan of 12 years seems more plausible. Then again, I do live in one of the more prosperous corners of the country, and maybe elsewhere the average car really does last 24 years... :shock:

JPatten
Posts: 88
Joined: Tue Jul 13, 2010 5:39 pm UTC
Location: Southeast USA
Contact:

Re: 1990: "Driving Cars"

Postby JPatten » Wed May 09, 2018 1:11 am UTC

I would say average Lifespan of a car, which would be middle of the road. I think it's Clark Howard who advocates that it takes 10 years to really get the worth out of a car. Usually, after 10 years repair costs start to add up and then instead of going down the cost of ownership starts rising again.
As in if you take the cost of the car (purchase, maintenance, gas etc...) then up to ten years the cost per mile generally is dropping as more and more miles accumulate. But after 10 years or so the cost can start going up as sometime after 10 years is when major systems have higher incidences of failure. of course that number MAY be higher now with increased reliability of cars. I would like to get at least another 2 years out of our van which would put it at 12 years old.
It is not uncommon for cars to last longer and some models MUCH longer.

gormster
Posts: 230
Joined: Mon Jul 23, 2007 6:43 am UTC
Location: Sydney

Re: 1990: "Driving Cars"

Postby gormster » Wed May 09, 2018 4:29 am UTC

Mikeski wrote:No idea about horses, but I do recall reading that, in the injuries-per-mile-traveled sense, you would be better off letting bicyclists go bare-headed, and having pedestrians wear those foam half-walnut helmets.


Putting people in danger makes them drive more safely, and vice versa. There would be fewer accidents if we didn't make people wear seatbelts, but their injuries would be much more severe. Same goes for bikes: there are more cycling accidents with helmet laws, but much, much fewer deaths and debilitating brain injuries.
Eddie Izzard wrote:And poetry! Poetry is a lot like music, only less notes and more words.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 3483
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1990: "Driving Cars"

Postby Soupspoon » Wed May 09, 2018 8:07 am UTC

JPatten wrote:I would say average Lifespan of a car, which would be middle of the road.

You only need two cars of that kind, going in opposite directions, and the lifespan of both will be drastically reduced. ;)

pscottdv
Posts: 61
Joined: Fri Feb 19, 2010 4:32 pm UTC

Re: 1990: "Driving Cars"

Postby pscottdv » Wed May 09, 2018 11:48 am UTC

[quote="JPatten"]When I got my License there was no "required" driving time or drivers ed required. All I had to do was pass the written test and the driving test. Which in my case consisted of backing out of the parking lot, driving down to the town square, around the courthouse (sort of a roundabout type deal) and then back up to the courthouse and park. Of course, that was basically 30 years ago now.[/quote

For a short period of time in Michigan, they dropped the requirement for the road test which means that the day I turned 16, in the middle of winter with ice on the roads, I went down to the secretary of state's office, passed an easy, 40-question, multiple-choice test and was handed a license to drive.

Boy was my sister mad about that: she had failed her road test twice before she got her license.

pscottdv
Posts: 61
Joined: Fri Feb 19, 2010 4:32 pm UTC

Re: 1990: "Driving Cars"

Postby pscottdv » Wed May 09, 2018 12:05 pm UTC

JPatten wrote:I would say average Lifespan of a car, which would be middle of the road. I think it's Clark Howard who advocates that it takes 10 years to really get the worth out of a car. Usually, after 10 years repair costs start to add up and then instead of going down the cost of ownership starts rising again.
As in if you take the cost of the car (purchase, maintenance, gas etc...) then up to ten years the cost per mile generally is dropping as more and more miles accumulate. But after 10 years or so the cost can start going up as sometime after 10 years is when major systems have higher incidences of failure. of course that number MAY be higher now with increased reliability of cars. I would like to get at least another 2 years out of our van which would put it at 12 years old.
It is not uncommon for cars to last longer and some models MUCH longer.


I think your numbers are ridiculously low. I've never owned a vehicle less than 10 years old and can assure you they are generally still very reliable at that age.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 3483
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: 1990: "Driving Cars"

Postby Soupspoon » Wed May 09, 2018 12:17 pm UTC

The chances of a vehicle that is 10 years old (and selected for it looking to be in good order) surviving another ten years isn't necessarily related to the chances of a new vehicle eventually reaching 20 years.

