snopes.com  

Go Back   snopes.com > Non-UL Chat > Techno-Babble

Reply
 
Thread Tools Display Modes
  #41  
Old 19 September 2017, 11:17 PM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

Seems every generation of engineers has to learn the hard way that all the technology in the world isn't enough if its design doesn't include the rest of the world. There's no magic prototype or government spending or manual or rules and laws or expert users or user interface or anything that's going to somehow transform a frog into a prince - no matter how wonderfully 'engineered*' it is, no matter how well it supposedly drives better than a human. All of those things have to be designed with the system by people who understand both very well. Otherwise, we're looking at two more decades of futzing around (technical term for trial and error) with these cars, just like we did with cell phones, the Internet, the automobile itself, the computer mouse, etc etc. It typically takes two decades. And that hasn't started yet because there are no self driving cars. (Maybe someone will come along, a magic Jobs or whatever - someone who finally gets it - to shorten that period by, what, maybe three years tops? But don't count on it. I'm as old as the mouse, for example, but it didn't really become a thing until I was out of high school. **)

* 'Engineered' in quotes because designing tech, without designing how it works in the real world with real people and businesses and (yes) laws and other tech, is not engineering. That's just making stuff. Big difference.

** And anyone who says 1984, fine, but the actual device is actually several years older than me. So, yep, two decades. Just a very very simple thing.

Last edited by ganzfeld; 19 September 2017 at 11:36 PM.
Reply With Quote
  #42  
Old 20 September 2017, 08:53 AM
erwins's Avatar
erwins erwins is offline
 
Join Date: 04 April 2006
Location: Portland, OR
Posts: 11,356
Default

Quote:
Originally Posted by GenYus234 View Post
What exactly would need testing on these closed system routes? The ability for a driverless vehicle to follow a path in a closed (or even limited access) system designed for it has been around for years if not decades in automated warehouses and localized delivery systems. Such systems even include collision avoidance and location broadcasting. What hasn't been fully tested is the ability for driverless vehicles to follow preexisting roads with human driven vehicles, pedestrians, and road hazards and irregularities.
I wanted to come back to this after I was thinking about it while driving home. It did not require a mysterious human factor to discover that a Tesla will drive smack into the side of a white truck with a white sky behind it. If existing closed systems already have working location broadcasting, that's great, but the system needs to be worked out among all of the car manufacturers, and we all have to work out the legal and privacy details, and it needs to be incorporated into cars. It's not like that part is done. And the cars are less safe without it. I don't want to be an alpha tester of whether, under certain conditions, autonomous cars won't see my car.

Tesla is selling a car that it claims you can get into and give a voice command and it will drive you there. And if you don't tell it anything, it will go to the location of the next appointment on your calendar. There is no regulation of this, even though there are regulations about what color the car's blinkers can be. They are selling this for use anywhere, but there is no established intervehicle communication system, which would be an important safety feature even in a completely closed system.

I'm not a Luddite here. I would be pretty happy were driverless cars to happen. I have a long commute, and I would love to be able to do things other than drive during it. But I also, as someone who spends a lot of time driving, am very concerned about driverless cars just showing up on the roads without regulations and without working out any of the crucial details for this step.

And the fact that the autonomous cars, which the engineers are so focused on being able to predict what human drivers will do, have no way of recognizing whether the car in front of it even has a human driver is utterly ridiculous. Particularly when *that* bit of information could be conveyed by something as simple as a barcode on the bumper.

The development of a self driving car that could interact with human driven cars and navigate ordinary roads was a great challenge that has pushed technology forward in lots of important ways. But just because a lot of progress has been made on the problem, that does not mean that it should be the real world approach to future transportation.

Last edited by erwins; 20 September 2017 at 09:08 AM.
Reply With Quote
  #43  
Old 20 September 2017, 11:05 AM
Don Enrico's Avatar
Don Enrico Don Enrico is offline
 
Join Date: 05 October 2004
Location: Hamburg, Germany
Posts: 7,296
Default

I'm still thinking about ways that could make driver-driven and driverless cars on the same road saver for everybody. We obviously need a way at least the driverless cars can communicate among each other. Even better when as many "old school" cars as possible are taking part in the communication as well. But there will alwys be the possibility of an old car without transmitter, a car with a broken transmitter, or an Amish horse-drawn carriage being on the road as well.

What about we have all transmitter-equipped cars transmitting not only their location ("I'm here!") and type ("I'm a 3.5 tons lorry!") and navigation plans ("I'm going to change to the right lane!"), but also what they "see" around them ("There's a car in front of me going 50 mph, and a car without a transmitter to my left"). That way, the transmittig cars would create an ever-changing map of the road they are on, and one car mistakeing a white lorry for the white sky with it's optical sensors would still know that the lorry is there because the other cars are saying so.

