Driverless Buses Ruled Out For London, For Now 84
An anonymous reader writes The office of the Mayor of London went into a bit of a panic this week after their own paper suggested that driverless buses could appear on the streets of the UK's capital at some point in the next four decades. The Mayor's office went so far as to suggest that they were really talking about driverless underground trains. Even more bizarre was the reaction of the city's taxi drivers' association — whose spokesperson claimed that the failure to deliver 'simple' software tasks such as speech recognition meant there was no chance of driverless buses appearing on London's streets.
I'm officially old I guess (Score:5, Interesting)
When I was younger, I worked on speech recogntion problems - well, expert systems and neural networks in general. It was the toughest nut our team had ever been tasked to crack, and we didn't crack it.
When the man on the street perceives speech recognition to be simple - and coming from a taxi driver, that's more than a little ironic, considering they're essentially human Traveler Salesman Problem solvers - you know technology has overtaken you beyond hope.
Me, I can't stop being complete blown away by what can be achieved today. Driverless cars are almost a reality everybody can buy, yet I still vividly remember MIT experimental self-driving trucks trying to hold a straight line on a closed circuit at 1 mph!
Speech recognition (Score:2)
Perhaps the problem of speech recognition is that we try to teach computers our language. We could also make a spoken system where the language and pronounciation is drafted for the task.
The second unnecessary difficulty of speech recognition is to convert sounds (triphones) into letters.
Re: Speech recognition (Score:2)
Anyone else notice it said next 4 decades? 40 years, yes everything will be driverless in 40 years. That's sort of obvious, we already have much of the driverless technology in many high end cars now, with them able to stay in a lane, adjust cruise control and even stop to avoid a collision.
Re: (Score:2)
AFAIK the problem of speech recognition is that a human can use the context to guesstimate what a garbled message might have meant, but a computer can't since it doesn't have a model about the subject. It would take a fully sapient computer to reach human-level speech recognition.
Re: (Score:1)
. . .
AFAIK the problem of speech recognition is that a human can use the context to guesstimate what a garbled message might have meant, but a computer can't since it doesn't have a model about the subject. It would take a fully sapient computer to reach human-level speech recognition.
Even then its one of the most difficult tasks. Fluid conversation-speech in particular requires a remarkable degree of co-ordination, and incredible processing power - requiring a system that can solve problems in incredibly tight real time windows. Effectively the brain can only do it by making a constant stream of predictions several seconds ahead. Speech is one of the areas where organic brains probably have to use quantum 'computing' to actually work.
ATO - GoA 4 (Score:5, Informative)
Unattended train operation is a reality -- see here https://en.wikipedia.org/wiki/... [wikipedia.org]
I wasn't aware of that. See also here https://en.wikipedia.org/wiki/... [wikipedia.org]
Re: (Score:2)
It's a trillion times easier than driving a car.
The existing train protection systems have a map of the track with speed limits, acceleration and braking gradients, and what not.
Moving the trains automatically is "solved" with a huge amount of engineering, but it's hardly AI. You still need a pair of eyes to monitor everything.
The "fuzzy" problems that probably need some kind of AI includes:
- Detecting obstacles on the track ( not that important, nothing is supposed to be near the tracks anyway.)
- O
Re: (Score:2)
How so? You don't even need a computer. Just make it so the train doesn't move if the doors aren't closed, the doors move with little force, and if they fail to close they re-open and try again in 5 seconds.
If anything is outside its normal parameters, hit the brakes, cut the power, and send an alarm.
Human drivers can't really do much else, eithe
Re: (Score:2)
How so? You don't even need a computer. Just make it so the train doesn't move if the doors aren't closed, the doors move with little force, and if they fail to close they re-open and try again in 5 seconds.
I've seen a few driverless trains around the world (e.g., in Paris, Copenhagen and at ORD in the US for transfer between terminals) and they usually operate with two sets of doors: one set on the train, and the other on the platform. This keeps people from accessing the track area except when the train is there to let them board. Combine this with obstruction detection when the doors are closing (without which millions of automatic doors wouldn't be safe) and I think we can say that this particular problem
Re: (Score:2)
25 people have died on the Portland Max tracks (driver operated trains). 54 people have died on BC's Skytrain tracks (automatic trains).