In your case, I suspect you're avoiding the clunkers on the market. Either with a superior eye or just a preference for the sturdily built models that do better than average.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Wed May 09, 2018 12:58 pm UTC

zemerick wrote:
sonar1313 wrote:If you're hoping to switch over to full auto any time soon, you'll likely be very disappointed. If I see it in my lifetime I'll be surprised. If it never comes, I wouldn't be surprised. I believe what you're describing is a utopian experience requiring complete and total connectivity between cars, who all tell each other every piece of information they come across and automatically update their maps and information accordingly. How else would they all have the completely accurate and up-to-date pothole database you describe?

We are over ten years into the smartphone revolution and still only 2/3 to 3/4 of the US population has one. That's for a device that costs a few hundred dollars, is typically replaced every couple years, and isn't all that complicated to engineer. Cars cost $25,000 and up, are insanely difficult to engineer, and are often replaced no more than once a decade or more. When they are, they're very often replaced by used cars. If every automaker came out right now with complete and total self-driving capacity and perfect total connectivity and put it on every single model tomorrow, it'd be 20-30 years before they became ubiquitous enough for the benefits of the full-autonomy world you describe. But the technology to really do that is still years off. And I don't think they come with all those benefits, either. Two six-lane roads intersecting, with all the traffic that entails, cannot move that traffic through the intersection at speed, both ways, collision-free, without some kind of major slowdown or management system. And autonomous cars will always be programmed to err on the side of caution.

I would add that it's not quite true that vehicle-miles-traveled (VMT) has been flat in the time frame you describe. Going from 2.95 trillion to 3.095 trillion is a 5%-ish jump. 145 billion is no small thing. Combined with the accident stats you quote, our crash rate is one accident per 491,582 miles traveled. Yes, it's up from one every 552,641 miles traveled. Still, that kind of failure rate would be considered absurdly outstanding in any other area, except maybe aviation. The medical industry should be so damn good at what it does, as humans are at driving. If that's the definition of being bad at something, then what are humans good at?


It indeed is a little ways away, but it doesn't have to be an all or nothing. As I mentioned, Tesla already has drastically improved the accident rate for their cars, and their system is incredibly simple. Nvidia recently unveiled a series of chips, cards, etc. that will drastically improve the capabilities of such systems. Some of those examples I gave would indeed require a whole networked system, but those were just extreme examples of what is capable, that humans can not even remotely compete with, ever.

Self-driving systems exist now, and more and more new cars are adding them. With an average age of cars under 12 years, this means we can expect a majority of cars to be automated in around 15 years, give or take. At around this point, the transitions can start. Things like only automated cars on highways. Automated fast lanes. Only automated vehicles in congestion zones.

None of that really matters though. My point was simply that we are NOT any good at all. We built a massive buffer system designed to try and allow us to make constant mistakes, but not have have an accident. I personally consider making more than 1 mistake per trip to be atrocious. To be adequate, it should be less than 1 a month or something. Excellent? 1 year, at most.

Further, my point on the self driving was more that the systems are actually very flexible, improving every day, and could easily meet your qualifications for "better driver".

Most of the time you see self-driving cars going slower than normal traffic isn't because that's how the system works. It's because they have the speed turned down while testing, to give the human inside more time to react when the car makes a mistake ( since it's being tested, IE isn't entirely ready for prime time. And before anyone tries to complain about testing on an open road...that's exactly how we teach and test humans too, because a simulation just can not compete with the real world. )

I'm also not saying we should force self driving on everyone here and now. I'm just saying that people should admit that we are not as capable as we like to think in driving cars, and that we should help not hinder efforts to develop automation.


I'd be curious to know what you call a mistake. A crash? Something that would've been a crash if not for the reaction of the other driver? Breaking the speed limit by 1 mph? Turning down the wrong street?

I also think that one mistake per trip is far too stringent a criterion. The best baseball/football/soccer/etc. players make multiple mistakes a game. Serena Williams makes dozens of unforced errors every match. The question is, did we get there safely or didn't we? Something like 99.999% of the time, if not more, the answer is yes. But if you stick with the one-mistake-per-trip criterion, and that's atrocious, then right now, there are no words for how horrifyingly, appallingly bad self-driving cars really are. After all, just as you say, they're being driven slowly to account for their inevitable screwup, no?

