Autonomous Cars, Yes or No?

[Paul D]

Originally Posted by Dave Brand

Really? Look up the definition of autonomous.

Autonomous: having the freedom to act independently (Oxford Dictionaries)

Note the word 'freedom' in there? Autonomous doesn't mean a compulsion to act independently, but simply the ability to do so. In the context of computer controlled cars, all that means is a car that is capable of driving itself. It does not mean a totally self-contained machine that is incapable of any external intervention, as you seem to be suggesting.

How do you suppose these cars will receive software updates, report faults, get information on road changes, traffic situations, etc? It can only be done via a capability to download information - and as soon as you give it the capability to download stuff, then potentially you let the hackers in too.

It's already been proven in the US, where a hacker (a young teenager I seem to remember, so imagine what an experienced hacker could achieve) demonstrated he could circumvent the systems on the latest GM vehicles and take partial control, allowing him to start engines, lock and unlock the doors and remotely operate other systems within the car.

[SidewaysFeltham]

Originally Posted by Akrapovic

The idea that all technology deskills is an idea that only works if you ignore the skills required to use new technology. The ability to actually use a computer is far more relevant than hand writing in today's world. This is not deskilling but rather a a transfer of skills to a more relevant area. This is why the idea that modern drivers are less skilled because they do not double d clutch is nonsense. Today's drivers are more skilled at dealing with the over congestion and information overload than older generations.

Thank you for providing links and then posting the information in the thread. Unfortunately there appears to be a bit of a breakdown of understanding on what we're discussing. You appear to be taking my posts as talking about the youngest possible generation of drivers, hence the mentioning of 17 year olds (you'll notice I haven't mentioned age brackets and certainly nothing that young). The original discussion was not about teenagers who have just passed their test, but the difference between the people who passed tests decades and decades ago, and the drivers of today. The 50s and 60s were even mentioned and some wildly inaccurate claims of the test being the same as 50 years ago were banded about.

You won't find me arguing that 17 year olds are better drivers than 50 year olds. I don't think anyone would (or has) made that claim. But I will dispute that a 25 year old is fundamentally a worse driver because he doesn't know how to use an outdated clutch system, despite passing a more stringent test in a more stressful environment. [IMG]file:///C:/DOCUME~1/MICHAE~1/LOCALS~1/Temp/msoclip1/01/clip_image001.gif[/IMG]

Q E D T P and any other letters you'd like. [IMG]file:///C:/DOCUME~1/MICHAE~1/LOCALS~1/Temp/msoclip1/01/clip_image001.gif[/IMG]

On the rest of your examples, I don't really agree. The example of not being able to fly a plane without a computer is a particularly bad one since it's all electronic and doesn't physically work without it. So if the onboard computer goes down, then the best pilot in the world isn't flying your plane. And kids using calculators is a good thing since the maths and especially science they are doing has advanced a lot since we were in school. I also don't agree these examples are relevant to the example of cars and clutches and only serve to driver is further off topic. [IMG]file:///C:/DOCUME~1/MICHAE~1/LOCALS~1/Temp/msoclip1/01/clip_image001.gif[/IMG]

Edit: just a note that I think the assumption about ages is most likely being tied to an assumption on my age. I am not a teenage driver (I wish). I'm not even a 20s driver. I'm well into my 30s. Whilst I would never say I am old and certainly not as experienced on the road as someone in their 60s. So I am not defending younger generations out of pride of my age, because that isn't my generation!

This was your opening statement: post#20

I completely disagree with the idea that a modern vehicle is "de-skilling" current drivers. Given that decades ago, a driving test involved driving up and down a runway and knowing the basic controls, and todays test is FAR more comprehensive, I don't think that's fair.

You then stated:

So I do not agree with the suggestion that we're deskilling drivers. The statistics show the opposite

Continually, you assert, driving standards are far better, thanks to the higher standards of current tests.