Of course, the idea probably falls flat due to the fact that all car manufacturers would have to agree on a transmitting system. And we can't even agree on how to design power outlets...
Reply With Quote
  #44  
Old 20 September 2017, 02:18 PM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

Quote:
Originally Posted by Don Enrico View Post
We obviously need a way at least the driverless cars can communicate among each other.
Yes, this and/or so much more. Many of the developers keep skipping these problems because they think they're easy. "Talk to each other? We have five hundred networked robots and products in a factory all working and moving in perfect harmony; compared to that it's a cinch. What needs testing?" No, actually, it's a completely different problem. As are lidar in weather, road map changes, insurance and liability, traffic laws, engaging and disengaging, telling where to go, pedestrian reactions, other driver anticipations, etc etc. The engineers don't yet realise that these have to be part of the design, not incidental to it or somehow crowbarred in later from existing (ie different) systems. Most of these developers see laws, for example, as just something to leave to others at best and a hindrance at worst. What they should be doing is encouraging laws that make sense for the cars and testing them in as close to possible as the real world so that they can adjust both to work together the way laws and cars do now (even though they're a bit of a kludge because engineers and lawmakers didn't do it right then either).
Reply With Quote
  #45  
Old 20 September 2017, 02:51 PM
GenYus234's Avatar
GenYus234 GenYus234 is offline
 
Join Date: 02 August 2005
Location: Mesa, AZ
Posts: 24,235
Default

Quote:
Originally Posted by erwins View Post
I wanted to come back to this after I was thinking about it while driving home. It did not require a mysterious human factor to discover that a Tesla will drive smack into the side of a white truck with a white sky behind it. If existing closed systems already have working location broadcasting, that's great, but the system needs to be worked out among all of the car manufacturers, and we all have to work out the legal and privacy details, and it needs to be incorporated into cars. It's not like that part is done. And the cars are less safe without it. I don't want to be an alpha tester of whether, under certain conditions, autonomous cars won't see my car.
The Tesla crash is a failure of deployment and human-machine interface. The system was never meant to be a completely self-driving car, but instead a series of assisting technologies. The failure was in not realizing that people would treat it as an autonomous system. The driver of Tesla was repeatedly warned by the car that he needed to drive it, but he ignored at least 8 warnings. Tesla has now updated the software so the car will slow to a stop when the driver seems to be using the system as a self-driving car. Essentially this is a case of the amount of technology falling into a danger valley between minimal technology where the driver pays full attention and maximum technology where the car can truly drive itself.

As Don Enrico points out, car to car communication should definitely be used, but only as additional information. It can't be depended on because of all the possible things that aren't communicating or because of potential human failures in communication. For example, the roll-out program would have to include every trailer, both commercial and private. Otherwise the system would depend on the human driver to alter their transponder to include the fact that their vehicle is now 20 feet longer and segmented. It would also have to work when the vehicle was not powered, such as when an RV user tows their car behind the RV.

ETA: I'm not saying that general deployment on existing roads is the best system, I'm saying that it is the system that is most likely to actually happen. Realpoltech as it were.

Quote:
Originally Posted by ganzfeld View Post
Yes, this and/or so much more. Many of the developers keep skipping these problems because they think they're easy. "Talk to each other? We have five hundred networked robots and products in a factory all working and moving in perfect harmony; compared to that it's a cinch. What needs testing?"
Or they are testing such systems.

Last edited by GenYus234; 20 September 2017 at 03:11 PM.
Reply With Quote
  #46  
Old 20 September 2017, 02:58 PM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

Yes, they have been testing such systems for more than five decades. Wake me when they start to get something that actually works with real people - i.e. real laws, real roads, real business models, etc. (g)You can't just slap those things on because they change the way the hardware has to work! That's why even though we've had robust vehicle-to-vehicle communication for decades we still aren't close to implementing it anywhere in the world.
Reply With Quote
  #47  
Old 20 September 2017, 03:41 PM
erwins's Avatar
erwins erwins is offline
 
Join Date: 04 April 2006
Location: Portland, OR
Posts: 11,356
Default

The part of the Tesla crash that I am talking about was the failure of the car to perceive another vehicle that was directly in its path. There is no reason to think that the car would have done anything differently if it was acting fully autonomously. And yes, it showed that the semi-autonomous systems in that vehicle were not sufficient for autonomous operation. How do we know whether any autonomously operating car on the road now has sufficiently safe systems? We can't, because we haven't even begun to try to define what is sufficiently safe, or to have requirements that the vehicles meet those standards.

I feel a bit like I'm in a human subjects experiment, and I vigorously do not give my consent, and I definitely know I have not been adequately informed.