So I think there's either a problem with that criterion, or the definition of a mistake. Also, I think that's not the reason they're being driven slowly. I doubt they will ever be programmed to break any rules, including and especially speed limits. They won't make a left immediately after the light turns red if the oncoming traffic used up all the green light time. Things like that. We get away with that because the police would never do anything else if they zealously enforced the precise speed limit and they'd never stop it from happening, and society is better off for it anyway. But if one big, wealthy company purposely programs its auto-driving vehicles to break laws? There would be problems. That's a much easier target than the millions of us - without hundreds of billions in revenue - who do the same.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Wed May 09, 2018 1:08 pm UTC

GlassHouses wrote:
Soupspoon wrote:
Not sure about what I' m arguing here, though, I just thought the quote quoted (in the midst of a hilking big wall of text) didn't quite ring true and needed investigation. Obviously you'd get different results in Cuba to those in Monte Carlo or Bahrain, and I don't know where the US strictly lies upon that spectrum (I suspect quite a bit of both, in extremis, with a large fuzzy middle-ground dominating the graphics).

I've seen the number 12 bandied about a few times recently, but as in "the average age of cars in the U.S. is 12 years." I find that hard to believe, I'd say an average lifespan of 12 years seems more plausible. Then again, I do live in one of the more prosperous corners of the country, and maybe elsewhere the average car really does last 24 years... :shock:


I drive through some pretty awful areas of the city on my commute. 24 years is a stretch for either definition of average, but the inner cities are absolutely loaded with evidence that the average lifespan of cars is probably longer than 12 years.

I think for purposes of this discussion, it's average age, not lifespan, that matters. Average lifespan is brought down by driving your shiny new car off the lot and smacking a tree a mile down the road. That doesn't really change the penetration into the market of a new technology. What matters is whether people adopt it, and they don't do so when they go looking for a $2000 set of wheels because that's the most they can afford. The question is not whether a Chevy Cobalt (or pick your model) lasts 6, 12, or 20 years, it's whether there's a Chevy Cobalt available to buy when someone wants a cheap used car. There always is.

User avatar
Sableagle
Ormurinn's Alt
Posts: 1807
Joined: Sat Jun 13, 2015 4:26 pm UTC
Location: The wrong side of the mirror
Contact:

Re: 1990: "Driving Cars"

Postby Sableagle » Wed May 09, 2018 7:25 pm UTC

sonar1313 wrote:If you're hoping to switch over to full auto any time soon, you'll likely be very disappointed. If I see it in my lifetime I'll be surprised. If it never comes, I wouldn't be surprised. I believe what you're describing is a utopian experience requiring complete and total connectivity between cars, who all tell each other every piece of information they come across and automatically update their maps and information accordingly. How else would they all have the completely accurate and up-to-date pothole database you describe?

We are over ten years into the smartphone revolution and still only 2/3 to 3/4 of the US population has one. That's for a device that costs a few hundred dollars, is typically replaced every couple years, and isn't all that complicated to engineer.
It'd be easier to achieve with fewer vehicles. Automate trams and buses and have them stay in touch and make and adjust their plans so that people don't miss a train by 30 seconds just because someone else stalled at a red light or whatever.

You know what else, humans do really, really well? Gun safety. Roughly 10 billion rounds are manufactured in the US each year, according to Wired, but still you only have ...
2016: 38,658
2015: 36,247
2014: 33,599
2013: 33,636
2012: 33,563
2011: 32,351
2010: 31,672
2009: 31,347
2008: 31,593
2007: 31,224
2006: 30,896

... total deaths, of which only ...
2016: 489
2015: 485
2014: 586
2013: 505
2012: 548
2011: 591
2010: 606
2009: 554
2008: 592
2007: 613
2006: 642

... were accidental, and a piffling ...
2013: 84,258
2012: 81,396
2011: 73,883
2010: 73,505
2009: 66,769
2008: 78,622
2007: 69,863
2006: 71,417

... non-fatal firearm injuries, according to http://www.gunpolicy.org/firearms/region/united-states
... so why restrict them at all? Obviously we're capable of handling them safely.
Oh, Willie McBride, it was all done in vain.

