Jump to content


Photo
- - - - -

driverless cars to hit california ( sorry!)


  • Please log in to reply
407 replies to this topic

#101 Kelpiecross

Kelpiecross
  • Member

  • 832 posts
  • Joined: October 10

Posted 08 October 2012 - 11:01

Program that lot.


Fair enough comments. And certainly at present (or in the next couple of years) it doesn't seem likely that a computer could match a human in situations of the types you mention. But I think the whole point of this debate comes down to - just how "clever" and "self-aware" are computers going to become in the next ten or twenty years? And this is the critical question.

Computers (and electronics in general) are evolving and improving at an unbelievable (and almost frightening) rate. We have gone fron a few transistors in a radio to billions in a computer in fifty or so years - even a tiny mobile phone has half a billion (or whatever - a bloody lot anyhow). Nobody can predict the future - this fact seems to be demonstrated continually these days. But surely the progression in computers would indicate that in ten or twenty years computers will be able to match humans in intuitive, lateral etc. etc. thinking - all the types of novel thinking that humans are good at. As well as this (when it comes to driving cars) a computer will always think logically and not be distracted. For instance it will probably never think to itself "You bastard - I 'll get you for that" - or think to itself at the traffic lights "I'll show this bastard how a proper car accelerates". Of course if computers do become "self-aware" its mind may wander onto thinking pervy thoughts about girl computers.

Or again, maybe computer science will virtually stall and computers never become noticeably more clever or "self-aware" than they are now.

Edited by Kelpiecross, 08 October 2012 - 11:05.


Advertisement

#102 MatsNorway

MatsNorway
  • Member

  • 2,033 posts
  • Joined: December 09

Posted 08 October 2012 - 11:34

I think you fundamentally misunderstand how computers can be used. The computer need not necessarily expect the pedestrian to behave in any given way, in fact it probably doesn't even need to determine that it's a pedestrian.


I think that to compete with a human paying attention in a city/urban area it will be hugely difficult for the computers to stay competitive. fully self going cars (goes anywhere) will not happen in in my life time.

As said it will be a evolution.

Volvo is going towards active braking when closing speeds are to high.
They are also going towards slipstreaming trains as mentioned above. Currently its only a idea.


One of the weaknesses is driver error from the front car.


next up (guessing) is active throttle to avoid/limit being rear ended by other cars. This should be easy as its only affecting forwards direction. wich should be well covered by sensors.
other deviced could be systems that flashes bright light to alarm the older car that its closing speed is considered accident likely. a bit OT.




This is partly an excuse to post some great Argentinian saloon car racing video ( thanks repco von brabham) but as Google is mega, mega rich how about it fits its driverless gear onto each of these cars and shows a similar race can be done by computers.

http://www.youtube.c...feature=related

The question of course is how will one car beat another one?? Well, fewer kids are keen on cars it seems but let teams of digigeek teens loose on the software to "tune " it to win but not crash and you may have a whole new generation of race competitors - and no worried about injuries parents.


I like these ideas. Driver with TC is better than just a driver. but a fully independend car is not neccearly going to be competitive.

Biggest challenge would be to get the race organisers to allow it to enter.




#103 Magoo

Magoo
  • Member

  • 2,435 posts
  • Joined: October 10

Posted 08 October 2012 - 12:53

Before we see fully autonomous cars, we will see more driver assistance -- and even driver intervention -- systems on human-operated cars than folks might be expecting. That was the other upshot of the story I posted earlier. Adaptive cruise control, lane departure warning and lane-keeping aids, blind spot indicators, and radar and camera-based collision warning systems, etc and so forth. And that's just the stuff available now. The warning systems of today are the intervention systems of tomorrow.




New study proves: your driving stinks | Mac's Motor City Garage





.


#104 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 08 October 2012 - 13:31

I have traction control on my Mazda - In the recent snow I got to the LH turn down my road and turned onto the slight upslope at which point the car gave up. A friend on the pavement was amused as I got out of the car and stood beside it as the wheels went round; I got back in, turned the off TC, rolled back a bit and drove up the street - Man beats computer.

Computers have a long way to go before they can cope with every situation...

Edited by Bloggsworth, 08 October 2012 - 13:32.


#105 Ross Stonefeld

Ross Stonefeld
  • Member

  • 56,898 posts
  • Joined: August 99

Posted 08 October 2012 - 13:39

I bet you do long division too.