Therefore if the foregoing is correct, then how come:


“Some 31,668 males – just over one-third of the total number of people disqualified from driving during this period – were between the ages of 20 and 30.
Males also outnumber females considerably when it comes to carrying out offences leading to a driving ban. In total 13,481 females were disqualified during this time period – less than 15 per cent of the overall total.
And in the 20-30 age group just 4,333 females were disqualified in the 12 month period, compared to 31,668 males.”


As I stated previously, the date of introduction of your whizzy new tests means the age groups who supposedly passed lie between 31 to 37 years old. As age increases, the numbers disqualified rapidly reduces. This data is all in the earlier references I cited.


Now this was amusing!


[quote On the rest of your examples, I don't really agree. The example of not being able to fly a plane without a computer is a particularly bad one since it's all electronic and doesn't physically work without it. So if the onboard computer goes down, then the best pilot in the world isn't flying your plane. [/quote]


You cannot really believe all airplanes now fly themselves, surely?
They can take off, land and make certain autonomous decisions (operating height, speed, heading) mainly on optimal fuel consumption grounds.
If the computer and back-up go down, however, then the pilots take over.
However, the more automatic the process becomes, then obviously, pilots are gradually de-skilled, since such skills are developed over time and honed each and every time they fly.

And kids using calculators is a good thing since the maths and especially science they are doing has advanced a lot since we were in school.

Except in order in order to program a calculator, the user needs a degree of basic maths skills. Even quite simple calculations require basic knowledge.


For example, on a simple calculator, compute the following:


1. In 2014 my company turned over £2,000,000. In 2015, my company turned over only £ 1,500,000. What is the percentage droop between 2014 and 2015? (i.e. how much less is the 2015 turnover expressed in percentage terms of reference).


2. I need to work out my VAT account. Invoices showing the total inclusive of VAT need (i) The actual sales price less VAT; and, (ii) The VAT.
VAT is 20%. Remember, usefully, there are 100 Pence per pound.


Seems the US Navy agrees concerning de-skilling...


“Mr McKinlay concluded: "Schools should teach navigation and map reading as life skills.


"The introduction of computers and calculators has not removed the need to understand numbers. The US Navy has started to teach celestial navigation again as a back-up skill.
"Navigation is where complex systems meet capable users." “


http://www.telegraph.co.uk/news/ukne...erts-warn.html

[Razor]

For long trips? Yep.

For when I have the need or feel to drive? (such as around town, etc) Nope

[Akrapovic]

I'm not making a large post as I'm on my phone, but the dissection of my post was utterly ridiculous and hasn't even been read properly. A perfect example being I'm
Not talking about teenagers and those young troublesome kids, but it keeps getting brought up.

And the pilot and plane example completely ignores what I even said. I never said a plane flies itself. I said a computer provides the link through electronics and without a computer, the connection is no longer there. I never said the computer flies it, just pointed out that the computer transmits the pilots actions. It's one of those lovely well known facts that the Eurofighter cannot actually fly without the computer and would just fall out of the sky.

But as long as we're not reading posts properly and just using condensending nonsense about all those "whizzy" tests (which people who are far mor competent than us have judged a good thing, and those horrible young drivers keep passing), then the debate is absolutely pointless. Shame, as until then it was a great debate with Mike (as always - he's brilliant to debate with. Respect!) buts it now lost its way. So as we don't continue to derail this I shall bow out before its out of control.

[fourWheelDrift]

Originally Posted by Akrapovic

And the pilot and plane example completely ignores what I even said. I never said a plane flies itself. I said a computer provides the link through electronics and without a computer, the connection is no longer there. I never said the computer flies it, just pointed out that the computer transmits the pilots actions. It's one of those lovely well known facts that the Eurofighter cannot actually fly without the computer and would just fall out of the sky.

