Washington Post: For the first time, Google's self-driving car takes some blame for a crash (https://www.washingtonpost.com/news/innovations/wp/2016/02/29/for-the-first-time-googles-self-driving-car-takes-some-blame-for-a-crash/)
How does a machine admit something?
(I get it, but it's a lousily written headline.)
Quoteits cars will learn that large vehicles are less likely to yield than other types of vehicles
In other words, the Goog is ramming its way through traffic, expecting others to yield.
Quote from: NE2 on February 29, 2016, 04:43:38 PM
Quoteits cars will learn that large vehicles are less likely to yield than other types of vehicles
In other words, the Goog is ramming its way through traffic, expecting others to yield.
Ramming...at 2 mph.
It says the Google car hit the bus in the center at the accordion fold. How could the bus possibly have yielded if it was already mostly ahead of the Google car?
Such reminds me of this
Farmers Insurance commercial from a couple of years ago.
Having a human operator there to intervene over a dozen times to prevent accidents doesn't seem like a very good track record to me.
Personally, I hope something like this never fully catches on.
Cell phones were supposed to be great, we could contact our loved ones when we wanted, but at the same time, the corporate world could contact you and tie you down to your job longer than before.
For many, drive time is the only time some people get to themselves, even if they are in heavy traffic, if the commute than becomes used as more time to spend at work as tasks will be done in the vehicle, I have to tell you, I really do not like that world.
Quote from: Max Rockatansky on March 02, 2016, 08:39:56 AM
Having a human operator there to intervene over a dozen times to prevent accidents doesn't seem like a very good track record to me.
Why not? You're dealing with computers, which while the programmers try to determine every type of scenario prior to being deployed in the field, there's always going to be something unusual or different that they come across and have to deal with.
If you can do something with 100% accuracy the first time, then you're better than everyone who's walked on the face of this earth.
Quote from: jeffandnicole on March 02, 2016, 09:52:23 AM
Quote from: Max Rockatansky on March 02, 2016, 08:39:56 AM
Having a human operator there to intervene over a dozen times to prevent accidents doesn't seem like a very good track record to me.
Why not? You're dealing with computers, which while the programmers try to determine every type of scenario prior to being deployed in the field, there's always going to be something unusual or different that they come across and have to deal with.
If you can do something with 100% accuracy the first time, then you're better than everyone who's walked on the face of this earth.
I'm not expecting 100% infallibility, but I would expect reliability. Does autopilot on a plane prevent 100% of crashes or adapt for every scenario?...of course not. The problem is that air planes have a much wider wider berth of traffic at 30,000 feet as opposed to traffic in down Los Angeles during rush hour. Basically from the test data that has been published thus far it doesn't suggest that the automated cars are nearly as reliable as Google thought they might be or what they spin in the public eye. I understand that there is a demand for such a technology with all the miserable commutes people have to endure especially in larger U.S. cities but should be proven reliable first. It just reminds me of all the hype behind flying cars in the 50s and 60s which at the end of the day couldn't ever be made practical due to several factors; one being that the skill level of every vehicle operator would have to be unrealistically high. Basically what Google needs to strive for if they truly serious is a system that has a proven track record that shows the automation is much more reliable than the average commuter not similar.
But to echo the sentiment of the poster previous to you, I don't think that I would ever be interested in automating my driving. I avoided having even a cell until 2009 when I was handed one by my company, that basically took away close to all my time out of the house where I could truly be left alone. I might be in the minority but it seems like most people are more willing to give up their personal autonomy in for 100% connectivity these days than not.
Quote from: NE2 on February 29, 2016, 04:43:38 PM
Quoteits cars will learn that large vehicles are less likely to yield than other types of vehicles
In other words, the Goog is ramming its way through traffic, expecting others to yield.
So would this be a violation of the first law of robotics?
Quote from: WillWeaverRVA on March 02, 2016, 10:30:59 AM
Quote from: NE2 on February 29, 2016, 04:43:38 PM
Quoteits cars will learn that large vehicles are less likely to yield than other types of vehicles
In other words, the Goog is ramming its way through traffic, expecting others to yield.
So would this be a violation of the first law of robotics?
Didn't Sky Net and the Cylons do the same thing? Does that mean we'll have Evil Google Car overlords ruling the Earth or this?
(https://www.aaroads.com/forum/proxy.php?request=http%3A%2F%2Fwww.geekzilla.com.br%2Fwp-content%2Fuploads%2F2012%2F05%2Fgoogle_street_view_car_wifi1.jpg&hash=6c3bfd837fd438c388d421b2bda8971022ad1856)
Quote from: MisterSG1 on March 02, 2016, 09:51:14 AM
Personally, I hope something like this never fully catches on.