#106 Tenmantaylor

Tenmantaylor
  • Member

  • 8,236 posts
  • Joined: July 01

Posted 08 October 2012 - 14:41

You're right magoo. Ford already have automatic stop features in their production cars: http://www.euroncap....veCityStop.aspx



#107 GrpB

GrpB
  • Member

  • 118 posts
  • Joined: February 09

Posted 08 October 2012 - 16:56

Before we see fully autonomous cars, we will see more driver assistance -- and even driver intervention -- systems on human-operated cars than folks might be expecting. That was the other upshot of the story I posted earlier. Adaptive cruise control, lane departure warning and lane-keeping aids, blind spot indicators, and radar and camera-based collision warning systems, etc and so forth. And that's just the stuff available now. The warning systems of today are the intervention systems of tomorrow.

.


The difference between fully autonomous and these driver aids is one of ideology, not technology. The problem with autonomous is that in the automotive landscape where they must function almost all vehicles can go 120-130 mph, accelerate, brake and corner hard, and are piloted by people that are often malicious or incompetent and often both.
Autonomous is easy when the overall system is controlled from the outset, as in a plant environment where autonomous tugs have been common for a long time. The problem with autonomous on public highways is that the basic/critical system parameters are uncontrolled, it pits silicon-based logic against humans who often act without logic. For example, the goal of the autonomous car merging on the highway to maintain some constant speed may be straighforward, the goal of the guy going WOT, 15 mph over the limit then slowing down to 5 mph under the limit, inexplicably, to prevent that same car from merging, is a common occurrence with infinite variation, is it realistic to expect an autonomous car’s logic to deal with this and countless other illogical things that people do?

Speed differential and intentional speed change is the primary cause of all the annoying, niggling things that people on highways do, speeding up to pass then slow down, slowing down then speeding up to keep people from passing, etc and which often blossom into full fledged road rage. If on public highways (no stop lights, no stop signs, traffic all flowing in one direction), cars were electronically limited to, say 75 mph by external communication with roadside speed limit transponders (replacing speed cameras, since people wouldn’t be able to speed anymore), then realistically all cars would end up in constant file at constant speed. From a practical/safety perspective, that would deal with most of the ‘issues’ that people have while driving without any automation required.

If that were step one, then autonomous on the highway would be easy and straightforward to implement, because people would be used to traveling at constant speed, with everyone at the same speed, with almost no driver inputs between entrance and exit of that highway. But step one would be unrealistic at best, because allowing some roadside box to limit speed is something that no consumer would ever ask for, and which no politician would push as a government mandate. In the US at least, we are quite happy and proud of the industry whose product is the number one cause of teen fatalities. To question the basic nature and use of public road infrastructure, as it relates to autonomous vehicles, is not realistic.




#108 munks

munks
  • Member

  • 343 posts
  • Joined: January 03

Posted 08 October 2012 - 19:05

I have traction control on my Mazda - In the recent snow I got to the LH turn down my road and turned onto the slight upslope at which point the car gave up. A friend on the pavement was amused as I got out of the car and stood beside it as the wheels went round; I got back in, turned the off TC, rolled back a bit and drove up the street - Man beats computer.


Wow, again you are the first one to think of this "deficiency" of traction control. Oops, strike that, it turns out there's a TC disable button in every car for a reason ...

#109 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 09 October 2012 - 08:00

Wow, again you are the first one to think of this "deficiency" of traction control. Oops, strike that, it turns out there's a TC disable button in every car for a reason ...


I thought it was there as an "Enable fun" button. Posters here keep telling us that computers are better than humans at controlling things; I was merely illustrating that at one simple thing they aren't, moving off from rest in difficult conditions they are useless; so why should we expect them to be so much better at much more complicated things - Or are we going to have to accept that when it is snowing or icy we will not be able to ask our autonomous car to take us anywhere? I think this is a very important question. You try to insult me in your post, but in your smugness you have totally missed the point, if a computer can't effectively move a car forward in snow who in the norther states of the US, Canada, or Europe is going to buy one, as it would be useless as a vehicle for transporting you or anybody else from A to B in the winter - Imagine if Sony sold you a television which you couldn't turn on if it was raining, and when you complained told you to suck it up because one day computers will rule the world?

#110 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 09 October 2012 - 09:15

And another thing. Your attempt at sarcasm hit on a significant truth, that's what the off switch is for; so, at what point, in icy conditions, part-way down a 1 in 12 hill with a blind bend at the bottom, will the autonomous car compute that it can't handle the conditions and cede control to the passengers? Half way through a hand of bridge, as the "driver" is pouring out a cup of coffee for his lady wife?

#111 MatsNorway

MatsNorway
  • Member

  • 2,033 posts
  • Joined: December 09

Posted 09 October 2012 - 15:30

As mentioned by bloggsworth one of the significant challenges would be to make it work in snow.

Now im sure the TC bloggsworth had was not the best there is. But the main issue with snow and ice is the sensors getting covered.

