8,000 cars per hour per lane on the highways of tomorrow?

Started by kernals12, September 26, 2024, 04:40:36 PM

Previous topic - Next topic

Max Rockatansky

Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?

Hey now, stop taking these 150 AI car threads seriously...


Quillz

Quote from: Max Rockatansky on December 21, 2024, 06:29:20 PM
Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?

Hey now, stop taking these 150 AI car threads seriously...
I mean, I don't know if this proposal was serious or not, but it does have merit. I have no doubt there will be further automation/AI in regards to driving in the future. But I approach things from the idea that reality is always much slower and more boring than what we hope. I mean, we can't even get RealID up and running, it keeps getting delayed. That's basically a piece of plastic. If we can't do that, we are gonna have super highways handling 8k+ cars an hour, fully AI controlled, with no issues?

It's the same thing with politics. You hear all these scary things that are going to happen, then in reality these things rarely happen, or they do but once removed from the sensationalism, find out aren't a big deal. Hell, look at the recent Apple Intelligence rollout. Turns out it was pretty lackluster. It just lets you do a few new things you couldn't do before. Not exactly the insane AI revolution we were told.

Roadgeekteen

I wouldn't mind ai helping TSA but there has to be some form of human checking.
My username has been outdated since August 2023 but I'm too lazy to change it

Max Rockatansky

Quote from: Roadgeekteen on December 21, 2024, 06:32:56 PMI wouldn't mind ai helping TSA but there has to be some form of human checking.

Whether it is robots or humans doesn't matter much.  It is all security and safety theater.

kernals12

Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?
AI, doesn't need to be perfect, it just needs to be better than humans.

Roadgeekteen

Quote from: kernals12 on December 21, 2024, 06:35:43 PM
Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?
AI, doesn't need to be perfect, it just needs to be better than humans.
I'm totally fine with AI for cars as human drivers aren't that good. Flying is already so safe so why mess up a good thing? Sure TSA is annoying but I don't really mind taking my shoes off briefly.
My username has been outdated since August 2023 but I'm too lazy to change it

Max Rockatansky

Quote from: Quillz on December 21, 2024, 06:32:07 PM
Quote from: Max Rockatansky on December 21, 2024, 06:29:20 PM
Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?

Hey now, stop taking these 150 AI car threads seriously...
I mean, I don't know if this proposal was serious or not, but it does have merit. I have no doubt there will be further automation/AI in regards to driving in the future. But I approach things from the idea that reality is always much slower and more boring than what we hope. I mean, we can't even get RealID up and running, it keeps getting delayed. That's basically a piece of plastic. If we can't do that, we are gonna have super highways handling 8k+ cars an hour, fully AI controlled, with no issues?

It's the same thing with politics. You hear all these scary things that are going to happen, then in reality these things rarely happen, or they do but once removed from the sensationalism, find out aren't a big deal. Hell, look at the recent Apple Intelligence rollout. Turns out it was pretty lackluster. It just lets you do a few new things you couldn't do before. Not exactly the insane AI revolution we were told.

I don't entirely dismiss the notion that some of this stuff is one day possible.  Like you said, the progression is often so slow that it isn't even worth putting much thought into. 

Trouble is we had a huge recent rash of similar K12 threads (which I don't think you active for) that spun up the forum populace. 

Quillz

Quote from: kernals12 on December 21, 2024, 06:35:43 PM
Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?
AI, doesn't need to be perfect, it just needs to be better than humans.
What does "better" mean in this context, though? My car already has adaptive cruise control, and can do emergency braking. It can stop the car faster than I can, is that better? As I said in another example, what if I deliberately want to take a less direct, more scenic route? Will AI allow me? If not, why not? Sometimes "better" doesn't mean the most efficient or most direct, it can be mean maximizing what I want to see. That's the kind of thing where you need a manual human brain to decide. Not a programmer who is deciding for me.

Also, we can and have had many Teslas involved in self-driving crashes. "Not perfect" translates to a lot of the same issues that occur with humans behind the wheel. What about software bugs that have affected medical equipment and led to deaths? AI has to be programmed by humans, and history has shown it's not possible to have bug-free software and AI.