I'm sure one or two on here are "real" pilots and could add more; however as a matter of interest I did a one hour "flight" on a modern 747 simulator with an instructor who was a recently retired 30 year Virgin Atlantic pilot. It's hard to overstate how interesting this was and I'd recommend it to anyone but one thing I did learn is that while the crew manage the systems, deal with emergencies and spent a lot of time talking to ATC they do precious little flying and indeed could do even less, essentially once the wheels leave the ground they can just tell the autopilot how high, what speed, how fast to ascend/descend, what heading to fly etc. until the aircraft is rolling gently down the runway and it's time to turn off onto the taxiway to the terminal.

Of course they do have to do a small amount of manual flying occasionally esp. landings to stay current and therefore qualified.

I don't think this has much to do with driving a car though, the key difference being that aircraft are not generally in close proximity, you can't jink three feet to the right or fumble when reaching for the brake pedal and cause an instant accident.

[SidewaysFeltham]

The example of not being able to fly a plane without a computer is a particularly bad one since it's all electronic and doesn't physically work without it.

Nonsense!

The pilot/s are able to fly any airplane even when and where the built-in autonomous systems fail. It is much harder, however.

Reference de-skilling:

Please actually read.

http://www.theatlantic.com/magazine/...etting/309516/

http://www.wsj.com/articles/SB100014...04202526288042

http://www.telegraph.co.uk/news/ukne...emergency.html

http://flightsafety.org/aerosafety-w...vel-automation

[Mike Harte]

Having moved the debate to aircraft, there are a few misconceptions. For example, the Typhoon Eurofighter is only really able to be flown straight and level without it's computer systems; the aircraft is notionally not aerodynamic, and it requires almost constant positioning refinements that the computer system provides.

You need to be careful when you say that any pilot can fly any plane, especially if you then add that they can do so when the automated system has failed. I have a lapsed pilot's licence, and even when it was current, I wouldn't have had a clue how to keep a Boeing 727 in the air, let alone a 747. Although commercial pilots now do virtually all their training in simulators, even in days gone by a 2 engine pilot would sit in the 2nd seat when upgrading to a 4 engine plane until they had shown sufficient skill in flying the plane. Even nowadays, an experienced captain moving from flying a Boeing 737 to an Airbus 320 will be required to undertake many hours in the simulator before he would be allowed in the cockpit, and even then he would need to fly in the right hand seat for quite a few flights before moving across the console.

And a word of caution about the interaction between on-board flight computers and pilots. Many of the recent aircraft losses have been caused by pilots trying to ignore what the instruments are telling them, and over-riding the autopilot. Virtually every fatal accident in recent years has been due mainly through human error, and my own opinion, for what little it's worth, is that pilots have been deskilled. They have become so reliant on instruments and computing power that they have lost some of the ability to be able to rely on their own ability. Although simulators are wonderful things, a trainee pilot can walk out of every crash landing that he makes; if he was in an actual plane, he wouldn't have that luxury.

Finally, thank goodness you say, certain aircraft at certain airfields have the ability to taxi to the runway, take off, climb, level off, descend, land and taxi to it's parking slot without any input from the pilot except for a few pushes on computer-like buttons. That is the theory; in practice, those airfields, known as Category IIIc, are awaiting certification. Most major airports will be Category IIIb which permits aircraft to take off and land in very poor conditions, and some aircraft and pilots are certified to land and permit the autopilot to take the aircraft off the runway onto the designated taxiway, at which point the pilot has to take manual control. But that will change shortly, I am sure; after all, BA has been able to use the current system for decades, even if other airlines are only now catching up.