Edited by MatsNorway, 09 October 2012 - 15:30.


#112 Lee Nicolle

Lee Nicolle
  • Member

  • 5,661 posts
  • Joined: July 08

Posted 09 October 2012 - 22:47

I know the auto engineeers on here will sqawk but modern driver aids? Traction control? Only good in average conditions. ABS brakes just lengthen braking distances for an experienced driver,, and is often known to fail. Especially the 'hi performance ' versions!! Stability control to me is bloody awful as the car does dumb things. Especially when towing. This in normal highway or hills driving situations. Might be good for very bad drivers but a remotely competent one will fight it. And some cannot be turned off easily either.
So driverless cars,, PLEASE.
There is some fair hoo ha about driverless ore trains in WA at the moment. And the consensus seems to be that they will not be allowed, or at least without a driver!!
Driver aids make drivers lazy and more likely to crash. The number of cruise control crashes a year is staggering, the car is driving you so the driver nods off and falls off a straight road!! In this country hundreds of times a year, I have nearly done it myself. Yes I use it too, it does take a load off of my back, but having my foot a foot from the brake pedal is inherently dumb and in Roo country I do not use it. Skippy is an unpredictable beast!!

#113 Greg Locock

Greg Locock
  • Member

  • 4,477 posts
  • Joined: March 03

Posted 09 October 2012 - 23:11

I know the auto engineeers on here will sqawk but modern driver aids? Traction control? Only good in average conditions. ABS brakes just lengthen braking distances for an experienced driver,, and is often known to fail. Especially the 'hi performance ' versions!! Stability control to me is bloody awful as the car does dumb things. Especially when towing. This in normal highway or hills driving situations. Might be good for very bad drivers but a remotely competent one will fight it. And some cannot be turned off easily either.


No squawking. The objective is to raise the standard of the bad to the average(or better). Have you ever learned how to drive with ABS?

#114 Dmitriy_Guller

Dmitriy_Guller
  • Member

  • 4,024 posts
  • Joined: July 01

Posted 09 October 2012 - 23:37

I know the auto engineeers on here will sqawk but modern driver aids? Traction control? Only good in average conditions. ABS brakes just lengthen braking distances for an experienced driver,, and is often known to fail. Especially the 'hi performance ' versions!! Stability control to me is bloody awful as the car does dumb things. Especially when towing. This in normal highway or hills driving situations. Might be good for very bad drivers but a remotely competent one will fight it. And some cannot be turned off easily either.
So driverless cars,, PLEASE.
There is some fair hoo ha about driverless ore trains in WA at the moment. And the consensus seems to be that they will not be allowed, or at least without a driver!!
Driver aids make drivers lazy and more likely to crash. The number of cruise control crashes a year is staggering, the car is driving you so the driver nods off and falls off a straight road!! In this country hundreds of times a year, I have nearly done it myself. Yes I use it too, it does take a load off of my back, but having my foot a foot from the brake pedal is inherently dumb and in Roo country I do not use it. Skippy is an unpredictable beast!!

Driving isn't racing, competent is good enough. The safety improvement from turning a good driver into a great driver is marginal. The safety improvement from turning a bad driver into a mediocre driver is much greater. Driving isn't racing, just not ****ing is pretty damn good.

#115 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 10 October 2012 - 07:56

No squawking. The objective is to raise the standard of the bad to the average(or better). Have you ever learned how to drive with ABS?


You mean there are drivers who can and drivers who can't - That's a new one on me. It should only come into use in extremis, if you call it into operation all the time you are an idiot.

#116 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 10 October 2012 - 07:58

The biggest improvement in the standard of driving amongst the young would come if they had to do compulsory time in an emergency/casualty unit near a major accident spot.

#117 carlt

carlt
  • Member

  • 1,043 posts
  • Joined: June 09

Posted 10 October 2012 - 08:32

The biggest improvement in the standard of driving amongst the young would come if they had to do compulsory time in an emergency/casualty unit near a major accident spot.


nah - only allow them to drive electric cars

according to most people they only go at walking pace and then only for 2 miles
- so safety sorted -

only old codgers and Republicans should be allowed to drive IC powered cars - with a legally mounted automatic assault rifle -

then we could all rest easy and feel totally safe

#118 munks

munks
  • Member

  • 343 posts
  • Joined: January 03

Posted 10 October 2012 - 14:54

only old codgers and Republicans


What's the difference?

#119 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 10 October 2012 - 15:38

What's the difference?


Ask Fidel Castro, he's an old codger, but definitely not a republican...

Edited by Bloggsworth, 10 October 2012 - 15:40.