Roadgeekteen

Quote from: Quillz on December 21, 2024, 06:39:40 PM
Quote from: kernals12 on December 21, 2024, 06:35:43 PM
Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?
AI, doesn't need to be perfect, it just needs to be better than humans.
What does "better" mean in this context, though? My car already has adaptive cruise control, and can do emergency braking. It can stop the car faster than I can, is that better? As I said in another example, what if I deliberately want to take a less direct, more scenic route? Will AI allow me? If not, why not? Sometimes "better" doesn't mean the most efficient or most direct, it can be mean maximizing what I want to see. That's the kind of thing where you need a manual human brain to decide. Not a programmer who is deciding for me.

Also, we can and have had many Teslas involved in self-driving crashes. "Not perfect" translates to a lot of the same issues that occur with humans behind the wheel. What about software bugs that have affected medical equipment and led to deaths? AI has to be programmed by humans, and history has shown it's not possible to have bug-free software and AI.
AI self-driving cars will very likely let you choose what route just like Google Maps lets you choose your routes. Or if they somehow don't there are obviously work arounds. Like if you want to take CA 1 between LA and SF just set your final destination to the Big Sur, and when you are almost there change your final destination to SF. But again I see no reason why they wouldn't let you change your route. Self driving taxis/Ubers are different.
My username has been outdated since August 2023 but I'm too lazy to change it

Quillz

Besides, I think most AI is best exemplified by things like CarPlay/Android Auto. Where I can use buttons on my steering wheel to tell the car what I want (play this song, etc.) Or my mapping app of choice can update me with live traffic (car crash ahead, turn here to save time). This is the perfect use case for AI/automation, because it's simple and yet has a clear, practical impact on my life. My driving experience is actively better with these services because they recognize things I am likely to do (play a song, get directions), and streamline them. I get to work faster with these services.

Problem is, a fully self-driving car that is deciding where I am going and what routes I am taking is not as useful in that case. Waze, for example, is the extreme example. It will redirect me... Onto side roads that have stop signs. Yes, I'm technically saving two minutes, but I'm really not when I have to sit at a major street and wait for traffic to clear. I've had several times I'm at a stop sign so long it just redirects me. Because these services still do not seem to understand that theoretical time savings is not the same thing as practical time savings. That's where you need a human brain and manual override. I'm going to understand the traffic flow of a local street better than Waze.

Max Rockatansky

Ehhh...that adaptive cruise control in my 2024 Corolla is the bane of my existence on access control roads.  It is way too sensitive and panics over a lot of non-hazardous scenarios.  I found how to turn it off but it requires deactivating the collision system every time I drive the car (which I only do on long freeway trips). 

Roadgeekteen

Quote from: Quillz on December 21, 2024, 06:43:52 PMBesides, I think most AI is best exemplified by things like CarPlay/Android Auto. Where I can use buttons on my steering wheel to tell the car what I want (play this song, etc.) Or my mapping app of choice can update me with live traffic (car crash ahead, turn here to save time). This is the perfect use case for AI/automation, because it's simple and yet has a clear, practical impact on my life. My driving experience is actively better with these services because they recognize things I am likely to do (play a song, get directions), and streamline them. I get to work faster with these services.

Problem is, a fully self-driving car that is deciding where I am going and what routes I am taking is not as useful in that case. Waze, for example, is the extreme example. It will redirect me... Onto side roads that have stop signs. Yes, I'm technically saving two minutes, but I'm really not when I have to sit at a major street and wait for traffic to clear. I've had several times I'm at a stop sign so long it just redirects me. Because these services still do not seem to understand that theoretical time savings is not the same thing as practical time savings. That's where you need a human brain and manual override. I'm going to understand the traffic flow of a local street better than Waze.
A fully self-driving car has uses that you might not think of. With them, you could do a red-eyed drive like a flight being able to go overnight without having to worry about staying awake. Also, they would be very big for safety simply because drunk people coming home from the bar could just get driving home.
My username has been outdated since August 2023 but I'm too lazy to change it

Quillz

Quote from: Roadgeekteen on December 21, 2024, 06:46:57 PM
Quote from: Quillz on December 21, 2024, 06:43:52 PMBesides, I think most AI is best exemplified by things like CarPlay/Android Auto. Where I can use buttons on my steering wheel to tell the car what I want (play this song, etc.) Or my mapping app of choice can update me with live traffic (car crash ahead, turn here to save time). This is the perfect use case for AI/automation, because it's simple and yet has a clear, practical impact on my life. My driving experience is actively better with these services because they recognize things I am likely to do (play a song, get directions), and streamline them. I get to work faster with these services.