User avatar
gmalivuk
GNU Terry Pratchett
Posts: 26440
Joined: Wed Feb 28, 2007 6:02 pm UTC
Location: Here and There
Contact:

Re: 1990: "Driving Cars"

Postby gmalivuk » Wed May 09, 2018 8:17 pm UTC

zemerick wrote:Nearly all traffic is caused either by poor human driving habits, or by accidents. We could handle 10 times the number of cars if they were driven by computer systems.
Huge [citation needed] on that figure.

I accept that the number of cars could increase a bit if all of them were connected and computer driven, but not by anything like a factor of 10.

And the increase in capacity would probably be more than countered by the increase in cars being used when we don't have to drive them ourselves. Ridesharing already has increased traffic problems in many cities, and this would add to that problem.
Unless stated otherwise, I do not care whether a statement, by itself, constitutes a persuasive political argument. I care whether it's true.
---
If this post has math that doesn't work for you, use TeX the World for Firefox or Chrome

(he/him/his)

Mikeski
Posts: 1018
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1990: "Driving Cars"

Postby Mikeski » Thu May 10, 2018 2:48 am UTC

pscottdv wrote:I think your numbers are ridiculously low. I've never owned a vehicle less than 10 years old and can assure you they are generally still very reliable at that age.

I think age (of the driver, not the car) has a lot to do with it.

If you grew up with crappy cars from the 70s and 80s, you might have a mental model of a 10-year-old car as "ancient". I still think of 100,000 miles on the odometer as "holy crap, take it to the scrapyard," even though I know current cars are good for far longer than that.

I sold a 1984 Ford in 1996, for not much more than its value as scrap metal. A bit over 80,000 miles, and it was done. Well-maintained, but it lived outdoors (no garage) in Minnesota. On the other hand, I traded in a 2000 Pontiac in 2009, and figured it was less than halfway through its life.

If you're younger or older than me, you may have a different mental model for "run-down jalopy".

Also, the US market was skewed by the "Cash for Clunkers" program, which removed 677,000 used cars from the market in the summer of 2009, in an effort to reduce pollution and boost the consumer economy thru new-car sales (spoiler alert: didn't work). So USAans and non-USAans will have differing views based on that, as well.

sonar1313
Posts: 138
Joined: Tue Mar 05, 2013 5:29 am UTC

Re: 1990: "Driving Cars"

Postby sonar1313 » Thu May 10, 2018 2:26 pm UTC

Sableagle wrote:
sonar1313 wrote:If you're hoping to switch over to full auto any time soon, you'll likely be very disappointed. If I see it in my lifetime I'll be surprised. If it never comes, I wouldn't be surprised. I believe what you're describing is a utopian experience requiring complete and total connectivity between cars, who all tell each other every piece of information they come across and automatically update their maps and information accordingly. How else would they all have the completely accurate and up-to-date pothole database you describe?

We are over ten years into the smartphone revolution and still only 2/3 to 3/4 of the US population has one. That's for a device that costs a few hundred dollars, is typically replaced every couple years, and isn't all that complicated to engineer.
It'd be easier to achieve with fewer vehicles. Automate trams and buses and have them stay in touch and make and adjust their plans so that people don't miss a train by 30 seconds just because someone else stalled at a red light or whatever.

You know what else, humans do really, really well? Gun safety. Roughly 10 billion rounds are manufactured in the US each year, according to Wired, but still you only have ...
2016: 38,658
2015: 36,247
2014: 33,599
2013: 33,636
2012: 33,563
2011: 32,351
2010: 31,672
2009: 31,347
2008: 31,593
2007: 31,224
2006: 30,896

... total deaths, of which only ...
2016: 489
2015: 485
2014: 586
2013: 505
2012: 548
2011: 591
2010: 606
2009: 554
2008: 592
2007: 613
2006: 642

... were accidental, and a piffling ...
2013: 84,258
2012: 81,396
2011: 73,883
2010: 73,505
2009: 66,769
2008: 78,622
2007: 69,863
2006: 71,417

... non-fatal firearm injuries, according to http://www.gunpolicy.org/firearms/region/united-states
... so why restrict them at all? Obviously we're capable of handling them safely.