Cell phones were supposed to be great, we could contact our loved ones when we wanted, but at the same time, the corporate world could contact you and tie you down to your job longer than before.
For many, drive time is the only time some people get to themselves, even if they are in heavy traffic, if the commute than becomes used as more time to spend at work as tasks will be done in the vehicle, I have to tell you, I really do not like that world.
It won't fully catch on until there's an interface that can allow a user to precisely place a vehicle where it needs to be.
There are many vehicles that need to be placed and moved and adjusted for reasons of need or personal preference. Shining your headlights on your campsite. Lining up a dumptruck to pour out in a specific spot (and all kinds of construction finagling). Getting the car in line with the most romantic view in front of which to [great, now there are children around and I have to edit my thoughts].
Quote from: Max Rockatansky on March 02, 2016, 10:27:40 AM
Quote from: jeffandnicole on March 02, 2016, 09:52:23 AM
Quote from: Max Rockatansky on March 02, 2016, 08:39:56 AM
Having a human operator there to intervene over a dozen times to prevent accidents doesn't seem like a very good track record to me.
Why not? You're dealing with computers, which while the programmers try to determine every type of scenario prior to being deployed in the field, there's always going to be something unusual or different that they come across and have to deal with.
If you can do something with 100% accuracy the first time, then you're better than everyone who's walked on the face of this earth.
I'm not expecting 100% infallibility, but I would expect reliability. Does autopilot on a plane prevent 100% of crashes or adapt for every scenario?...of course not. The problem is that air planes have a much wider wider berth of traffic at 30,000 feet as opposed to traffic in down Los Angeles during rush hour. Basically from the test data that has been published thus far it doesn't suggest that the automated cars are nearly as reliable as Google thought they might be or what they spin in the public eye. I understand that there is a demand for such a technology with all the miserable commutes people have to endure especially in larger U.S. cities but should be proven reliable first. It just reminds me of all the hype behind flying cars in the 50s and 60s which at the end of the day couldn't ever be made practical due to several factors; one being that the skill level of every vehicle operator would have to be unrealistically high. Basically what Google needs to strive for if they truly serious is a system that has a proven track record that shows the automation is much more reliable than the average commuter not similar.
Isn't that what they're striving for? I don't know of any car for sale that's fully automated yet. You're acting like these cars are all over the place, rather than the limited number that are being used for exactly what you're saying.
The difference between flying cars (which never really existed and were mostly Jetsons creations and sci-fi dreams) and automated cars are night and day. And it had a lot more to do with what would be necessary to actually build a flying car.
Many people aren't envisioning automated cars because they are too narrowly thinking of their capabilities. Then again, many people aren't even aware of automation technologies in use already, even in their own cars. Give it 30 years, and these same people will wonder how we got around by actually driving ourselves.
Since the first function of a GPS is to get you lost, I have faith that we will never see a self-driving car. The variable are just too great.
And, why would you want one anyway. Driving is fun. If you don't like driving, take the bus.
I'm eager to watch the absurd political battle that ensues when a country obssessed with NASCAR has its driving privileges revoked.
Quote from: NE2 on February 29, 2016, 04:43:38 PM
Quoteits cars will learn that large vehicles are less likely to yield than other types of vehicles
In other words, the Goog is ramming its way through traffic, expecting others to yield.
Which, to be fair, is exactly what human drivers do, and it works just fine until the driver you're expecting to yield is distracted and doesn't.
There are many situations in driving where there's following the law to a T, and then there's what's realistic. A hyper literal interpretation of the law about yielding when turning into traffic would be that you need to wait for a big enough space where you can get up to speed without any vehicles behind you needing to slow down before you can turn. In many real world circumstances, however, the amount of traffic on the road you're turning onto is such that you'll be waiting all day for a gap this big if that's what you insist on. So, in the interest of keeping everyone moving (there may well be other vehicles behind you waiting to turn), you find a gap which is big enough that you can safely enter it and give the driver of the next vehicle ample time to react to you doing so, but not big enough that no reaction on their part is required.
All that said, I find the idea of the Google car inching out into traffic at 2 MPH to be... odd. Seems to me it shouldn't be advancing that slowly, since this represents excessively timid behavior. The bus probably didn't yield because the driver didn't realize the Google car was pulling out, or assumed that the Google car would automatically stop to avoid a collision.
This is perhaps a potential pitfall of mixing human and non-human drivers: the non-human drivers will, of course, in the name of safety be programmed to do whatever they can to avoid a collision. But the human drivers will recognize this and some will take full advantage of it, acting like they can cut off the robot car all they want because the robot car will get out of their way and not fight back.
I would expect this to hold especially true in areas where aggressive driving is common and normal.