Advertisement

#120 BRG

BRG
  • Member

  • 11,446 posts
  • Joined: September 99

Posted 10 October 2012 - 19:16

ABS brakes just lengthen braking distances for an experienced driver,, and is often known to fail. Especially the 'hi performance ' versions!!

You clearly have very poor ABS on your car. And only the very very best drivers can beat a decent ABS. I certainly can't and I doubt if you can either. In my view ABS should be mandatory on all new cars (which it pretty much is nowadays) and is the most effective safety aid since seat belts. My ABS cuts in maybe once in 2 months or so, but at least once it has saved me from rear-ending the guy in front when my attention wandered for a moment. Which leads me to :-

Driver aids make drivers lazy and more likely to crash. The number of cruise control crashes a year is staggering, the car is driving you so the driver nods off and falls off a straight road!! In this country hundreds of times a year, I have nearly done it myself. Yes I use it too, it does take a load off of my back, but having my foot a foot from the brake pedal is inherently dumb and in Roo country I do not use it. Skippy is an unpredictable beast!!

A very articulate argument in favour of driverless cars where if the 'driver' nods off, it won't matter!

#121 gruntguru

gruntguru
  • Member

  • 5,175 posts
  • Joined: January 09

Posted 10 October 2012 - 23:36

Computers may well have a different approach, but different humans have different approaches at different times and in different situations. Trains run on well defined tracks, no possibility that they can take a different route, yet irrational people still manage to wander into their path; you cannot consider the apparent efficiency of the computer in isolation, it is part of a world full of irrational people, semi-rational animals and falling objects with no pretence of rationality at all...

It is nonsense to suggest that another irrational (or even a rational) human has any better chance than a computer - of predicting the behaviour of an irrational human.

#122 mariner

mariner
  • Member

  • 1,359 posts
  • Joined: January 07

Posted 11 October 2012 - 06:46


I have been assuming that a big benefit of driverless cars is improved traffic flow but I'm wondering now if that is only true in certain conditions.

I would imagine that in US style gridiron road networks with only traffic lights driverless would boost flow and speeds. However on roads like in the UK I'm now not so sure.

Yesterday I went round a local roundabout/traffic circle at Welwyn Garden City ( Tony Matthews will know the one). It is just off an expressway and is about 80 metres across. There are five entry/exit roads and an entry to a huge supermarket squeezed into the space. Each approach rod has three lanes and there are six sets of traffic lights.

It only works becuase people change lanes constantly and even cut across multiple lanes. If driverless cars are set to avoid collisions at all costs ( and hopefully they would be ) then I can't see them coping with such complex traffic interweaving very well. Some degree of judgement of the risks being taken by other drivers is needed to keep traffic flowing on that roundabout. This involves non linear questiosn like " is the car driven by an old lady or is it a 10 year old BMW M3 with a dodgy looking driver?"

People use such judgements to take calculated risks in complex driving situations. They do get it wrong sometimes but it can help traffic flows

How do you programme a driverless car to take risk? - can you morally do so and if you did who gets sued after the accident?

#123 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 08:21

The "Name that tune" argument is specious. If thousands of tunes start with the same two notes and the computer selects a different answer to a human, then it has chosen a correct but different answer. if you mean that a human being can recognise a likely candidate from, for example, the very first riff of "Have You Seen Your Mother", yes I agree that computers will be unlikely to manage that.


You have unwittingly, as far as you are concerned, hit the nail on the head. Choosing from unlikely scenarios is precisely what a driver has to do in an emergency, make a best estimate and act on it; computers are seriously crap at estimates, they only do 0 & 1, not much room for doubt there.

#124 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 08:29

All these examples are extreme rather than commonplace. 90% of people won't notice these things. The computer is to help the idiots not the clairvoyants like yourself.


It is in the extreme that we are talking about and the extreme is where accidents happen. The Air France crash off Brasil happened because the computer got lost because a pitot tube iced up, the plane kept climbing, lost it and ceded control to the pilots who had not been trained to deal with such a situation "As it was not necessary, the computer has complete control..." No-one is saying that in perfect conditions when everyone is behaving themselves the autonomous car won't work, the argument is about what happens when things aren't what the computer expect to have to deal with, and will it be able to "think" its way out of the situation.

#125 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 08:32

It is nonsense to suggest that another irrational (or even a rational) human has any better chance than a computer - of predicting the behaviour of an irrational human.


A rational person will expect an irrational person to behave irrationally and an irrational person sees a rational person as irrational so will expect an irrational person to behave in the same way as they do...

#126 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 08:37

I have been assuming that a big benefit of driverless cars is improved traffic flow but I'm wondering now if that is only true in certain conditions.

