Google promises autonomous cars for all within five years Sergey Brin is promising Google's self-driving cars will be available for everyone within five years, and says that his company's current fleet of vehicles has managed to drive 50,000 miles without humans having to take the wheel. Google has over 300,000 miles of automated-driving testing under its belt already, he said, and …
"...are you feeling lucky?"
Every time you get into a car (or even out of bed for that matter), it's a question of feeling lucky.
Is the dude in front of me an old geezer with the vision of a mole and the reactions of cheese? Are they drunk? on drugs? morons? hoons?
Is there wild animals about to leap out in front? Did I just drive over a nail? Are the kids fighting on the back seat? Did my phone just ring?
The real question is of the risk. 90% of accidents are caused by humans. Can a robot with perfect reflexes, perfect attention and who can see 360 degrees as well as in the dark do a worse job than human drivers?
There will still be accidents and sometimes the robot will be at fault and sometimes people will die but people are dying right now anyway and less people will die with self drive vehicles.
The only real threat from self drive cars is to the coffers of the roadside tax collectors
Every time you get into a car (or even out of bed for that matter), it's a question of feeling lucky.
I drive every day on one of the most dangerous stretches of road in Australia. It's almost entirely single carriageway with very few overtaking lanes. Much of it is atop of three metre high embankments to avoid flooding. Armco barriers are unknown. It's filled with psychopath truck drivers that don't have a tachograph and consequently drive stupid hours, out of state holiday makers towing tin snails behind them, local yokels that wouldn't survive ten minutes in London, very large clueless animals that randomly wander out in front of you, no equivalent to an MOT for any vehicles and yet somehow I survive. Add a vehicular Blue Screen of Death to that list and I may have second thoughts.
Typical human exaggeration of skill, chance and certainties. Statistically, you and every other meatbag on the planet are terrible drivers.
And it's not exactly going to be a Windows box running in the backseat, it will be a solid, road-tested-to-fuck embedded system with chips that are at least 3 generations behind and selected for proven reliability. Look at the aviation industry...they are still using old Pentiums because they have been used day, day out for years in other applications, the software will have been tested and logged for years with the hardware and then finally approved - the systems are known inside out. Plane crashes are rare. Systems and autopilot malfunctions are even rarer. Almost all crashes are down to human error. And so it goes with the upcoming road going 'autopilots'.
"...and less people will die with self drive vehicles."
Fewer people will die.
Also, Google navigation sent me into the middle of a forest the other day on farm tracks. It was only after zooming in to the max that I saw my destination and where Google had put me had quite a few contour lines between it...I should have realised as the roads went from M to A to B to bridle path...hum.
It doesn't work by following navigation - GPS is nothing like accurate enough.
It compiles Street View, GPS, previous tracks, a video archive, radar and live video feeds to do several things.
1. It knows roughly where you are thanks to GPS
2. It compares where you look like you are from the video to make the GPS more accurate
3. It compares previously driven tracks with yours
4. It has a 3D 'memory' of the area, so can identify things that aren't usually there: be they cars in front or pedestrians waiting to cross
5. It senses distance to other objects and adjusts speed accordingly.
6. Ideally it would be talking to nearby vehicles and they'd all coordinate responses, but that's not going to happen for a long while.
Anyone with have a critically-thinking brain can see the safety inherent in driverless cars...but it just takes one robot car while the technology is in its infancy to set the unsure middle mass against it. I'm sure it will be fine up until people accept these things and everyone buys them in droves (they will be super safe after all), but a fatal crash will occur at some point and the question comes over culpability; who is responsible, the vehicle occupant? The manufacturer? The software designer? The chip makers? I for one welcome our new driverless overlords, but these are serious questions that will need to be considered sooner rather than later.
Also, don't expect an easy ride (he he) from taxi drivers either.
My biggest issue is not whether the robot will "know" when there's something wrong - it's whether it will know when something is *likely* to go wrong.
Humans, for all their issues, and vulnerabilities, and flaws, are good at recognising patterns - as thorne said, there's an element of not knowing what the people around us on the road are doing - but we can look at driving styles, techniques, driver behaviour - hell, even the type of car - and make a lot of judgements about how the driver will react before we even start to have a problem. And we can do that extremely quickly, for all the cars on the road around us. That sort of recognition requires a huge amount of computing power, and some pretty amazing AI in order for a machine to do it effectively.
Eg. If I'm driving along and come up to an old clapped out commodore - a P plate on the back, tyre rubber sprayed over the rear guards, mismatched wheels, a passenger in the front seat with his feet out the window and the stereo blaring, I can make judgements about how he will react if I try and overtake. I know he might not be paying full attention to the road, and that he may make poor decisions, sudden lane changes without indicating, etc. A computer will require massive amounts of power to see these same things - assuming a meatbag has told the computer to look for these issues at all.