Problem is, a fully self-driving car that is deciding where I am going and what routes I am taking is not as useful in that case. Waze, for example, is the extreme example. It will redirect me... Onto side roads that have stop signs. Yes, I'm technically saving two minutes, but I'm really not when I have to sit at a major street and wait for traffic to clear. I've had several times I'm at a stop sign so long it just redirects me. Because these services still do not seem to understand that theoretical time savings is not the same thing as practical time savings. That's where you need a human brain and manual override. I'm going to understand the traffic flow of a local street better than Waze.
A fully self-driving car has uses that you might not think of. With them, you could do a red-eyed drive like a flight being able to go overnight without having to worry about staying awake. Also, they would be very big for safety simply because drunk people coming home from the bar could just get driving home.
Yes, that's all true. But again, my take is I will believe all of this when it happens. Practical rollouts are very slow and gradual. And that's assuming these things even work, there will no doubt be many software bugs. Apple earlier this year showed off the next-gen CarPlay, said it would be up and running this year. Looks like that won't happen, not a single car manufacturer has gone ahead with it.

And you still have to deal with humanity. What happens if a drunk person decides to just drive home anyway? The NFL, for example, offers all of their players on-demand Uber. Specifically to counter the issue of drunk players. We saw what happened with Henry Ruggs.

Roadgeekteen

Quote from: Quillz on December 21, 2024, 06:50:45 PM
Quote from: Roadgeekteen on December 21, 2024, 06:46:57 PM
Quote from: Quillz on December 21, 2024, 06:43:52 PMBesides, I think most AI is best exemplified by things like CarPlay/Android Auto. Where I can use buttons on my steering wheel to tell the car what I want (play this song, etc.) Or my mapping app of choice can update me with live traffic (car crash ahead, turn here to save time). This is the perfect use case for AI/automation, because it's simple and yet has a clear, practical impact on my life. My driving experience is actively better with these services because they recognize things I am likely to do (play a song, get directions), and streamline them. I get to work faster with these services.

Problem is, a fully self-driving car that is deciding where I am going and what routes I am taking is not as useful in that case. Waze, for example, is the extreme example. It will redirect me... Onto side roads that have stop signs. Yes, I'm technically saving two minutes, but I'm really not when I have to sit at a major street and wait for traffic to clear. I've had several times I'm at a stop sign so long it just redirects me. Because these services still do not seem to understand that theoretical time savings is not the same thing as practical time savings. That's where you need a human brain and manual override. I'm going to understand the traffic flow of a local street better than Waze.
A fully self-driving car has uses that you might not think of. With them, you could do a red-eyed drive like a flight being able to go overnight without having to worry about staying awake. Also, they would be very big for safety simply because drunk people coming home from the bar could just get driving home.
Yes, that's all true. But again, my take is I will believe all of this when it happens. Practical rollouts are very slow and gradual. And that's assuming these things even work, there will no doubt be many software bugs. Apple earlier this year showed off the next-gen CarPlay, said it would be up and running this year. Looks like that won't happen, not a single car manufacturer has gone ahead with it.

And you still have to deal with humanity. What happens if a drunk person decides to just drive home anyway? The NFL, for example, offers all of their players on-demand Uber. Specifically to counter the issue of drunk players. We saw what happened with Henry Ruggs.
Calling an uber takes more time and money than just clicking the self drive option in the car. And eventually maybe some cars won't even have personal driving as an option.
My username has been outdated since August 2023 but I'm too lazy to change it

english si

Quote from: kalvado on December 21, 2024, 05:07:02 PM
Quote from: kernals12 on December 21, 2024, 05:02:32 PMThis study says that freeway lanes for autonomous cars could carry up to *12,000* vehicles per hour. 
I think 25 thousand is a better number. It looks nicer, and is equally realistic
Why stop at 25k? 100,000 is nice and round!