I would imagine that in US style gridiron road networks with only traffic lights driverless would boost flow and speeds. However on roads like in the UK I'm now not so sure.

Yesterday I went round a local roundabout/traffic circle at Welwyn Garden City ( Tony Matthews will know the one). It is just off an expressway and is about 80 metres across. There are five entry/exit roads and an entry to a huge supermarket squeezed into the space. Each approach rod has three lanes and there are six sets of traffic lights.

It only works becuase people change lanes constantly and even cut across multiple lanes. If driverless cars are set to avoid collisions at all costs ( and hopefully they would be ) then I can't see them coping with such complex traffic interweaving very well. Some degree of judgement of the risks being taken by other drivers is needed to keep traffic flowing on that roundabout. This involves non linear questiosn like " is the car driven by an old lady or is it a 10 year old BMW M3 with a dodgy looking driver?"

People use such judgements to take calculated risks in complex driving situations. They do get it wrong sometimes but it can help traffic flows

How do you programme a driverless car to take risk? - can you morally do so and if you did who gets sued after the accident?


I know that roundabout well, I rounded it every morning on the way to work for 7 years. The first few it had no traffic lights and I never saw an accident, in the first 2 weeks after the installation of the lights there were about a dozen accidents - They should have left it as it was, having 7 lights for 5 roads was very confusing and somewhat defeated the point of having a roundabout in the first place.

You weren't on your way to the old Lotus workshop at Panshanger were you?

#127 Tony Matthews

Tony Matthews
  • Member

  • 17,498 posts
  • Joined: September 08

Posted 11 October 2012 - 11:42

You weren't on your way to the old Lotus workshop at Panshanger were you?

I used to go there a couple of times a week, but didn't use that roundabout, I went under the Welwyn viaduct from the A1. My favourite RaB is the Magic at Hemel Hempstead.

#128 gruntguru

gruntguru
  • Member

  • 5,175 posts
  • Joined: January 09

Posted 11 October 2012 - 12:23

A rational person will expect an irrational person to behave irrationally and an irrational person sees a rational person as irrational so will expect an irrational person to behave in the same way as they do...

:)
Thanks. Makes sense now. Perhaps I perceived you as something you weren't? :drunk:

#129 Tony Matthews

Tony Matthews
  • Member

  • 17,498 posts
  • Joined: September 08

Posted 11 October 2012 - 13:05

The Air France crash off Brasil happened because the computer got lost because a pitot tube iced up, the plane kept climbing, lost it and ceded control to the pilots who had not been trained to deal with such a situation ...

The official verdict was pilot error, pure and simple.

#130 Dmitriy_Guller

Dmitriy_Guller
  • Member

  • 4,024 posts
  • Joined: July 01

Posted 11 October 2012 - 13:26

You have unwittingly, as far as you are concerned, hit the nail on the head. Choosing from unlikely scenarios is precisely what a driver has to do in an emergency, make a best estimate and act on it; computers are seriously crap at estimates, they only do 0 & 1, not much room for doubt there.

I think it's safe to say that you're not a computer scientist working at Google.

#131 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 13:39

The official verdict was pilot error, pure and simple.


The knee-jerk reaction is to blame the pilots, even if it wasn't, ask the families of the Chinook which crashed at the Mull of Kintyre. The Air France pilotswere not trained to deal with that emergency as it wasn't supposed to happen when the computer was flying the aircraft.

Qantas Flight 72 - The Australian Transport Safety Bureau investigation found fault with one of the aircraft's three Air Data Inertial Reference Units and a previously unknown software design limitation of the Airbus A330's fly by wire flight control primary computer (FCPC). So much for the infallibility of computers.

As it will be the same when an autonomous car crashes, the driver will be blamed - Too much money invested in the technology to blame it for the crash. And there is nothing in the slightest bit simple in "Pilot Error", get yourself a copy of The Human Factor in Aircraft Accidents by David Beaty ISBN-10: 1853104825.

When the manufacturers accept that they can put the cars on the road under the legal condition of Strict Liability, I will be happy to sit in one. Note my earlier statistic, 54% of commercial aircraft accidents were caused by human error, that means 46% were aircraft failure. 0.009% of Americans were injured in traffic accidents in 2010, about 4.1 fatalities per 100million vehicle miles travelled, that is pretty low, which makes me think that autonomous cars are more about making money than any human factor.



#132 Ross Stonefeld

Ross Stonefeld
  • Member

  • 56,898 posts
  • Joined: August 99

Posted 11 October 2012 - 14:04

There were also oddities in the system, like the flight controls not being in unison. So one guy was pulling up and another guy pushing down.