We don't even have fully automated trains yet outside of subway/monorail systems despite trains not needing to worry about free form turns, merging, etc, since all of that is done by track switches and not the train. Additionally, the railroad controls what types of trains are allowed on it, so there is no issue of mixing old and new equipment as with cars. Even most "driverless" subway systems have an operator on board at all times. So I suspect driverless cars are pretty far off.
Quote from: bzakharin on March 03, 2016, 12:53:59 PM
We don't even have fully automated trains yet outside of subway/monorail systems despite trains not needing to worry about free form turns, merging, etc, since all of that is done by track switches and not the train. Additionally, the railroad controls what types of trains are allowed on it, so there is no issue of mixing old and new equipment as with cars. Even most "driverless" subway systems have an operator on board at all times. So I suspect driverless cars are pretty far off.
This is a very good point.
If disrupting train technology was sexier, Google could have made a lot more impact by now if they'd put a fifth this much energy into automatic train control.
Put one of these on. (http://fallout.wikia.com/wiki/Protectron_%28Fallout_4%29)
Kidding aside, why not just scrap self driving cars altogether? This is prototype technology that should be done under control in a lab, not just drop it onto the road.
Quote from: noelbotevera on March 03, 2016, 07:23:19 PM
Put one of these on. (http://fallout.wikia.com/wiki/Protectron_%28Fallout_4%29)
Kidding aside, why not just scrap self driving cars altogether? This is prototype technology that should be done under control in a lab, not just drop it onto the road.
..... Do you really think they didn't spend
years developing this in a lab before they started testing it in the real world? It may not be working with 100% reliability yet, but technology rarely if ever does. At some point you need to move from the lab to the test phase. They didn't just "drop it on the road".
Quote from: jeffandnicole on March 02, 2016, 11:27:31 AM
Quote from: Max Rockatansky on March 02, 2016, 10:27:40 AM
Quote from: jeffandnicole on March 02, 2016, 09:52:23 AM
Quote from: Max Rockatansky on March 02, 2016, 08:39:56 AM
Having a human operator there to intervene over a dozen times to prevent accidents doesn't seem like a very good track record to me.
Why not? You're dealing with computers, which while the programmers try to determine every type of scenario prior to being deployed in the field, there's always going to be something unusual or different that they come across and have to deal with.
If you can do something with 100% accuracy the first time, then you're better than everyone who's walked on the face of this earth.
I'm not expecting 100% infallibility, but I would expect reliability. Does autopilot on a plane prevent 100% of crashes or adapt for every scenario?...of course not. The problem is that air planes have a much wider wider berth of traffic at 30,000 feet as opposed to traffic in down Los Angeles during rush hour. Basically from the test data that has been published thus far it doesn't suggest that the automated cars are nearly as reliable as Google thought they might be or what they spin in the public eye. I understand that there is a demand for such a technology with all the miserable commutes people have to endure especially in larger U.S. cities but should be proven reliable first. It just reminds me of all the hype behind flying cars in the 50s and 60s which at the end of the day couldn't ever be made practical due to several factors; one being that the skill level of every vehicle operator would have to be unrealistically high. Basically what Google needs to strive for if they truly serious is a system that has a proven track record that shows the automation is much more reliable than the average commuter not similar.
Isn't that what they're striving for? I don't know of any car for sale that's fully automated yet. You're acting like these cars are all over the place, rather than the limited number that are being used for exactly what you're saying.
The difference between flying cars (which never really existed and were mostly Jetsons creations and sci-fi dreams) and automated cars are night and day. And it had a lot more to do with what would be necessary to actually build a flying car.
Many people aren't envisioning automated cars because they are too narrowly thinking of their capabilities. Then again, many people aren't even aware of automation technologies in use already, even in their own cars. Give it 30 years, and these same people will wonder how we got around by actually driving ourselves.
Well considering I'm a Californian the test vehicles are a much more common sight, but by no means do I consider them an everyday thing....I've only seen one of the test vehicles maybe two or three times in my travels. 30 years is probably what it will take for the technology to become reliable and affordable enough to become main stream. The interesting thing to me is the way this technology is presented by most media sources is as if it's something the masses will be able to use in just a year or two...it might be available to extent but likely cost an arm and a leg until the price is driven down. A similar comparison I would make is internet service. I remember my Dad talking about how Dad in the late 80s was always trying to promote the internet as a reliable news source, a place you could do business, find entertainment, shop, ect. The proof of concept was there at the time but the technology was incredibly clunky and incredibly expensive at the time. Over the next 25 years it became much more reliable, efficient and much affordable....which brings us to talking on an internet message board. In the vein of the topic at hand there has been many advances in alternate fuel sources in cars such hybrids, full on battery powered motors and even hydrogen. With those technologies were kind at a mid-point where most of them have proven reliability, have had a market presence and for the most part are becoming much more affordable. Hell the manual transmission is on the verge of going extinct completely from the current automotive market.