Other than the part about not being restricted at all (they already are to some extent, as are cars of course) I agree completely. Gun safety is another thing humans are remarkably good at, the breathless coverage of every stray bullet notwithstanding. There are tens of millions, possibly over a hundred million, gun owners in the US (not counting those who have them illegally), and as you can see almost all of them manage to handle them without putting a bullet in someone else.

wumpus
Posts: 531
Joined: Thu Feb 21, 2008 12:16 am UTC

Re: 1990: "Driving Cars"

Postby wumpus » Thu May 10, 2018 5:10 pm UTC

Mikeski wrote:2015: ~0.80/1000 mi (1 failure per 1244 miles)
2016: ~0.20/1000 mi (1 failure per 5127 miles)
2017: ~0.18/1000 mi (1 failure per 5596 miles)

The 90/90 rule of engineering is lumbering out of hibernation, as the computer-driven cars are reaching... 1% of human ability.


https://www.dmv.ca.gov/portal/wcm/conne ... OD=AJPERES

Presumably that is where the data comes from, and what little data they show makes nearly all disengagements appear to be driving at least as safely as a human.

The other assumption is that "all humans" drive at "human ability".

I can assure you that in my first 5,000 miles of driving, a Waymo computer would take the wheel out of my hands a number of times. I can't imagine that the typical teenager can manage 5,000 miles without more than a few "disengagement events". Certainly any use of a cell phone would qualify as a "sensory failure" (something that barely existed when I was a teen).

My 82 year old father is even worse. I *know* that he can't drive 5,000 miles without driving through a red light. I'm not even sure he goes 500 miles between such acts (in the few times I've ridden in the car with him I've seen him do it. And he's crashed at least once this way (and will go into extreme conspiracy theory insisting that his enemies set him up).

It is politically impossible to assemble data on how many "disengagement events" a drunk driver will do, but I can only hope that people who like to keep living (and keep family members alive as well) + the alcohol industry can prevent MADD and their allies from keeping operating a self-guided car while drunk DWI/DUI.

I'd recommend comparing crash data to crash data not "disengagement events" to crash data. I'd also recommend that there are clearly people who shouldn't be allowed to touch the steering wheel. The 90/90 rule applies for humans as well, assuming that 10% of the population causes 90% of the crashes. Currently, it is next to impossible to revoke driving rights from people for anything less than multiple DUIs (and even then they still drive) as it leaves anyone not living in NYC, DC, Chicago, and the SF/Bay area essentially unemployed and shut in his house. Obviously, expecting someone to buy a Tesla (or similarly cost self-driving car) is still not an option for all but the richest of dangerous drivers, drunks, and the elderly, but hopefully that can change.

Convincing the other 90% to let the computer drive is a completely different issue. I'd be more than happy to turn over driving to Tesla's computer for a drive taking more than 4 hours (and grab the wheel when it asks) but don't see much use for it. A much more advanced system where I can simply surf the web while commuting would be another story, but I'm sure that I'd always want manual control (maybe I should start learning to ride a motorcycle).

Mikeski
Posts: 1018
Joined: Sun Jan 13, 2008 7:24 am UTC
Location: Minnesota, USA

Re: 1990: "Driving Cars"

Postby Mikeski » Thu May 10, 2018 6:01 pm UTC

wumpus wrote:I'd recommend comparing crash data to crash data not "disengagement events" to crash data.

That requires data on fully-autonomous cars, as a "disengagement" with a human to take over usually does not result in a crash... but it's also not a crash that the computer avoided, it was a crash that a human avoided.

wumpus wrote:The 90/90 rule applies for humans as well, assuming that 10% of the population causes 90% of the crashes.

That's the normal 90/10 rule. (Sometimes 80/20).

The 90/90 rule is a different rule used among engineering types: "the first 90% of a project takes 90% of the schedule. The remaining 10% of a project takes the other 90% of the schedule."

That is, management assumes that after you've solved at the easy problems at a certain rate, you can solve the remaining difficult problems at the same rate. And the more problems you want solved, the closer the schedule gets to infinity.

There was a 400% improvement in Google Car's driving ability from 2015 to 2016.

Then there was a ... 10% improvement from 2016 to 2017.

Apparently, they're done with the easy 90% and moving into the hard 90%.

YTPrenewed
Posts: 67
Joined: Tue Sep 07, 2010 12:09 am UTC

Re: 1990: "Driving Cars"

Postby YTPrenewed » Thu May 10, 2018 8:58 pm UTC

Unclevertitle wrote:This. This is why I didn't actually get a driver's license until I was 24 even when the legal driving age in my state is 16.

I wasn't quite 24, but I didn't bother with the beginner's permit until I was 19, or the full license until I was 22, because I figured I could just take public transit, and figured any place not urban enough for public transit would be boring anyway.