[chillibowl]

a couple of new developments:

a fleet of about a dozen automated trucks just completed a run across Europe.

http://qz.com/656104/a-fleet-of-truc...across-europe/

there will be more and more of these exercises coming up soon but one thing i found interesting was this:

'But the convoys must first successfully navigate Europe’s bureaucracies. Bruijn’s next goal is to get everyone to sign off on a roadmap for the next five years. This would address the technical problems, such as the inability for trucks from different brands to platoon together (they all use different wifi systems).'

looks like competition and desire for each manufacturer to develop their own system will cause delays (more negotiation) to allow trucks from different makers to communicate with one another so effectively their ability to coexist on the same roadways together...i would assume this will also be an issue for their road car manu counterparts.

also a scary thought for those worried about employment...trucks that drive themselves and pick up cargo from automated shipping ports...

second article about some experts concerns about self driving cars.

http://hosted.ap.org/dynamic/stories...04-08-15-37-19

several of the issues have already been brought up in this thread but one which hasent is '-Self-driving cars can't take directions from a policeman.'

while this might not seem like a big issue i think a similar problem would exist for road workers or anytime a person is out there redirecting traffic.

its one thing to expect manus to develop software that can understand signs and road markings but to expect them to develop software that deals with understanding human behaviour and actions is entirely a bigger problem to solve and probably one outside their expertise.

[wnut]

Can't wait for this to be 100% self drive:

http://www.abc.net.au/news/2016-04-2...-fault/7348812

[wnut]

Had not realised they had 1.5M km on Google self drive cars.

Still, first crash. Looks pretty basic.

https://www.youtube.com/watch?v=ZFndKRwZ6MI

https://www.youtube.com/watch?v=I9T6LkNm-5w

[chillibowl]

so these google cars....are the cameras on top just for diagnostic purposes or do they form part of the self driving technology?

[Tim Falce]

Originally Posted by wnut

Had not realised they had 1.5M km on Google self drive cars.

Still, first crash. Looks pretty basic.

https://www.youtube.com/watch?v=ZFndKRwZ6MI

https://www.youtube.com/watch?v=I9T6LkNm-5w

It must have had the Asian or African cab driver program running, pull out then indicate, expect everyone to stop for you. Bus driver was trained in London, don't stop or give way to anything.
Nothing new then.

I how you exchange insurance details with a driverless car when you have a coming together, do you stick a note under the wipers?

[wnut]

Tesla autonomous car fatality.

https://www.teslamotors.com/blog/tragic-loss

[wnut]

Driver reportedly watching Harry Potter when he was killed.

http://www.news.com.au/technology/in...a5eaf25436b6a5

[SidewaysFeltham]

I picked that up, too, WNUT.

http://www.telegraph.co.uk/news/2016...ile-in-autopi/

Seems those of us with serious reservations were not too far wrong...

[Paul D]

Just what I alluded to in an earlier post - the technology, in my opinion, just isn't good enough yet. I see that the spokesperson from Tesla attempts to mitigate the situation by saying that neither the car nor the driver saw the white truck.

Well, we know for sure the car didn't see it, but how exactly do they know the driver didn't see it? He's dead, so we can't ask him can we? It's entirely possible he either saw it too late to act, because he wasn't paying full attention to the road as he believed all the hype that these cars are incapable of error, or, he simply never saw it at all, for the same reason. Had he been actually driving the car, I'm sure he'd have seen it.

I find it hard to believe that a human driver (at any level above moron) would not see a white semi pulling across their path. That's not necessarily saying the outcome would ultimately be any different, but at least a driver would brake/swerve to attempt to minimise the impact, and not simply carry on into it at full speed as it appears the Tesla did.

I saw some figures somewhere that said this was the first fatality in something like 120 million miles of autonomous driving, compared to a fatality every 90 million or so (in the USA) for humans driving. Not a fantastic improvement really is it? Especially not when you consider that the figure for humans takes account of every crap driver we all see out there on a daily basis - if you're actually a decent, skilled driver, then you can reasonably expect to beat the figure for humans by a considerable factor. I'd guess the figure for deaths amongst highly skilled drivers is somewhere between 500 million and a billion miles per death. Only a guess, of course, and I doubt such a statistic exists, but we can all agree that the figure amongst highly skilled drivers is going to be much higher than the figure for all drivers, and the figure for autonomous.