One function of legislation is that it can serve as a way of gaining either a mutual agreement, or, at least assent that the majority has agreed on something and is strong enough to carry the day. We as a society need to work out these systems and problems, and the vehicle designers need to know the answers to these questions so they can be Incorporated in the designs, *before* the cars are put into widespread, unrestricted use.
Reply With Quote
  #48  
Old 20 September 2017, 03:43 PM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

Oh, and by the way, no that Cadillac system and ones like it in the article Genyus linked don't count as testing the type of system I was talking about. The National Automated Highway System Consortium started in 1994. It took them only three years to demonstrate robust, safe, automated car following. It was twenty years ago last month:
https://www.youtube.com/watch?v=C9G6JRUmg_A
(See especially from 4:20. -- ETA not to mention the prediction at 9:22. He was right - we did have all the hardware - and software - needed to do this in 2007. Turns out that's not the hard part.)
Why wasn't 20 years enough for this existing tech to be used? Luddites? Baloney. It has to work along with the rest of... well, everything and everyone else! That turns out to be a much much harder problem than it seems. You can't just say "well this is the lane where they'll run so go to it".

Last edited by ganzfeld; 20 September 2017 at 03:54 PM.
Reply With Quote
  #49  
Old 20 September 2017, 04:24 PM
GenYus234's Avatar
GenYus234 GenYus234 is offline
 
Join Date: 02 August 2005
Location: Mesa, AZ
Posts: 24,235
Default

Quote:
Originally Posted by erwins View Post
I feel a bit like I'm in a human subjects experiment, and I vigorously do not give my consent, and I definitely know I have not been adequately informed.
If it makes you feel any better, this is far from the only experiment you are unwillingly part of. Building codes, environmental standards, food production, software design, security, politics, the list is nearly endless.
Reply With Quote
  #50  
Old 21 September 2017, 12:54 AM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

Yes, that's exactly why we have a very strict set of rules (in the US, rooted in the constitution) by which (many of) those things are developed. Engineers seem to have a hard time understanding that 1) laws need to be engineered too; it's not a matter of just firing them up 2) regulation doesn't (necessarily) hinder progress of technology because regulation is part of technology and always has been. Zero technology exists without it. The first thing that happened when fire was discovered was people realised it doesn't even come close to doing what it's supposed to do without regulation.
Reply With Quote
  #51  
Old 21 September 2017, 01:40 AM
erwins's Avatar
erwins erwins is offline
 
Join Date: 04 April 2006
Location: Portland, OR
Posts: 11,356
Default

Quote:
Originally Posted by GenYus234 View Post
If it makes you feel any better, this is far from the only experiment you are unwillingly part of. Building codes, environmental standards, food production, software design, security, politics, the list is nearly endless.
Nope, as ganzfeld explained, those are very different. I refer you also to the last paragraph of post 47. Those systems you name, and ones like them, are the counterexample to what is happening with vehicle automation.
Reply With Quote
  #52  
Old 21 September 2017, 02:18 PM
GenYus234's Avatar
GenYus234 GenYus234 is offline
 
Join Date: 02 August 2005
Location: Mesa, AZ
Posts: 24,235
Default

I have never argued for unregulated testing. My argument has been that large-scale closed system testing is impractical and would probably be largely ineffectual at this point and that on-the-road testing is the most likely way that the technology will advance.
Reply With Quote
  #53  
Old 21 September 2017, 06:03 PM
DawnStorm's Avatar
DawnStorm DawnStorm is offline
 
Join Date: 11 March 2003
Location: Montgomery County, MD
Posts: 15,304
Driver

Apropos of nothing, some people in this area have been disguising themselves as car seat/camouflaging themselves and driving around. The drivers' hands are in their laps. The goal is to gauge people's reactions to a 'driverless' car.
Reply With Quote
  #54  
Old 22 September 2017, 04:15 PM
RichardM RichardM is offline
 
Join Date: 27 March 2004
Location: Las Cruces, NM
Posts: 3,882
Default

A difference between the Republican party and the Democratic party is a reliance on civil law suits versus regulation/laws. The Republicans say that civil law suits will force manufacturers to develop safe products. The Democrats say that only regulation will do that.

That said, almost all engineers follow almost all laws in regard to their professional work. To do otherwise would invite both criminal and civil law suits.

As to considering law and the autonomous car, the IEEE has a group considering this. I am sure they are working on every thing they can. But engineers are human. Their managers however, are not.
Reply With Quote
  #55  
Old 22 September 2017, 11:21 PM
erwins's Avatar
erwins erwins is offline
 
Join Date: 04 April 2006
Location: Portland, OR
Posts: 11,356
Default

Yes, those on the right tend to argue that the free market and civil litigation will sort it all out. And then they place caps on damage awards in civil litigation, and press for litigation protection for companies in various ways. I'm more a fan of trying to figure out the basic safety rules ahead of time, and, as I've alluded to, the legislative process serves a function of communal agreement that can have great value.
Reply With Quote
  #56  
Old 23 September 2017, 04:00 AM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

I know about the ieee's thing but, frankly, I don't think anyone has a handle on this, including them (and I think they know that too, unlike the lots of others in that area). I don't think we're going to have any of the kinds of automation these companies are touting for a long time.