The biggest risk IMO is that it'll probably be the government who decides what the autonomous car needs to "see", and in their opinion as long as it stays 5km/h below the limit no-one will ever die. Which will, presumably, be proved wrong immediately and horrifically. Then we'll never see autonomous cars on Aussie roads ever again, because it will, inevitably, be everyone's fault except the government regulators...
...Can a robot with perfect reflexes, perfect attention and who can see 360 degrees as well as in the dark do a worse job than human drivers?
Can a robot with moving parts that break or wear out, programmed by notoriously-fallible humans do a worse job than human drivers?
There, fixed it for you.
Your car is NOT already a 'robot' controlled by software... some of the operations of its systems are controlled by a robot ie engine management, anti-skid assisted braking systems, seatbelt tightening, radiator water flow, climate control air conditioning etc. None of the basic things YOU need to do say for passing your driving test and conducting yourself safely along the public highway are ALREADY done by a 'robot'. Automatic systems do not constitute the full autonomy being suggested.
don't forget the cameras and mikes inside to monitor your behaviour, to better understand how the meat in the can behaves on a long trip, what reaction this or that advert has on them... they turn away, trying to look at another window advert? / getting green at watching the same advert? / putting a black bag over his/her head? / etc.
all perfectly anonymised results, being fed to the google central and carefully selected business partners (minus the usual statutory disclaimers about gov agencies, spooks, local council, insurance companies, etc. And tax authorities, in case you happen to be sniggering at how you managed to hide this little extra income.
but no, this would NEVER happen, nosir. Like they said just over 120 years ago humans will NEVER be able to fly. They still can't, right?
OK, there's a discussion to be had about privacy with email, search results, tracking cookies and so on. Personally, I trust them enough. Definitely not entirely, certainly not without caveats and care - but enough that I can live with them.
But what I really like about Google is they're not afraid to try for the big stuff. Project Glass might be bobbins, it might be awesome, it might never even work at all - but it takes guts to take a punt on something like that. Imaging significant amounts of the planet from space down to street level is a staggering task to even contemplate, let alone actually do (albeit imperfectly, but it's still impressive). Now there's not-far-from-commercially available self-driving vehicles. ROBOT FRICKIN' CARS. Cars you can get into and say "take me home" and they ACTUALLY DO.
I have come, with time, to accept that the promised future of hoverboards and replicators is unlikely to occur, but I'll settle for a self-driving car and a wearable AR device. Probably with AdBlock+ installed, mind you.
Maybe it'll be able to drive your car for you on simple roads like interstate highways, but navigating city streets? Google Maps may be way better than Apple Maps, but they still contain enough errors that you don't want your car following their instructions and driving you into a cornfield or a creek.
How the hell is it going to deal with construction zones? Sometimes even I, a human with a lot more vision recognition ability than even Google can muster, has to slow down and think about exactly where the hell the high school dropout who placed the cones and signs is trying to tell me to drive.
Nevermind driving in rain, snow, or fog. And I'm only thinking about the US. Traffic lights and stop signs make for pretty simple intersections. Having a computer navigate a complex roundabout seems a lot more difficult. The narrow roads in some areas of the UK will present big problems too. Sometimes the only way two cars can pass without busting mirrors or worse is to drag the side of one or both vehicles in the hedges (I've always wondered if that's where the expression "hedging your bets" originated) Is a Google driven car going to be willing to hit a few hedges without tripping its collision avoidance? Or will it stop in the middle of the road and force its human to take over?
I think these problems can eventually be solved, but if overzealous idiots try to push this technology out to the general public before it is REALLY ready, all they'll do is harden a public perception that it is dangerous and/or worthless and should be banned and avoided, and set back actual availability by a couple decades.
They've already clocked up thousands of miles, in every weather conditions and traffic conditions with only one accident recorded which was caused by a human in the other car rear ending them at the lights.
How do you prove them safe to drive on the road without getting them on the road?
That said I think they must be on crack to think they'll be on the road in five years. The effort to get laws changed to allow self drive cars will take at least twice as long as that. It's not a technical issue but a bureaucratic issue. New laws written, new standards drafted, new testing regimes and a mountain of red tape that will dwarf most soverign nations, then multiply that by every country you want to sell them into cause their all different.
If that doesn't kill off self drive vehicles, nothing will.
"How the hell is it going to deal with construction zones? Sometimes even I, a human with a lot more vision recognition ability than even Google can muster, has to slow down and think about exactly where the hell the high school dropout who placed the cones and signs is trying to tell me to drive."
Simples, the "high school dropout", and every other human, will be banned from erecting any sort of barrier on the roads. After all, there'll be construction robots that can do it correctly, in a way the robot cars will understand.
I do love every time we have a story on auto-cars we get comments along the lines of "it won't handle a junction or something unexpected".
Really, you seriously don't think they thought of those things? You think they're pushing to put these into mass production but didn't do any testing apart from on quiet dual carriageways? Come on get serious, they are already legal in a couple of states.
You massively underestimate the state of modern computing.