A transit workers' strike reminded me that it's better to have my license and not need it than to need it and not have it. (I still felt defeated in a sense to realize I might need it at all, but I've since learned to somewhat enjoy driving from time to time.)

I nearly failed the parallel parking component of my exam. That's fine for me, because my hometown had no parallel parking spaces I ever needed to use, but that license IS good for the whole province, and I'm not sure it should be. (Didn't want to say anything, of course.)

I suspect the reason they don't make driving tests harder is because driving, rightly or wrongly, is perceived as a necessity, at least in some ways, for some people. It's part of the same reason carbon taxes are opposed so vehemently. If they're not going to tax carbon to reduce flooding deaths, why would they make driving tests harder to reduce car crash deaths?

wumpus
Posts: 531
Joined: Thu Feb 21, 2008 12:16 am UTC

Re: 1990: "Driving Cars"

Postby wumpus » Fri May 11, 2018 5:12 pm UTC

gmalivuk wrote:
zemerick wrote:Nearly all traffic is caused either by poor human driving habits, or by accidents. We could handle 10 times the number of cars if they were driven by computer systems.
Huge [citation needed] on that figure.

I accept that the number of cars could increase a bit if all of them were connected and computer driven, but not by anything like a factor of 10.

And the increase in capacity would probably be more than countered by the increase in cars being used when we don't have to drive them ourselves. Ridesharing already has increased traffic problems in many cities, and this would add to that problem.


For highway traffic, this should be trivial to compute by determining the spacing of cars possible by a human, and the spacing of cars possible by a human.

For US highway traffic, assuming 70mph traffic (sorry about imperial units, but metric units are only used to repair cars in the USA):
4 seconds (what I needed to write on my driving test in the 1980s): ~400 ft
2 seconds (what is currently recommended) ~200 ft
1 second (typical in Maryland driving) ~100ft
0.1 second (possible for a computer?) -10ft?

A factor of 10 assumes that either the drivers aren't tailgating (which likely only happens in places that *don't* need more roads) or the computer simply has *zero* spacing between cars. A factor of 3 to 5 seems possible, although how you would be certain that a human didn't slip into your convoy is beyond me. Don't forget the aerodynamic advantages to all that drafting as well.

Mikeski wrote:
wumpus wrote:I'd recommend comparing crash data to crash data not "disengagement events" to crash data.

That requires data on fully-autonomous cars, as a "disengagement" with a human to take over usually does not result in a crash... but it's also not a crash that the computer avoided, it was a crash that a human avoided.

wumpus wrote:The 90/90 rule applies for humans as well, assuming that 10% of the population causes 90% of the crashes.

That's the normal 90/10 rule. (Sometimes 80/20).

The 90/90 rule is a different rule used among engineering types: "the first 90% of a project takes 90% of the schedule. The remaining 10% of a project takes the other 90% of the schedule."

That is, management assumes that after you've solved at the easy problems at a certain rate, you can solve the remaining difficult problems at the same rate. And the more problems you want solved, the closer the schedule gets to infinity.

There was a 400% improvement in Google Car's driving ability from 2015 to 2016.

Then there was a ... 10% improvement from 2016 to 2017.

Apparently, they're done with the easy 90% and moving into the hard 90%.


Being able to certify that a driverless car can drive with less crashes and less death than the cumulative rate for human-driven cars should be a huge turning point for such cars, and it is likely they are there (if not, those 10% gains might take awhile). Even so, there are issues between:

Replacing drivers with obviously limited driving skills (teenagers, elderly, drunks) <- we should be here
Replacing taxi drivers (Uber's obvious goal) , <- this isn't clear at all
Acting as electronic soccer moms (strong political push, from the same faction expected to push child safety) <- whether there is a difference between this and e-Uber is a political question
replacing drivers in general (expect a huge pushback, at least until people who didn't get a driver's license before their 17th birthday have a voting majority). On the other hand, limiting highways during rush hour to automated controls might be required (see above) in places that could really use the congestion relief. <- The computers are *almost there*, but retrofitting current cars would be a nightmare. Expect this to be decades away (although retrofitting the cars is probably easier than building new highways).


Return to “Individual XKCD Comic Threads”

Who is online

Users browsing this forum: No registered users and 40 guests