Ed Self-Driving Cars: Pros, Cons, and Predictions

[bruto]

Edited by jimbob: 

quote of subsequently moderated content removed

I, on the other hand, am pretty confident the new technology will continue to improve apace, and become very reliable and mechanically functional. I just don't think we'll it's the right solution to the right problems. For at least a little while, a genie works better out of the bottle than in.

 

[Gulliver Foyle]

I, on the other hand, am pretty confident the new technology will continue to improve apace, and become very reliable and mechanically functional. I just don't think we'll it's the right solution to the right problems. For at least a little while, a genie works better out of the bottle than in.

Given that the industry that has used self driving fot decades, the airline industry, only uses it as an aid and still has human pilots in charge who can override at a touch of a button if needed and still have to do the most complicated and accident prone driving, I'd say self driving is at best decades away from being viable on the ground.

And even then the best use scenario is for short distance urban mass transit on predictable routes with segregated lanes, i.e. light rail and bus routes separated from normal traffic.

 

[Rolfe]

Have a look at the videos made by Tesla drivers in the USA who have the latest software, though. It's pretty impressive. I really don't know how far that principle (which is quite different from the way the existing robotaxis work) will be able to go, but they're certainly doing better than short distance urban driving on predictable routes.

 

[bruto]

Given that the industry that has used self driving fot decades, the airline industry, only uses it as an aid and still has human pilots in charge who can override at a touch of a button if needed and still have to do the most complicated and accident prone driving, I'd say self driving is at best decades away from being viable on the ground.

And even then the best use scenario is for short distance urban mass transit on predictable routes with segregated lanes, i.e. light rail and bus routes separated from normal traffic.

Probably so, but I think the possibility exists that there will be a reversal of a sort as the technology progresses. At this point, the self driving car is required to work well in the existing environment, but when it becomes sufficiently usable to start being seen as a necessity, the environment becomes the thing that must change to accommodate it. I imagine this is quite a way off, but the decades roll by pretty quickly.

 

[Brainster]

That said, though... He does raise some good points, like that 'safer than regular cars' isn't as good an argument as people imagine it to be. Cars are incredibly dangerous, so that's kind of like saying 'safer than smoking cigarettes', isn't it?

Unless you are going to ban cars entirely, then safer than human-driven cars is actually a very strong argument for driver-less vehicles.

 

[sir drinks-a-lot]

I finally decided to try taking a Waymo a few weeks ago in SF. The cost wasn't much different than Lyft or Uber - it amy have even been slightly more expensive. I figured I'd try it for the novelty. Overall, it was pretty cool. You unlock the doors with the app and get inside, then a robot-voice tells you the ground rules. You can choose from some music playlists, or roll your own. The route I took was a nice, scenic, four miles through the mission area of the city. It handled everything pretty well, including some blind hills. It also seemed to comprehend pedestrians waving it to "go ahead" in crosswalks.

Being a resident of SF, of course I've heard about coning and some of the issues. I've seen a couple issues myself. One, where a way was in some sort of confused mode, sitting pretty far back in a left turn lane with its blinker on, but not moving. It stayed there for a while, and cars were going around it.

The second issue I saw was even worse. The car was confused and sort of sitting in place, but also moving around a bit. It would pull forward a few feet, back up the sam amount, and occasionally turn a little bit during the process. Still, nothing major, as it was on the side of a not very busy street. I've heard about them interfering with emergency vehicles on occasion. Not sure how serious this might turn out to be.

Overall, I think they have a lot of potential, especially in SF - which might not be a great test case for other cities.

 

[The Great Zaganza]

Tesla doesn't stand a chance against the tech coming out of China, as it doesn't have the money, manpower or scale of production.
It will be left behind, technologically, if it isn't already.

 

[Rolfe]

Tesla doesn't stand a chance against the tech coming out of China, as it doesn't have the money, manpower or scale of production.
It will be left behind, technologically, if it isn't already.

Maybe, but I don't see it yet. The main difference with Tesla is that it doesn't seem to require a detailed map of a defined area with clear markings, and a geofence to prevent it going beyond that. I'm not aware what method(s) are under development in China.

 

[Rolfe]

Unless you are going to ban cars entirely, then safer than human-driven cars is actually a very strong argument for driver-less vehicles.

There is an issue though. If an accident occurs due to human error, other human beings are more disposed to understanding and accepting this - even though they may be fulminating with rage against the idiot responsible. Idiots are a fact of life, we didn't create them on purpose. Conversely, if a driverless car causes an accident, the nature of the fury is going to be different.

It may be a temporary issue, but at the moment I think people are more accepting of "drunk driver causes crash" than they are likely to be of "newfangled tech causes crash".

 

[The Great Zaganza]

Maybe, but I don't see it yet. The main difference with Tesla is that it doesn't seem to require a detailed map of a defined area with clear markings, and a geofence to prevent it going beyond that. I'm not aware what method(s) are under development in China.

China has way, way, waaaay more data it can use to train self-driving algorithms on, and even less data protection.
They are also investing heavily in A.I..
Tesla simply cannot compete.

 

[Rolfe]

China has way, way, waaaay more data it can use to train self-driving algorithms on, and even less data protection.
They are also investing heavily in A.I..
Tesla simply cannot compete.

That may well be so, but I haven't seen it yet.

 

[Disbelief]

I finally decided to try taking a Waymo a few weeks ago in SF. The cost wasn't much different than Lyft or Uber - it amy have even been slightly more expensive. I figured I'd try it for the novelty. Overall, it was pretty cool. You unlock the doors with the app and get inside, then a robot-voice tells you the ground rules. You can choose from some music playlists, or roll your own. The route I took was a nice, scenic, four miles through the mission area of the city. It handled everything pretty well, including some blind hills. It also seemed to comprehend pedestrians waving it to "go ahead" in crosswalks.