"You think they're pushing to put these into mass production but didn't do any testing apart from on quiet dual carriageways?"
No, but some American roads are vastly different from some British roads
.How do you think one of these would cope with a "Single Track Road with Passing Places"?
What about those "Unsuitable for HGVs" roads that idiots still drive lorries down because that's where their Sat Nav said to go?
Can the computers read "Road Closed, Diversion" signs?
If the navigation data is wrong, will the computers be able to cope with the changes?
Don't get me wrong, I do like the idea of these vehicles and I think they're excellent for urban/ motorway driving etc, however there's still a lot that needs to be taken into consideration before they're ready for a world-wide roll out.
All these objections are valid difficulties the system would have to deal with (eventually).
But they are minor problems once you've got the thing able to drive itself - plenty of solutions for all these scenarios spring to even my dulled mind.
-
The real problem for this is going to come when the first robot car kills someone (or at least the first person is killed in an incident involving a robot car) - even if the overall safety record is vastly superior there will be a hell of a fuss and politicians tend to take the easy way out in these situations and bow to the noisy protests.
You've been watching too much TV. the USA might on average have wider roads but they still have narrow 1-lane roads and busy cities... if a car can self-drive in a city then that's pretty much the ultimate test.
Having good maps is all well and good but people don't need that. Just two squishy eyeballs is all we need to navigate these fiendishly difficult tasks.
Sure on a 1-track road with passing places it will be SAFE - it will not drive into the oncoming car but will slow to a safe stop. But it MIGHT not know what to do, however then you would simply take over; nobody is suggesting these are not going to allow human intervention.
So I imagine the car would safely slow to a halt and then you sort it out until the other guy stops honking at you.
I do love every time we have a story on auto-cars we get comments along the lines of "it won't handle a junction or something unexpected".
Sure, it'll handle 99.9% of everything it runs into. It's that 0.1% that humans will deal with better in many cases.
Look, I have faith that computer driven cars can be much safer than human driven cars. They don't get distracted, bored, angry, sleepy, drunk, etc. which account for the bulk of accidents. They will mostly avoid the accidents that humans get into. However, they'll also get into some accidents most humans WON'T, because of that tiny percentage of things it won't handle correctly that for humans is not that hard. That's going to be their achilles heel.
I think cars will be computer driven with a human required to be in the driver's seat and required to be ready to take over at a moment's notice the second the computer becomes uncertain of what it should do for well over a decade before we can sleep in the back seat while the car drives us home.
This post has been deleted by its author
If I am going somewhere with a railway station or a good bus service I take that because it is so hard to read a book while driving. I like the idea of self driving cars, but I would want it dual control with me in charge for years before I have the nerve to read a book while a machine drives. The obvious way to start earning my confidence is a video showing the camera's view of the road marked up with the machine's idea of what it thinks is outside.
I live near a narrow windy road. The speed selected by locals depends on the depth of the pot holes hidden around the next corner. Anyone know if Google can drive like a local? This includes remembering refuse collection day and if the hedges look really neat, dodging the tractor in the middle of the road with a hedge cutter.
Perhaps with that mindset we should ban anyone who is not local from driving on roads.
Perhaps the system will be smarter than a mindless automaton who follows direction like some meatsacks do wtih TomTom and end up in the sea. Perhaps if you bought one it would be able to register the shock of the first time it goes over a pot-hole and knows where it is for future use. Just like a local would...
Not long until there is a whole generation who has never driven a car.
If they work out well safety-wise, manual driving might be forbidden except in emergencies.
No one with driving skills is going to kill motor-racing stone dead rather abruptly!
Private ownership of cars might give way to phone-app hiring of cars by required seating or carrying capacity, they drive to you, Taxi drivers aren't going to be happy.
Top Gear will be nothing like we know it... O.K that might be a good thing.
Where are my car keys? You don't need them Grandad!
"No one with driving skills is going to kill motor-racing stone dead rather abruptly!"
I agree with your other points, but not this one.
Has horse racing stopped because horses are no longer the primary means of long-distance travel in the UK? Has sailing stopped because sailing ships are no longer the dominant kind of sea vessels?
I am not sure why commentators here assume that a driverless car is wholly driverless .. that it's all or nothing. Think about cruise control .. useful in certain circumstances, not in others. Or automatic gearboxes .. many of these have a semi manual override now.
In the same way i would expect a driverless car could be used on a motorway (freeway for the colonials) but switched to manual 'town and country' driving when required.
For long distance driving this would be a god-send .. I look forward to it happening.
Interesting:
USA 40,000 deaths, population 314 million
UK: under 2000 deaths, population 62 million.
How does the USA manage to have five times the rate of road deaths as the UK, given that the UK is such a crowded little island? Perhaps the US should be looking for some low-tech measures to reduce the casualty figures before worrying about driverless cars - or is the right to kill people with a car enshrined in the US Constitution along with the right to kill people with automatic weapons?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