Re: (Score:3)
Re: (Score:2)
Right - so how many of those suicides would have been prevented if a driver saw someone on the track and was able to stop the train successfully?
I've been on the Max where we stopped and I saw a whole set of clothes/shoes on the platform - there was a kid in his undies about a mile up the track that the driver saw, did a very hard stop quickly enough and was able to get help for this youth.
See what I'm saying? Train tracks are probably the most controlled environment for AI to exist, but if you can't handle
Re: (Score:3)
Probably not many. There's not many spots on the Skytrain track where you can see the track "about a mile up", especially coming into stations. The design of the track is recessed, which doesn't help either. Additionally, if I recall correctly most of the suicides have been of the "throw yourself in front of the train as it enters the station" variety. There are closed circuit cameras monitoring the stations (not to mention transit police some of the time), and they DO stop the trains if something goes
Re: (Score:2)
Re: (Score:1)
Right - so how many of those suicides would have been prevented if a driver saw someone on the track and was able to stop the train successfully?
Trains with drivers don't tend to stop before they run over people, especially suicides. They can't. There's a lot of kenetic energy, and very modest traction on metal wheels/rails. So probably zero or there about.
Re: (Score:2)
You didn't even read my post - I was on a train where they did stop for someone that was on the tracks - it was a very bone jarring stop too - like so fast that if I wasn't holding on to something for dear life I would have broken my nose.
I'll give you they can't stop for everyone, but there would be conditions where they could and should.
Re: (Score:2)
There is technology out there that could detect humans/animals even in the darkest portions (in tunnels etc) well in advance, outside human drivers' visual range. However whether or not that would make a difference is a big question, you can't stop a chunk of steel weighing in at 10T in a matter of seconds - well, you could (rocket boosters and whatnot) but then the meat bags inside the train would be omelets.
Re: (Score:2)
The alternative hypothesis is that as someone who has chosen to commit suicide by train, would you prefer to watched in your final moments by a train driver, or would you prefer a little privacy
Re: (Score:3)
Well first.. those two numbers are very similar.
Second- as he points out, a lot of them were suicides. Suicides also occurred on the Portland.
These are exceptions which will be figured out- and once they are there (and they will be) will never be "rookie" drivers or "sad drivers because they had a death in the family" or "old drivers" or "sleepy drivers" again.
And if you are suicidal enough to jump in front of a train (a grisly way to die), then you are probably going to find another way to suicide (like j
Re: (Score:2)
I would just like to point out that the AI required to drive a train that effectively requires no steering nor split-second decisions is completely different to what would be required to acceptably drive a car.
Re: (Score:2)
Taxi drivers have a reputation (whether deserved or not I will not speculate) for having heavy accidents. However I don't see why speech recognition is necessary. A keyboard and screen, or a touch screen would be fine to say where you want to go.
Re: (Score:2)
Why is he wrong? I agree speech recognition is a complex problem, but today its only marginally better than when I was a child and I was born in 76. I love to play around with Google now and show it off to friends - but its not perfect - I'd say its about 60%-70% on simple stuff - Google now seems about as good as the IBM speech recognition system I got to play with in the 90's and that was pretty mind blowing.
I also used do document imaging at a university, and even the best OCR engines with the best image
Re: (Score:2)
Speech recognition has been in computers since OS/2 Warp and MacOS9. It's a 'solved' problem. However we speak much slower than we either think or type/move mouses so it's a bit of a solution looking for a problem (and with mobile there are some practical uses eg. driving a car but it's still weird to talk to a device in public as if it were your butler; heck it's weird to talk to a human butler). What isn't solved very well is understanding natural language and having a 'conversation' with a computer.
Trains sound like a good idea. (Score:3, Interesting)
There have been a number of drivers' strikes that I'm sure make them unpopular. No doubt management would leap at the chance to be rid of them. The hard part will be keeping the union from finding out too soon and taking preemptive protest action against redundencies.