Personally for me, I tend to have a passion for driving but certainly not commuting. In that thirty years if I am still alive I likely would be in my mid-60s towards the end of my workforce lifespan. I probably would consider the technology if it was something affordable but I wouldn't go out of my way to over pay for it. But then again that's why I don't live in some of the bigger cites like San Diego and L.A. anymore, the quality on life just in general much less the commutes is gradually more and more soul crushing...at least for me. BUT, no automated car is ever going to take away the enjoyment of a mountainside highway climb in a performance oriented car.
Quote from: noelbotevera on March 03, 2016, 07:23:19 PM
Put one of these on. (http://fallout.wikia.com/wiki/Protectron_%28Fallout_4%29)
Kidding aside, why not just scrap self driving cars altogether? This is prototype technology that should be done under control in a lab, not just drop it onto the road.
You can only go so far with proving a technology in simulated lab scenarios. Take a look at the history of rail, automotive and even powered flight technology. Most of the advances that led to those technologies becoming mainstream were done in the real world and were much more haphazard than anything we would accept by modern standards. Even thought I'm incredibly skeptical about where the automated technology is at, it has progressed enough that it needs to field tested to push it forward. I honestly don't see how Google is going to get an ROI anytime soon but it's their money and they have the deep pockets to keep dropping cash into research for a long time.
Self driving technology will is going to be implemented step by step. It's a process that's already underway. There are cars that brake automatically in some situations. Other cars sort of parallel park themselves. Lane departure and blindspot warning features are becoming more common. The elements are in place for the next logical step: computer control in a semi-contained environment like a rural freeway during dry weather.
They will keep adding features so there are more and more situations where the computer is the better option for vehicle control for most people. It's going to be a generation-long process and even then, we are probably only going to end up with a semi-autonomous system available to us. But just that is going to save thousands of lives and thousands more serious injuries. Not to mention millions of dollars in damages.
With that as a goal, this little nudge by Google's car is such a minor hiccup along the way. The engineers and programmers are going to learn a lot from troubleshooting this minor failure that will prevent more calamitous incidents with their vehicles in the future.
Quote from: bzakharin on March 03, 2016, 12:53:59 PM
We don't even have fully automated trains yet outside of subway/monorail systems despite trains not needing to worry about free form turns, merging, etc, since all of that is done by track switches and not the train. Additionally, the railroad controls what types of trains are allowed on it, so there is no issue of mixing old and new equipment as with cars. Even most "driverless" subway systems have an operator on board at all times. So I suspect driverless cars are pretty far off.
Yeah, we definitely need to do this in a tiered approach:
- Automate grade-separated metro lines (no crossings). Already done in some areas (Vancouver is very famous for its automated, 3-line system).
- Automate railroads (metro and light rail lines as well) that do not run in traffic but do have at-grade crossings.
- Automate light rail lines and streetcars that run in mixed traffic.
- Automate buses on fixed routes
- Automate buses doing demand response services (DART/dial-a-ride/paratransit)
- Automate regular automobiles
Sliding up the difficulty scale a bit between light rail and buses, but it's doable.
There's been talk here in Seattle about automating our future light rail lines (which could be built to be fully grade-separated) and leaving the current one (with its at-grade crossings) as a driver-operated one until technology catches up.
https://www.youtube.com/watch?v=odA_EVwVycU
Quote from: Bruce on March 05, 2016, 12:54:34 AM
Quote from: bzakharin on March 03, 2016, 12:53:59 PM
We don't even have fully automated trains yet outside of subway/monorail systems despite trains not needing to worry about free form turns, merging, etc, since all of that is done by track switches and not the train. Additionally, the railroad controls what types of trains are allowed on it, so there is no issue of mixing old and new equipment as with cars. Even most "driverless" subway systems have an operator on board at all times. So I suspect driverless cars are pretty far off.
Yeah, we definitely need to do this in a tiered approach:
- Automate grade-separated metro lines (no crossings). Already done in some areas (Vancouver is very famous for its automated, 3-line system).
- Automate railroads (metro and light rail lines as well) that do not run in traffic but do have at-grade crossings.
- Automate light rail lines and streetcars that run in mixed traffic.
- Automate buses on fixed routes
- Automate buses doing demand response services (DART/dial-a-ride/paratransit)
- Automate regular automobiles
Sliding up the difficulty scale a bit between light rail and buses, but it's doable.
There's been talk here in Seattle about automating our future light rail lines (which could be built to be fully grade-separated) and leaving the current one (with its at-grade crossings) as a driver-operated one until technology catches up.
Won't driverless buses require crimeless cities?