Quote from: Max Rockatansky on December 20, 2024, 10:38:39 PMAnd yet the HSR line here has a whole bunch of stops planned for non-dense areas.  That corridor should have directed lined between Los Angeles and San Francisco as a proof of concept.
Serving the Central Valley and providing some regeneration is worth the time penalty of a slightly longer route, not least because it is much less challenging to build than the crow-flies line that remains mountainous. The planned service for a fully-fleshed out line sees most trains skipping lower density/smaller city stops (though they still get served with an hourly or better service). You can get high speed trains close together (the biggest problem is switches, followed by station dwell times) and so you can run frequent services on several service patterns on a two-track railway (especially if stations are 4-track).

However, building the Central Valley section first and using that as test of principal for HSR in the US is silly - you have some 'beet field' stops alongside two decent-sized intermediate cities, fail to get to either metropolis and leaving a huge amount of capacity that is awaiting later phases to make use of it. It will look like a white elephant.

Quillz

Quote from: Roadgeekteen on December 21, 2024, 06:52:18 PMAnd eventually maybe some cars won't even have personal driving as an option.
Which demonstrates my point. What I don't want to happen. You always need to have some kind of manual override. What happens if the self-driving feature has a software bug? What if it does something you don't want it to do? Going 100% fully automated is going to have a lot of practical issues, and because no one ever will create truly perfect software, sooner or later the issues with that will reveal themselves.

A simple example is when Apple releases public beta software. There is always some weird or obscure issue that shows up because people have tons of use case scenarios that the programmers at Apple can't anticipate. Or how people will find bugs in Microsoft Windows 20 years after a particular version was released. The difference is those kind of bugs won't lead to car crashes and potential deaths.

Max Rockatansky

Quote from: english si on December 21, 2024, 08:03:39 PM
Quote from: kalvado on December 21, 2024, 05:07:02 PM
Quote from: kernals12 on December 21, 2024, 05:02:32 PMThis study says that freeway lanes for autonomous cars could carry up to *12,000* vehicles per hour. 
I think 25 thousand is a better number. It looks nicer, and is equally realistic
Why stop at 25k? 100,000 is nice and round!

Quote from: Max Rockatansky on December 20, 2024, 10:38:39 PMAnd yet the HSR line here has a whole bunch of stops planned for non-dense areas.  That corridor should have directed lined between Los Angeles and San Francisco as a proof of concept.
Serving the Central Valley and providing some regeneration is worth the time penalty of a slightly longer route, not least because it is much less challenging to build than the crow-flies line that remains mountainous. The planned service for a fully-fleshed out line sees most trains skipping lower density/smaller city stops (though they still get served with an hourly or better service). You can get high speed trains close together (the biggest problem is switches, followed by station dwell times) and so you can run frequent services on several service patterns on a two-track railway (especially if stations are 4-track).

However, building the Central Valley section first and using that as test of principal for HSR in the US is silly - you have some 'beet field' stops alongside two decent-sized intermediate cities, fail to get to either metropolis and leaving a huge amount of capacity that is awaiting later phases to make use of it. It will look like a white elephant.

One of the early concepts proposed was to run the HSR south from Pacheco Pass in the median of I-5.  The rationale for swinging the line east was supposedly having to deal with tunneling in Tejon Pass near the San Andreas Fault.  Considering much of the line around Pacheco Pass is in similarly seismically prone terrain that argument seemed faulty to me.

What I think really drove getting the HSR in the Central Valley was Fresno buying in lock stock and barrel as a form of urban renewal.  That and there probably was an assumption that farmers wouldn't fight against eminent domain tooth and nail.  The latter assumption turned out to be categorically false. 

kernals12

#67
Quote from: Quillz on December 21, 2024, 08:32:18 PMGoing 100% fully automated is going to have a lot of practical issues, and because no one ever will create truly perfect software, sooner or later the issues with that will reveal themselves.