#133 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 16:04

There were also oddities in the system, like the flight controls not being in unison. So one guy was pulling up and another guy pushing down.


So, double pilot error...

#134 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 16:24

I think a lot of posters are missing the fact that the access to memory and use thereof are fundamentally different in humans and computers; I don't mean phisically, but rather in their raison d'étre. A computer's memory bank is the dictionary in which useful stuff is in a neatly stored array and accesible by inputting the right code; the human memory does not serve that purpose at all, it is a random access area by which we can predict our future thoughts and actions, it allows us to compare what we are faced with, or want to achieve, with what we have done and experienced before and the concomitant result; it allows our decisions to be precedent based. Oh well, you can program that in, or can you? The computer can only reference what it has been told to reference whereas the human brain can make reference to experiences it has never had by stealing another's predicament and the memory of how they got out of it. An illustration: During a French GP at Rheims, Fangio lifted off when nobody expected him to and as a consequence safely passed Mike Hawthorn's off; when asked after the race why had he lifted off he said "I could smell new mown grass, which meant somebody must have gone off the circuit." Similarly, when coming out of the tunnel at Monaco he suddenly started barking and successfully avoided a considerable accident at the chicane, which was littered with telegraph poles; when asked he replied "When I come out of the tunnel all I usually see is pink in the grandstands, that time all was brown, which meant I was looking at the back of people's heads, so there must have been an accident at the chicane." It is the ability to instantaneously connect two memories of totally different origin, and from them posit a 3rd situation, that distinguishes the human brain from that of a computer.

Edited by Bloggsworth, 11 October 2012 - 16:24.


#135 blkirk

blkirk
  • Member

  • 265 posts
  • Joined: March 00

Posted 11 October 2012 - 18:23

Bloggsworth,

If you haven't read the book "How We Decide", you would probably find it very interesting. It covers many aspects of our current understanding of how the human brain works. There are long sections on the emotional intuition side of decision making. It covers things such as radar operators distinguishing the glowing dots of incoming missiles from the identical glowing dots of friendly airplanes, firefighters making counterintuitive lifesaving decisions in the face of panic, quarterbacks choosing the open receiver in fractions of a second, and pilots learning how to fly (and land) a jet liner when all of their controls are dead.

On the flip side, I think IBM did a remarkably good job with Watson in emulating the intuitive side of human thinking. I would argue that being better at Jeopardy than the best human is a far, far more difficult task for a computer than driving a car better than the vast majority of humans.

Edited by blkirk, 11 October 2012 - 18:26.


#136 GSpeedR

GSpeedR
  • Member

  • 35 posts
  • Joined: March 12

Posted 11 October 2012 - 18:23

I think a lot of posters are missing the fact that the access to memory and use thereof are fundamentally different in humans and computers; I don't mean phisically, but rather in their raison d'étre. A computer's memory bank is the dictionary in which useful stuff is in a neatly stored array and accesible by inputting the right code; the human memory does not serve that purpose at all, it is a random access area by which we can predict our future thoughts and actions, it allows us to compare what we are faced with, or want to achieve, with what we have done and experienced before and the concomitant result; it allows our decisions to be precedent based. Oh well, you can program that in, or can you? The computer can only reference what it has been told to reference whereas the human brain can make reference to experiences it has never had by stealing another's predicament and the memory of how they got out of it. An illustration: During a French GP at Rheims, Fangio lifted off when nobody expected him to and as a consequence safely passed Mike Hawthorn's off; when asked after the race why had he lifted off he said "I could smell new mown grass, which meant somebody must have gone off the circuit." Similarly, when coming out of the tunnel at Monaco he suddenly started barking and successfully avoided a considerable accident at the chicane, which was littered with telegraph poles; when asked he replied "When I come out of the tunnel all I usually see is pink in the grandstands, that time all was brown, which meant I was looking at the back of people's heads, so there must have been an accident at the chicane." It is the ability to instantaneously connect two memories of totally different origin, and from them posit a 3rd situation, that distinguishes the human brain from that of a computer.


This description of computer function was obsoleted 30+ years ago. Even your iPhone is capable of learning and adapting.

#137 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 18:41

This description of computer function was obsoleted 30+ years ago. Even your iPhone is capable of learning and adapting.


Yeah right, I'm about to let my iPhone drive my car - A facile answer with absolutely no depth to it.

#138 carlt

carlt
  • Member

  • 1,043 posts
  • Joined: June 09

Posted 11 October 2012 - 18:55

Yeah right, I'm about to let my iPhone drive my car - A facile answer with absolutely no depth to it.