[chillibowl]

for sure there will be more of these accidents as the tech evolves.

that said though, as there are more cars on the road with this tech i would expect there to be an decrease in accidents and mortality rates (perhaps even a an exponential decrease)...so the question for me isnt 'why did the tesla driver fail to see the other vehicle?' rather it is 'would this accident/fatality have been avoidable if the other vehicle (or both vehicles) had an autopilot system and/or an ability for both cars to communicate with each other?'

most definitely agree though that these systems have a long way to go and part of me questions why so much of this development is happening on public roadways with each company working on their own parallel and possibly non complimentary systems. there is too much of a rush to get a finished product to market here and that is in itself a very dangerous thing.

[fourWheelDrift]

To represent this as a self driving or autonomous car fatality is totally inaccurate. According to news reports the car had a beta version of software designed to work in a motorway type environment to allow the car to adjust it's speed and change lanes, this is a much more limited amount of "autonomy" than a self driving car has. The software in question was a beta test version and only used by those who opted to do so and was supposed to be supervised by the driver with his hands on the wheel and his eyes on the road at all times.

Of course sooner or later there will be a fatality involving a self driving car, that is not really the issue, the question is whether they turn out to be safer than the human driven variety, at the current state of the art maybe maybe not but in time they certainly will be.

[chillibowl]

Originally Posted by fourWheelDrift

To represent this as a self driving or autonomous car fatality is totally inaccurate....

lol true true and good point.

im reminded of stories back when GPS systems first became available and people blindly following their GPS screen into obviously visible disasters.

no doubt, in the short term there will be and uptick in accidents as drivers wrap their heads around the fact that adaptive cruise control, lane departure, and brake assist are not synonyms for 'self-driving car'.

[Akrapovic]

Originally Posted by fourWheelDrift

To represent this as a self driving or autonomous car fatality is totally inaccurate. According to news reports the car had a beta version of software designed to work in a motorway type environment to allow the car to adjust it's speed and change lanes, this is a much more limited amount of "autonomy" than a self driving car has. The software in question was a beta test version and only used by those who opted to do so and was supposed to be supervised by the driver with his hands on the wheel and his eyes on the road at all times.

Of course sooner or later there will be a fatality involving a self driving car, that is not really the issue, the question is whether they turn out to be safer than the human driven variety, at the current state of the art maybe maybe not but in time they certainly will be.

To add to this, it has been said that the amount of miles covered by Teslas self-driving system is over double what humans cover between accidents.

[wnut]

Originally Posted by Akrapovic

To add to this, it has been said that the amount of miles covered by Teslas self-driving system is over double what humans cover between accidents.

Yup, and 77.8625% of all statistics are made up!

[275 GTB-4]

Originally Posted by wnut

Yup, and 77.8625% of all statistics are made up!

Lies, damn lies...and Statistics!

[mattcanary]

Originally Posted by Akrapovic

I won't dissect your post as I feel that takes us far too far off topic and tbh is a little disrespectful (I feel when people do that, they do it purely to create arguments), isn't a fair thing to do, but I disagree with basically the entire thing, especially -



This just isn't true, I'm sorry. I work with driving instructors and examiners who have said it's barely recognizable what drivers go through now compared to what they did before. Mentioning what is the same ignores what is not. The theory test did not even existed until the mid 90s and has massively evolved into an entirely different animal, and then the introduction of the Hazard Perception is a completely new concept, nevermind being principally the same as something 5 decades ago. The Independent Driving section was only introduced in the last few years as well. Despite these additional elements being introduced, the theory test being introduced, the highway code evolving massively, and roads becoming more congested and complex in layout, the pass mark has remained the same - 45-50% (varying more on location than anything else). So todays drivers are coping with more advanced test with more stressful environments, with the same pass rate. This does not coincide with a lack of driving skills.