Quote from: Quillz on December 21, 2024, 06:39:40 PM
Quote from: kernals12 on December 21, 2024, 06:35:43 PM
Quote from: Quillz on December 21, 2024, 06:27:12 PM
Quote from: Roadgeekteen on December 21, 2024, 11:44:07 AMIf ai took over the tsa I'm never flying again. I don't trust ai to keep airplanes safe.
Everything needs some kind of manual backup. Yes, the planes actually CAN fly themselves for the most part, but you still need a pilot as a backup. You always will. It's the same reason you can't fall asleep behind the wheel of your Tesla.

It's like the old Jurassic Park issue. Automation is great, but what happens if/when it fails and you've got no manual backups? Just because your automation works, doesn't mean the car next to you does. Since I've had my Subaru, which uses adaptive cruise control, I've found I am still overriding it a lot. Not because it doesn't work, but because it works too well: it will stop so suddenly it can be really jerky. I've found doing a manual override lets me get a smoother slowdown, and it will still always apply an emergency break anyway to prevent a crash.

As I said earlier in this thread, my takeaway is that all this stuff is cool in theory, but I'll believe it when it happens. Not to mention we can't even agree on what the "highway of tomorrow" will be. Remember the experiment of putting solar panels on them and having cars drive over them to generate electricity? It didn't work out in practice. Will "highways of tomorrow" be fully computer controlled like what we saw in "Minority Report," and thus be overridden if/when the powers that be end up not liking you or think you've done something wrong? What if I want to take a scenic route, but my smart car only lets me go the most optimal route?
AI, doesn't need to be perfect, it just needs to be better than humans.
What does "better" mean in this context, though? My car already has adaptive cruise control, and can do emergency braking. It can stop the car faster than I can, is that better? As I said in another example, what if I deliberately want to take a less direct, more scenic route? Will AI allow me? If not, why not? Sometimes "better" doesn't mean the most efficient or most direct, it can be mean maximizing what I want to see. That's the kind of thing where you need a manual human brain to decide. Not a programmer who is deciding for me.

Also, we can and have had many Teslas involved in self-driving crashes. "Not perfect" translates to a lot of the same issues that occur with humans behind the wheel. What about software bugs that have affected medical equipment and led to deaths? AI has to be programmed by humans, and history has shown it's not possible to have bug-free software and AI.

If we're talking about safety, I'd say "better" can be measured based on number of accidents per mile travelled.

vdeane

Quote from: Roadgeekteen on December 21, 2024, 06:43:19 PMAI self-driving cars will very likely let you choose what route just like Google Maps lets you choose your routes. Or if they somehow don't there are obviously work arounds. Like if you want to take CA 1 between LA and SF just set your final destination to the Big Sur, and when you are almost there change your final destination to SF. But again I see no reason why they wouldn't let you change your route. Self driving taxis/Ubers are different.
Even Google Maps won't let you do that on mobile, and I don't think others do even on desktop.  Incidentally, this route planning software is already here on electric cars.  If any of them have that option, I haven't heard of it.
Please note: All comments here represent my own personal opinion and do not reflect the official position of NYSDOT or its affiliates.

Roadgeekteen

Quote from: vdeane on December 21, 2024, 08:50:07 PM
Quote from: Roadgeekteen on December 21, 2024, 06:43:19 PMAI self-driving cars will very likely let you choose what route just like Google Maps lets you choose your routes. Or if they somehow don't there are obviously work arounds. Like if you want to take CA 1 between LA and SF just set your final destination to the Big Sur, and when you are almost there change your final destination to SF. But again I see no reason why they wouldn't let you change your route. Self driving taxis/Ubers are different.
Even Google Maps won't let you do that on mobile, and I don't think others do even on desktop.  Incidentally, this route planning software is already here on electric cars.  If any of them have that option, I haven't heard of it.
Just do a multi stop route among the route you want to go on. It takes more work as most normals don't care about this stuff but it's nowhere near impossible.
My username has been outdated since August 2023 but I'm too lazy to change it

Max Rockatansky

I'm all for the 150 MPH AI cars if they carry the normals away from my favorite back roads. 

english si

Quote from: Max Rockatansky on December 21, 2024, 08:43:03 PMOne of the early concepts proposed was to run the HSR south from Pacheco Pass in the median of I-5.  The rationale for swinging the line east was supposedly having to deal with tunneling in Tejon Pass near the San Andreas Fault.  Considering much of the line around Pacheco Pass is in similarly seismically prone terrain that argument seemed faulty to me.
That's a different question to the Central Valley. I-5 is in the Central Valley.