I would suggest that this is the one that is facile and lacks depth

GSpeedR was merely concise and directly to the point

#139 Kalmake

Kalmake
  • Member

  • 455 posts
  • Joined: November 07

Posted 11 October 2012 - 21:17

In Name That Tune identifying the song or all possible songs is almost trivial for a computer. Song identification is of interest to copyright holders. Google for example uses it to find and remove material from Youtube.

In this clip I found, they are given clues to help narrow it down. That part is similar to what Watson did.

Advertisement

#140 Lee Nicolle

Lee Nicolle
  • Member

  • 5,661 posts
  • Joined: July 08

Posted 11 October 2012 - 21:35

The knee-jerk reaction is to blame the pilots, even if it wasn't, ask the families of the Chinook which crashed at the Mull of Kintyre. The Air France pilotswere not trained to deal with that emergency as it wasn't supposed to happen when the computer was flying the aircraft.

Qantas Flight 72 - The Australian Transport Safety Bureau investigation found fault with one of the aircraft's three Air Data Inertial Reference Units and a previously unknown software design limitation of the Airbus A330's fly by wire flight control primary computer (FCPC). So much for the infallibility of computers.

As it will be the same when an autonomous car crashes, the driver will be blamed - Too much money invested in the technology to blame it for the crash. And there is nothing in the slightest bit simple in "Pilot Error", get yourself a copy of The Human Factor in Aircraft Accidents by David Beaty ISBN-10: 1853104825.

When the manufacturers accept that they can put the cars on the road under the legal condition of Strict Liability, I will be happy to sit in one. Note my earlier statistic, 54% of commercial aircraft accidents were caused by human error, that means 46% were aircraft failure. 0.009% of Americans were injured in traffic accidents in 2010, about 4.1 fatalities per 100million vehicle miles travelled, that is pretty low, which makes me think that autonomous cars are more about making money than any human factor.

An experienced commercial pilot friend of mine says that you never trust the computer, you should moniter what the plane is doing at all times. Too many planes have been lost because the computer has gotten lost!! And they sometimes do erratic things. He says it makes flying a plane easier and less physical BUT you have to be aware. H ehas had to take over on a few occasions. And these are 20 seat jets worth millions of dollars. Not a 30 grand car that has to make a whole pile more decisions every minute.
We all have experience with computers, the GPS that trys to send you into the ocean, onto closed roads and gets all upset when you go onto a road it says is not there.

#141 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 21:36

In Name That Tune identifying the song or all possible songs is almost trivial for a computer. Song identification is of interest to copyright holders. Google for example uses it to find and remove material from Youtube.

In this clip I found, they are given clues to help narrow it down. That part is similar to what Watson did.


I might be more amenable to your blandishments if you actually suggested how the autonomous car will get round these problems, rather than merely assert that the iPhone is very clever, so clever that it puts a Thames bridge in the mddle of a street and Buckingham Palace quite a way from where it actually happens to be; now I'm sure that you will respond by saying that this is a trivial problem, but considering that some of the best brains in the world work for Apple, and probably no less competent than those working on programs for autonomous cars, so where does that leave us, absolutely certain that there will be no bugs in the software, or are you proposing an acceptable level of collateral damage?

Edited by Bloggsworth, 11 October 2012 - 21:37.


#142 Dmitriy_Guller

Dmitriy_Guller
  • Member

  • 4,024 posts
  • Joined: July 01

Posted 11 October 2012 - 22:15

I think a lot of posters are missing the fact that the access to memory and use thereof are fundamentally different in humans and computers; I don't mean phisically, but rather in their raison d'étre. A computer's memory bank is the dictionary in which useful stuff is in a neatly stored array and accesible by inputting the right code; the human memory does not serve that purpose at all, it is a random access area by which we can predict our future thoughts and actions, it allows us to compare what we are faced with, or want to achieve, with what we have done and experienced before and the concomitant result; it allows our decisions to be precedent based.

You consistent misconception about the state of the art of computer algorithms is more rigid than you claim computers to be. What you keep pointing out in this thread are not the technical limitations of computers, but rather the limitations of your understanding of what computers can do now, or may theoretically be able to do with enough R&D.

Artificial Intelligence algorithms like neural networks are capable of both learning, and of coming to correct conclusions in unintuitive and unprogrammed ways. Computers didn't beat the humans at backgammon by having some human program in the winning strategy, or by brute-forcing through every possible combination; they developed the optimal strategy all on their own after humans set up the learning algorithm. And all that was done years ago, and we're still just scratching the surface.

You can keep repeating the same obsolete nonsense about what computer are capable of doing, or you can do some research, and maybe even develop some appreciation for the science of artificial intelligence.

Edited by Dmitriy_Guller, 11 October 2012 - 22:15.