"Self driving" is a stupid term. The car goes where it wants to go? No; nobody wants that. Someone still has to tell it lots and lots of things besides just where to go. The only difference between that and just driving is the way those instructions are given. Say exactly what it does, not "autopilot" or "super duper cruise control" or some other stupid term either.

Here's another dumb term: "human error". At best, it's an admission of the system's failure. Say exactly what happened. What can the engineers do to prevent it? The excuse that the driver (or passenger or whatever you call the person in the driver's seat of a car that's supposed to be quasi-self-driving) was warned - once or eight times or a hundred times - is pathetic. We've had systems that warn people of all kinds of things for over a hundred years. We know that they often fail when they are designed the way these "warnings" are. It's just CYA. They don't know how to let the driver know what's happening in anything close to, for example, the tactile feedback one gets with one's hands on the wheel; being able to feel the car taking control or handing it back. That's just one example of hundreds of things that still need to be designed. (See that video I linked to, by the way, of the 1950's vision of this. These postwar engineers had a real grasp of the type of system that works for real.)

I don't think "real world testing" really means much (is this another dumb term? probably) when they don't know what to test; they still don't have a handle on all the things that can go wrong.
Reply With Quote
  #57  
Old 23 September 2017, 06:11 AM
crocoduck_hunter's Avatar
crocoduck_hunter crocoduck_hunter is offline
 
Join Date: 27 May 2009
Location: Roseburg, OR
Posts: 11,220
Default

I thought "real world testing" meant "try operating it in the real world, then try to come up with ways to fix all the things that go wrong."
Reply With Quote
  #58  
Old 23 September 2017, 06:47 AM
ganzfeld's Avatar
ganzfeld ganzfeld is offline
 
Join Date: 05 September 2005
Location: Kyoto, Japan
Posts: 22,945
Default

Well, it's an ideal, not a reality. There's nothing wrong with testing in a lab. That's the real world too. You can't test something in a fantasy world so that's right out. As soon as it becomes a test (rather than a passive observation - which can also interfere, yes, Dr Goodall), it's not that ideal real world. Yes, we try to get as close as possible but it's more important what is being tested rather than sticking to unrealistic ideals.

To "try operating it in the real world, then try to come up with ways to fix all the things that go wrong" is the futzing (aka trial and error) I talked about. Yes, that does work but it takes a lot longer. Again, that's why it took two decades between the first cell phone call and the cell phone proper. Other than the size, there was almost nothing about the 1993 phone that couldn't be done with 1973 technology. It's a lot harder than it seems at first. Yes, real-world testing had to be done but just making a whole bunch of cell phones and seeing how they fail doesn't really move things forward much. (cf Iridium)

It's more about being realistic - being skeptical - than being in the "real world". (That's hard when using investors' money. Which probably, I'm guessing, goes back to what RichardM was saying.)
Reply With Quote
  #59  
Old 23 September 2017, 05:12 PM
RichardM RichardM is offline
 
Join Date: 27 March 2004
Location: Las Cruces, NM
Posts: 3,882
Default

ganzfeld, I have say that it appears you are not familiar with any type of engineering testing. Unfortunately, I'm not sure how to explain it to you.

With autonomous vehicles, not only do the usual things have to be tested, such as the windows, transmission, etc. but the sensors and the software. These are the components that actually can be lab tested the easiest.

The problem is that without complete vehicle to vehicle communication, the autonomous vehicle can not protect against another vehicle operating at an unsafe speed coming through an intersection and hitting the autonomous vehicle if the crossroad is shielded from the sensors because of a building on the corner. Nor can a human operated vehicle stop this sort of crash. You, man or machine, can't avoid what you can't see.

And even with complete inter-vehicle communication, such vehicles can not protect completely against trees falling on them or even people running into traffic. Again, some thing a human operator can not avoid either.

Another type of crash that can not be avoided is the human operated vehicle running into a stationary vehicle. That has happened to me more than once.
Reply With Quote
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is On

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Republican Convention: 20,000 sign petition demanding right to carry guns at event firefighter_raven Soapbox Derby 8 29 March 2016 02:13 PM
Spider on shoulder leads to "driverless" car crash involving school bus wanderwoman Crash and Burn 26 27 September 2015 05:42 PM
CVS changes name, stops tobacco sales early snopes Business Bytes 7 11 September 2014 10:32 PM
Thunder stops eggs from hatching snopes Old Wives' Tales 2 22 July 2014 05:37 PM
Beat police stops with air bag deployment snopes Automobiles 21 24 August 2007 03:23 PM


All times are GMT. The time now is 06:09 PM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2017, vBulletin Solutions, Inc.