Re: (Score:3)
Let me guess, an american who has never seen public transport :)
Hint: the OP is talking about in-city transportation.
Re: (Score:2)
Wrong :): EU, and of coure I've seen public transport, but not used it on a daily basis except in my school time. You are true, my comment was on goods transport.
Re: (Score:2)
Re: (Score:1)
Public transport is passenger transport.
America doesn't have much public transport because the car companies bought as much of it up as possible and scrapped it, and used every other lobbying power they had to ensure people had little alternative to private cars.
In Europe that was not possible as governments see it as one of their responsibilities to ensure adequate public transport is available.
Re: (Score:3)
US moves 10 times as much over rail as Europe does, over 25% of all freight is moved by rail in the US
I suspect that this difference may be in large part due to the more widespread use of water-based transport in the EU; it's a lot more efficient than even rail (provided you've got a suitable river going in the right direction or are close to the sea, which describes more of the EU than the US).
Re: (Score:2)
Re: Trains sound like a good idea. (Score:3)
There are already some driverless underground railways in operation. For example in Nuremburg and even in London. Buses would be possible when driverless cars are working. In most cases talking to the driver is limited to asking for a ticket by specifzing the destination. A ticket machine can provide the same service.
Re: Trains sound like a good idea. (Score:2)
In London it's not necessary to talk to bus drivers. The fare is a flat rate (£1.45) regardless of distance and cash isn't accepted, only smartcard, credit card and some other prepayment method.
Re: (Score:2)
Actually, the technology is just about there for no-driver subway and commuter rail trains. Japan could probably start implementing this on their subway systems probably within a decade (they're already doing this on monorails and automated guideway transit systems).
Re: Trains sound like a good idea. (Score:2)
I'm on a driverless train right now, the Docklands Light Railway on London. It's been running without a driver since 1987, the one accident was minor, and under manual override.
A Wikipedia list was posted above of similar systems.
Link fixed (Score:2)
Well done to timothy for replacing blog link with actual news article link in summary.
How many drivers? (Score:4, Interesting)
Tried googling, couldn't find anything much other than job adverts.
How many professional drivers are there in the UK or US? Including bus, taxi, cab, private mini bus, postal, delivery and haulage? My guess would be 500,000 to a 1,000,000 in the UK alone.
That's a lot of jobs that could be lost to autonomous driving.
Re: How many drivers? (Score:3)
Taxi drivers could be replaced by smartphone apps, you'll tell your smartphone where you want to go and a taxi arrives to take you do the des
Re: (Score:1)
Only because jobs are lost doesn't mean its bad. The problem arises when it gets to distribute the advantage gained from the cost cuts. Do you give it to the rich? Do you build a welfare state? Perhaps invest into better life quality? Luxurious government systems (like Democracy -- I think democracy *is* a luxury)? Less work hours? Instead of cutting 80% of the jobs, you could make 5 people work one day a week. The question "who pays that?" misses the point: progress pais that, we only need to determine how
Re: (Score:2)
Re: (Score:2)
At least the UK has NHS.
Re: (Score:2)
Re: (Score:2)
You also need to account for the 20% of the population that live in rural areas that aren't usually serviced by busses.
Re: (Score:1)
Oh god, it will be just as horrific and devastating as when so many veterinarians, stablehands, blacksmiths, buggy makers, and drivers lost their jobs after cars replaced horses!
Re: (Score:3)
Seems to me a lot of jobs were lost when we gave up on horse-drawn carriages also.
So, should we be required to keep professional drivers employed in spite of the job being completely pointless in a few years to a decade?
False equivalence (Score:4, Insightful)
Nowadays the job market for untrained and potentially uneducated job is *shrinking*. This is not the same as back when horse cariage were gone and automobile came in.
There is a high chance that untrained and uneducated job lost today, are definitively lost thru job market shrinkage. Think about that. Think about what that means for the economy as a whole when 100.000 jobs are lost. Nothing good for the economy or for the social stability.