#143 gruntguru

gruntguru
  • Member

  • 5,175 posts
  • Joined: January 09

Posted 11 October 2012 - 22:23

so where does that leave us, absolutely certain that there will be no bugs in the software, or are you proposing an acceptable level of collateral damage?

As previously stated in this thread - the technology should proceed when injury/death rates are lower for the driverless car. I would suggest the case will become overwhelming when the difference is an order of magnitude or more. I think it will be a long time before the choice (self-drive or driverless) is removed completely, so superior drivers like yourself can continue to take the safer option.

#144 Bloggsworth

Bloggsworth
  • Member

  • 7,438 posts
  • Joined: April 07

Posted 11 October 2012 - 22:51

As previously stated in this thread - the technology should proceed when injury/death rates are lower for the driverless car. I would suggest the case will become overwhelming when the difference is an order of magnitude or more. I think it will be a long time before the choice (self-drive or driverless) is removed completely, so superior drivers like yourself can continue to take the safer option.


How is the accident rate going to get to be lower without releasing into the wild large numbers of the beasts, only then will we know if they are safer or not, so I ask again, in the initial stages will there be an acceptable rate of collateral damage?

#145 Greg Locock

Greg Locock
  • Member

  • 4,477 posts
  • Joined: March 03

Posted 12 October 2012 - 00:23

Rather than join in the never ending squabble again, here's some information.

Yesterday I went to a conference on computer vision. If you look on youtube there are several videos of pedestrian avoidance demonstrations, in which the pedestrians in a given street scene are surrounded by handy coffin sized boxes, and I was learning about how those coffin sized boxes work, how they are made, and what you do with them.

Rather to my surprise the motion prediction is done with filters, so the (non mathematical) assumption is that the current path, speed and radius of curvature are the best guess for the forseeable future. That is, it doesn't assume that every pedestrian will immediately start to run directly at the car, it assumes that they will carry on doing what they are doing. Of course, as the child starts running for the ball, the prediction will be in error until a new prediction is made.

Rather more scary for the paranoid was the face recognition software, and if I understood it correctly, a given face can be edited out of a video feed, and a new face substituted, in real time, on a laptop computer. That wasn't demonstrated as an application, but I saw each part of that in other applications.


#146 pugfan

pugfan
  • Member

  • 174 posts
  • Joined: August 09

Posted 12 October 2012 - 02:43

The official verdict was pilot error, pure and simple.


Indeed, all they had to do to avoid the accident was to switch the autopilot back on...

#147 bigleagueslider

bigleagueslider
  • Member

  • 838 posts
  • Joined: March 11

Posted 12 October 2012 - 04:13

Here's one way to consider this situation. Human error is the main cause of many traffic accidents. And those human errors are mostly attributable to lack of attention or poor judgement. An automated system may be able to prevent many of these accidents. But can an automated system prevent accidents due to external factors such as road hazards, errant pedestrians, etc.? Automated controls are mostly reactive in nature, and they would not be able to foresee potential problems as well as a good driver could.

#148 carlt

carlt
  • Member

  • 1,043 posts
  • Joined: June 09

Posted 12 October 2012 - 07:56

How is the accident rate going to get to be lower without releasing into the wild large numbers of the beasts, only then will we know if they are safer or not, so I ask again, in the initial stages will there be an acceptable rate of collateral damage?


Overpopulation problem solved ?

#149 Ross Stonefeld

Ross Stonefeld
  • Member

  • 56,898 posts
  • Joined: August 99

Posted 12 October 2012 - 11:17

The Apple maps problem is, shockingly, caused by poor human input. On the designer, not user, end.

#150 Tony Matthews

Tony Matthews
  • Member

  • 17,498 posts
  • Joined: September 08

Posted 12 October 2012 - 13:09

Automated controls are mostly reactive in nature, and they would not be able to foresee potential problems as well as a good driver could.

Yet.

I don't like the idea of flying in aircraft that are soley under the control of computers, especially, as is apparently the case, the humans up the sharp end have not been trained to take over and fly it at high altitude. I don't like the idea of being in a car that is solely under the control of a computer, unless I am in the position to take control the instant that I feel impending doom, and it doesn't look like the computer is doing a good job, in which case I might just as well be driving.

However, I have learned to accept that computers, sensors and related gizmos are going to get better and better, to the point where I will be happy to give up control most of the time. I may not really enjoy that scenario, assuming I'm still here to experience it, but having had several red-face moments with technology - auto-focus? Impossible! Oh... - I now have faith in engineers to do the supposed impossible, given time.

At least I didn't proclaim, just before the event, that man would never walk on the moon, like one eminent British astronomer.