Serving Fresno and Bakersfield directly by following the CA99 corridor rather than the I-5 bypass-the-lot corridor makes sense. There's capacity for stopping trains and these are not small places. Serving them with Haute-Picardie-esque 'beet field' stations on I-5 is really bad. Other issues with the route (like not just following CA99 almost all the time to save on land costs) are other issues with the route than the serving of the Central Valley properly.

Serving Palmdale is much less of a good idea - hence the engineering (there's zero way a road median route through those mountains on any pass would ever work - too tight horizontal and vertical curvature for high speed (or even mid-speed), too steep grades for any railway, despite high speed rail being able to deal with steeper grades than other passenger rail) 'earthquakes' reasoning, rather than economic justification.

Max Rockatansky

Quote from: english si on December 22, 2024, 09:46:30 AM
Quote from: Max Rockatansky on December 21, 2024, 08:43:03 PMOne of the early concepts proposed was to run the HSR south from Pacheco Pass in the median of I-5.  The rationale for swinging the line east was supposedly having to deal with tunneling in Tejon Pass near the San Andreas Fault.  Considering much of the line around Pacheco Pass is in similarly seismically prone terrain that argument seemed faulty to me.
That's a different question to the Central Valley. I-5 is in the Central Valley.

Serving Fresno and Bakersfield directly by following the CA99 corridor rather than the I-5 bypass-the-lot corridor makes sense. There's capacity for stopping trains and these are not small places. Serving them with Haute-Picardie-esque 'beet field' stations on I-5 is really bad. Other issues with the route (like not just following CA99 almost all the time to save on land costs) are other issues with the route than the serving of the Central Valley properly.

Serving Palmdale is much less of a good idea - hence the engineering (there's zero way a road median route through those mountains on any pass would ever work - too tight horizontal and vertical curvature for high speed (or even mid-speed), too steep grades for any railway, despite high speed rail being able to deal with steeper grades than other passenger rail) 'earthquakes' reasoning, rather than economic justification.

My understanding on why the HSR switches from the Union Pacific to Southern Pacific was so there could be a dual Hanford-Visalia station.  Mind you, the station (along the one slated for Madera) isn't close to the downtown center of either city.

PColumbus73

https://www.aaroads.com/forum/index.php?topic=35143.msg2951239#msg2951239

Since we're bringing up the idea of chaining vehicles together... again, see the question I posed in the '150 MPH Highways' discussion that went unanswered:

QuoteVehicles entering and exiting the highway are going to create unpredictable gaps in traffic. Assuming they're all AI, they would have to manipulate the cars so they can merge or exit without colliding, so either traffic on the highway gets throttled to fit incoming traffic, or all the ramps are essentially metered, including traffic transitioning from another freeway onto the 150 MPH highways.

And as referenced, each vehicle has its own destination, so in order to chain them efficiently, you'd have to connect cars going to similar destinations. But how does that happen when cars are freely entering and exiting randomly?

Quote from: kernals12 on September 27, 2024, 04:51:25 PM
Quote from: michravera on September 27, 2024, 01:35:24 PM
Quote from: thspfc on September 26, 2024, 05:21:43 PMThis is at least 30 years away from being a topic for legitimate discussion on a road forum.

Basic math says that, with 2 seconds of following time, you can only do 1800 vehicles per hour (that's neglecting the length of a vehicle). Cut the time to 1 second and you double the throughput to 3600. Yeah, cut the following time to 0.45 seconds and you'd get 8000, but at 20 m/s (72 km/h), that's only 9 m between the fronts of vehicles. Chevy Suburbans are just 5.73 m. That not only gives only 0.16 s of reaction time, but requires that lead vehicles not break much faster than the trailing vehicle. There's only roughly the width of a lane between bumpers.