Giving examples like "most of todays drivers who commit X offense" is meaningless, since they'd have failed their test 50 years ago too according to the description you've given, and yet here we are with bad drivers of all ages. How can this possibly be? And since new drivers are only allowed to total 6 penalty points in their first 2 years, they already have less of a buffer to work with than the more experienced drivers.

I have provided many a statistic to show that driving has not gotten worse. You have dismissed these as baseless and without value, but replaced it with anecdotal evidence. I don't see how this has any more value than the facts from government studies, which show accidents declining despite the massive increase of cars. Sorry if you disagree, but then we'll have to agree to disagree on that point.

-----------------------

On the topic of driverless cars - currently there's a problem where the driver is still legally in charge of the vehicle, even with the computer driving. This sounds fine before you consider what would happen in an accident really, as you'd have to take control of the vehicle extremely quickly.

How does this link to the self parking cars? Who is responsible should you bump something using those?

I can see this working if everything was self driving. But with a bunch of humans in the equation, you'll never be able to tell what the person is actually going to do. And what happens when they come across a police chase? I know it sounds mental, but that's the sort of thing a human reacts to. You could make it react accordingly to blue flashing lights, but a car trying to escape the police...? Hmmmm.

There is a place for these cars in this world, without a doubt. But I do enjoy my daily commute, and it'll be sad if that goes.

Cars are physically far, far easier to drive nowadays than in the sixties though.
Also, how strictly are people taking driving tests marked now, compared to the sixties? Are they allowed more leeway? Maybe they're not?
There are more elements in the driving-test nowadays. However, that does not necessarily mean that the driving test is harder to pass nowadays.

[mattcanary]

Originally Posted by Akrapovic

All of this anecdotal evidence is lovely but it just isn't supported by the statistics. You could argue we're less skilled, or you could point out that people don't need to know how to double d clutch since it isn't a relevant skill. What IS a relevant skill however is the ability to know the computer on your car rather well, which is something that many many of my elderly friends struggle with. So what exactly is the point in being skilled in things which aren't relevant?

Once again, I'll point out that the current driving test contains far more content than it ever has done, as well as introducing a theory test, hazard perception test and takes place on far more crowded and stressful roads than in the past. Yet the pass mark remains the same. Those terrible deskilled drivers of today are taking in more information and ticking more boxes than previous generations, and having the same success rate. I know that might hurt some of the older generations pride, but that's just the facts. Sorry guys!

Ticking boxes is certainly more prevalent in the year 2018. Due to an infatuations with stats and numbers, and little thought about context.

[mattcanary]

Originally Posted by Akrapovic

So people are only passing exams because of a prescribed pass rate that just be achieved? We give people a chance to prove they're good and then we claim the test is rigged when they do? That's some mind bending gymnastics you have to do to justify that. And scarily similar to what we say about school exam results today too. We know for a fact that there is more stuff in the exams and the results are better than ever. Must be proof the exams are easy right? We just love to tell younger generations just how rubbish they are in this country. I think nostalgia is like heroin for older generations.

You don't need my word that the driving test is more advanced than it used to be. We know there's more in it, because there is. Theory Test, Hazard Perception, Independant Driving and soon we'll have GPS Following. These didn't exist before. More boxes have to be ticked to pass now. That isn't my opinion. The tests are also done on more busy roads than in the past. And as the test results from around the nation show, it's harder to pass in busy areas. The Scottish Highlands has a higher pass rate than central London.

And thirdly, if you're driving fast enough in your driving test that the road holding ability of the vehicle comes into the equation, you've already failed your driving test. You shouldn't be answer near the limit of the vehicle.

The speed limit is low compared to the capability of most cars nowadays, compared to the fifties and sixties.
Old cars are harder to drive, and especially so at higher speeds.
Additionally, speed limits are often lower nowadays!, adding to the ease of driving.

I would add though that I make more mistakes when I drive slowly, than when I drive at a reasonable speed, (as I lose concentration and become complacent).