Re: (Score:2)
Re: (Score:2)
It's spelled heighth; you know, like "length, width, heighth and weighth."
Re: (Score:3)
"So what?" (Score:2)
Yes, and the buggy-whip makers won't be pleased, either.
That argument does not pass the "So what?" test, and never has. Technology advances, and society changes; it's why we're not still all running around dressed in skins, hunting down our food with rocks. When it does, some types of job inevitably become less-sought, or even redundant. If that job's what you do, or me, that's just tough; the world doesn't owe anyone a living. If your job goes away, you go look for another. Yes, there's a social argument f
Re: (Score:2)
Well, from Norway the total figures for land transport of goods and people is 4,6% of the employment, though that may include related service like loading and unloading. However, there's also work in the primary industries (agriculture, forestry, mining) and many operating trucks and such in production industries that are also potential targets for automation. The biggest productivity boost is that people could use the time they spend driving for other things though. Limiting myself to personal cars (not ta
Future Shock (Score:2)
Good Job London (Score:2)
touchscreen underwear.
Any other recommendations?
Re: (Score:2)
(files patent)
Driverless trains (Score:1)
...are already quite common. That's nothing new. Fully automated subway trains have been deployed all over the world. I recently rode one in Barcelona (the new L9/L10) and it was quite nice. That should be (relatively) uncontroversial. Busses navigate extremely complex environments. Subway trains navigate remarkably simple, controlled environments.
Re: (Score:2)
We already have them in London on the Docklands Light Railway. Been running since 1987
http://en.wikipedia.org/wiki/D... [wikipedia.org]
Even some deep level Tube lines were built to have driverless trains but the Unions won't let Tfl introduce them.
Re: (Score:1)
I find this the most interesting comment in the thread so far. I was wondering, how are driverless buses within 4 _decades_ unlikely?
Of course they are unlikely, because the union will stop them.
Re: (Score:2)
Why do you need speech recognition? (Score:2)
Why would you need speech recognition on a London bus? You never talk to the driver. You get on, touch your Oyster card to the reader, and get off when you get to your stop. That's it. It's a flat rate fare. You can't even use cash on them anymore - you have to use an Oyster card.
Comment removed (Score:4, Funny)
Re: (Score:2)
Buses are technically easier to automate, but (Score:2)
Trains are actually the easiest vehicle to automate, because they run on fixed tracks. The automation system need only worry about merge points, plus what is ahead and what is behind the train. Buses are easier to automate than cars because they run on fixed routes at relatively slow speeds, but the liability exposure of a busload of people in an accident is much greater.
Re: (Score:1)
I imagine the elimination of blind spots is another thing that will be an even bigger advantage for busses than cars.
Public transport will be obsolete (Score:1)
Once driverless car technology has sufficiently matured, there will be no need for buses, underground trains, or any other current public transport system. City authorities should be planning for driverless cars (including driverless "taxis" for those who do not own a car) instead of continuing to think about and invest in soon-to-be-obsolete modes of transportation.
In addition to the regularly mentioned benefits of driverless cars, such as reduction in road deaths, and freeing commuters to use travel time
Re: (Score:2)
Once driverless car technology has sufficiently matured, there will be no need for buses, underground trains, or any other current public transport system.
Are you sure about that? You seem to be assuming that everyone will be travelling from and to different places and that there will be no concentrations of people attending the same location at the same time. It's been my experience that people don't work like that. I also suspect that the price that these vehicles would charge would make them rather less economic than you think. Unless there's evidence that what you propose would be cheaper than public transport currently is, or that there will be no common
Couldn't be worse. (Score:1)
I've seen bus drivers take a corner without considering the other lane, and wipe out a driver and passenger in a truck, waiting in the turn lane. I've seen a bus driver carelessly activate the bus-stairs-convert-to-wheel-chair-lift before it was safe, completely knocking over an elderly wheel-chair bound person onto the concrete, head first . . . and then just sit there, not doing anything, requiring myself and another passenger to jump off and assist the person.
I don't see how automation can do much worse.