Being a resident of SF, of course I've heard about coning and some of the issues. I've seen a couple issues myself. One, where a way was in some sort of confused mode, sitting pretty far back in a left turn lane with its blinker on, but not moving. It stayed there for a while, and cars were going around it.

The second issue I saw was even worse. The car was confused and sort of sitting in place, but also moving around a bit. It would pull forward a few feet, back up the sam amount, and occasionally turn a little bit during the process. Still, nothing major, as it was on the side of a not very busy street. I've heard about them interfering with emergency vehicles on occasion. Not sure how serious this might turn out to be.

Overall, I think they have a lot of potential, especially in SF - which might not be a great test case for other cities.

SF was chosen specifically for the challenging environment with the assumption that other cities would be easier once many of the difficulties were overcome. At least that's a significant reason why Cruise chose it, and I would assume the same for Waymo.

 

[crescent]

There is an issue though. If an accident occurs due to human error, other human beings are more disposed to understanding and accepting this - even though they may be fulminating with rage against the idiot responsible. Idiots are a fact of life, we didn't create them on purpose. Conversely, if a driverless car causes an accident, the nature of the fury is going to be different.

It may be a temporary issue, but at the moment I think people are more accepting of "drunk driver causes crash" than they are likely to be of "newfangled tech causes crash".

Tying in with this:

Self driving cars, even if they have fewer crashes overall, may still get into crashes that people could have avoided. Conversely, they could avoid crashes that people don't avoid. The patterns of crash types will be different, and that may result in misunderstanding. You see a driverless car get into a wreck that a human would have avoided, and it gets hard to comprehend that the system may have fewer crashes overall than humans do.

 

[The Great Zaganza]

That may well be so, but I haven't seen it yet.

The biggest advantage or disadvantage China has, depending on how you look at it, is that the US is a ◊◊◊◊◊◊◊◊ country when it comes to road infrastructure.
That means that Chinese self-driving cars might have a hard time in the US, but they will do fine in Europe.

 

[Rolfe]

Tying in with this:

Self driving cars, even if they have fewer crashes overall, may still get into crashes that people could have avoided. Conversely, they could avoid crashes that people don't avoid. The patterns of crash types will be different, and that may result in misunderstanding. You see a driverless car get into a wreck that a human would have avoided, and it gets hard to comprehend that the system may have fewer crashes overall than humans do.

That is probably a better and clearer explanation than mine. People are likely to accept the sort of accidents people cause, even if a driverless car could have avoided them, and go batcrap crazy about accidents caused by self-driving cars which a human driver would have been able to avoid. That will apply even if the latter category of accident is demonstrably less common than the former.

It's similar, though not identical, to the current situation with EV fires. ICE cars are demonstrably 20 to 60 times more likely to catch fire than EVs. Three hundred a day in Britain alone. They're so common they're often not reported, or consigned to the inner pages of a local rag with some sort of human interest hook like "local Mum's lucky escape" or "buses re-routed last Tuesday". It takes a really spectacular incident like the Luton car park fire to get on the front page. (And even then someone will try to blame it on an EV.) In contrast an EV fire gets plastered all over the front pages of the Mail and the Sun and so on, even if nobody was injured. It seems to be human nature to accept the adverse events we're used to but to get very aerated about any adverse events associated with something new, even if they're much less frequent.

 

[Rolfe]

The biggest advantage or disadvantage China has, depending on how you look at it, is that the US is a ◊◊◊◊◊◊◊◊ country when it comes to road infrastructure.
That means that Chinese self-driving cars might have a hard time in the US, but they will do fine in Europe.

Likely to be a disaster in Britain, then.

 

[Brainster]

There is an issue though. If an accident occurs due to human error, other human beings are more disposed to understanding and accepting this - even though they may be fulminating with rage against the idiot responsible. Idiots are a fact of life, we didn't create them on purpose. Conversely, if a driverless car causes an accident, the nature of the fury is going to be different.

It may be a temporary issue, but at the moment I think people are more accepting of "drunk driver causes crash" than they are likely to be of "newfangled tech causes crash".

That people will react irrationally is understandable but we should base policy on what causes the least total harm while providing the maximum possible benefits.

And what (other than lawsuits and letters to the editor) are the consequences of people not accepting "newfangled tech causes crash?" Are concerned local citizens going to form a lynch mob to string up the offending Waymo?

 

[Rolfe]

It's an issue that will probably have to be recognised and addressed. At the moment the unfounded hysteria about EV fires is dissuading people from buying EVs, leading to political consequences such as the current call to relax EV mandates because "people aren't as keen to buy EVs as we thought they would be." A bit more education about the real frequency and effects/consequences of EV fires might have helped. Perhaps a lesson to be taken forward.

 

[Disbelief]

Tying in with this:

Self driving cars, even if they have fewer crashes overall, may still get into crashes that people could have avoided. Conversely, they could avoid crashes that people don't avoid. The patterns of crash types will be different, and that may result in misunderstanding. You see a driverless car get into a wreck that a human would have avoided, and it gets hard to comprehend that the system may have fewer crashes overall than humans do.

The whole idea behind self driving is to avoid situations before they even occur. While the pattern type will change, in theory it should be greatly reduced because the autonomous car will be more likely avoid situations where there really is no good option.

 

[Rolfe]

Someone pointed out that as the proportion of self-driving cars increases, so that self-driving cars will often be interacting with other self-driving cars, safety should increase a lot as each car should be able to predict what the other will do to a fairly high degree of accuracy,