It'd be a train basically

And considering K12 is acknowledging that we are essentially reinventing trains, it's worth noting that how moving cargo via train involves several rail yards where rail cars are sorted and assembled into train sets. To apply this to personal vehicles would be combining the wait times of ferries with toll plazas as drivers would have to wait in these yards for other people who are going to similar destinations. The closest example is probably people who drive their cars through the English Channel, either by ferry or by train through the Channel Tunnel. Whatever value there is in 'chaining' a bunch of vehicles together would have to offset the delays that would come with sitting in a staging area for an undetermined amount of time.

Secondly, there was an incident at a drone show in Orlando where they malfunctioned and caused at least one injury. I think that's relevant in this discussion because the people that are advocating for cars to be connected to each other through AI and traveling at highway speeds (either 70-150 MPH) seem to believe that AI malfunctioning is either not that big of a deal, or an acceptable risk. So, depending on how you look at it, a malfunctioning AI 'chain' is either going to be an inconvenience when cars go through a dead zone, or we may have an increase in multi-car pile-ups.

kernals12

Quote from: PColumbus73 on December 22, 2024, 10:45:18 PMhttps://www.aaroads.com/forum/index.php?topic=35143.msg2951239#msg2951239

Since we're bringing up the idea of chaining vehicles together... again, see the question I posed in the '150 MPH Highways' discussion that went unanswered:

QuoteVehicles entering and exiting the highway are going to create unpredictable gaps in traffic. Assuming they're all AI, they would have to manipulate the cars so they can merge or exit without colliding, so either traffic on the highway gets throttled to fit incoming traffic, or all the ramps are essentially metered, including traffic transitioning from another freeway onto the 150 MPH highways.

And as referenced, each vehicle has its own destination, so in order to chain them efficiently, you'd have to connect cars going to similar destinations. But how does that happen when cars are freely entering and exiting randomly?

Quote from: kernals12 on September 27, 2024, 04:51:25 PM
Quote from: michravera on September 27, 2024, 01:35:24 PM
Quote from: thspfc on September 26, 2024, 05:21:43 PMThis is at least 30 years away from being a topic for legitimate discussion on a road forum.

Basic math says that, with 2 seconds of following time, you can only do 1800 vehicles per hour (that's neglecting the length of a vehicle). Cut the time to 1 second and you double the throughput to 3600. Yeah, cut the following time to 0.45 seconds and you'd get 8000, but at 20 m/s (72 km/h), that's only 9 m between the fronts of vehicles. Chevy Suburbans are just 5.73 m. That not only gives only 0.16 s of reaction time, but requires that lead vehicles not break much faster than the trailing vehicle. There's only roughly the width of a lane between bumpers.

It'd be a train basically

And considering K12 is acknowledging that we are essentially reinventing trains, it's worth noting that how moving cargo via train involves several rail yards where rail cars are sorted and assembled into train sets. To apply this to personal vehicles would be combining the wait times of ferries with toll plazas as drivers would have to wait in these yards for other people who are going to similar destinations. The closest example is probably people who drive their cars through the English Channel, either by ferry or by train through the Channel Tunnel. Whatever value there is in 'chaining' a bunch of vehicles together would have to offset the delays that would come with sitting in a staging area for an undetermined amount of time.

Secondly, there was an incident at a drone show in Orlando where they malfunctioned and caused at least one injury. I think that's relevant in this discussion because the people that are advocating for cars to be connected to each other through AI and traveling at highway speeds (either 70-150 MPH) seem to believe that AI malfunctioning is either not that big of a deal, or an acceptable risk. So, depending on how you look at it, a malfunctioning AI 'chain' is either going to be an inconvenience when cars go through a dead zone, or we may have an increase in multi-car pile-ups.

Computers can run trillions of calculations per second, I think that forming and dismantling platoons on the fly is well within their capabilities.



Opinions expressed here on belong solely to the poster and do not represent or reflect the opinions or beliefs of AARoads, its creators and/or associates.