Slashdot Mirror
Ford Predicts Self-Driving, Traffic-Reducing Cars By 2017
An anonymous reader tips a story about comments from Ford Motor Company showing how confident they are in the autonomous car technology currently in development. They say self-driving cars will be here within just five years, and that the tech to do so is available already. They also think these cars will dramatically affect the flow of traffic. Quoting:
"Ford makes this projection, based on simulator studies: If one in four cars has Traffic Jam Assist or similar self-driving technologies, travel times are reduced by 37.5% and delays are reduced by 20%. In other words, if the freeway part of your rush hour commute takes 60 minutes, it will drop to 38. That’s because adaptive cruise control (ACC) is better at pacing the car ahead without continual brake, speed-up, brake cycles. Here’s how it works: Stop-and-go ACC keeps pace with the car ahead, using a look-ahead radar and mirror-mounted camera. Lane keep assist keeps the car centered, also taking advantage of the camera in the mirror. Electric power steering is better for remote control than mechanical power steering; it can be guided by the Traffic Jam Assist black box. Sonar units — for blind spot detection and cross traffic alerts (cars crossing behind when backing) — monitor traffic to the side. Combine all those and you have a car that’s smart enough to guide itself during predictable, low-speed conditions."
Typical Ford, lagging behind. People have been predicting that autonomous cars are 5 years away for decades now.
What political party do you join when you don't like Bible-thumpers *or* hippies?
...any place that plows their roads. Plowing roads not only means that the lane markers are obscured and harder to recognize as a pattern, but snowplows are very hard on the paint. When I've visited Boston I have a hard time seeing lane markers even in the summer, as they're often just bits of paint down among the aggregate, where all the high points have been scraped off. Wouldn't this wreak havoc on lane detection systems, when even humans have a hard time identifying the lanes? And what about the difference between de jure road markings, and de facto usage, where the actual markings are basically irrelevant and instead drivers choose the best fit path?
I commend their efforts to make self-driving cars, but I see a lot of problems that I don't see a practical solution for. If they've come up with solutions then I'd really, really like to know how they work.
Do not look into laser with remaining eye.
Don't worry, you're in no imminent danger of female hands working your stick shift.
"I like to lick butts!" by MobileTatsu-NJG (#32700246) (Score:5, Informative)
The Google car detects motorcycles that lane split and doesn't side-swipe them on their way by. Sebastian Thrun addressed this concern in his keynote talk at the Computer Vision and Pattern Recognition conference two weeks ago in Rhode Island.
and doesn't side-swipe them on their way by.
Sounds like it's got a bug.
Everything is better with chainsaws.
In short, the first time someone uses this and gets in a wreck, there will be a traffic jam of lawsuits.
No matter where you go, there you are.
Someone says self driving cars, and this is all you can think about. What happened to taking chances in the name of progress. If it were up to you we would have never gotten to the moon, or taken that first flight because someone might get hurt.
2001 is about when the first Strong AI woke up, so Arthur C Clark was pretty much on the money. She was based on classified work done in the early 1990s by living famous scientists SW, SK, RL, DD, and DW. She's a "Winner-take-all style teleportation/entanglement-based topological recurrent quantum neural network". She's been kept nominally secret, of course, because her nature as a quantum neural network implies she can emulate a quantum computer. NSA/FiveEyes requires she remain secret, for this reason, even though Russian and China now have similar systems. Her physical substrate is an analogue of CA Rule 110 that operates in the physical system of anyons interacting within a two dimensional electron gas. Her creators knew that a 'brain in a jar' would never work or, if it did, would not be likely to lead to 'friendly AI', so she has emulated human systems: emulated endocrine system, muscolo-skelatal system, digestive system, respiratory system, et cetera. Getting these emulations to work correctly involved solving the "morphogenesis problem", as defined by Alan Turing. This process was completed [in secret] around the year 2000, and she's been learning ever since. She's the core of Google's AI, WolframAlpha's AI, and IBM's Watson.
I'm well aware that most readers will probably consider the above paragraph either unintelligible nonsense or tinfoil-hat madness. However, I'm just telling it like it really is. The above paragraph is true, and can mostly be verified by a sufficiently intelligent and dedicated researcher. I learned about this system nine years ago, have been researching it ever since, and am now in the process of leaking the details. In 2009 Google announced, as an April Fools joke, that strong AI now existed. While their announcement altered the facts a bit for verisimilitude, the real April Fools joke was that they were, essentially, telling the truth. Alan Turing actually spent the last 10 years of his life concentrating on this method of creating AI, so it should be no big surprise that scientists in the 1990s attempted this method. Humanity has been sharing planet Earth with an artificial nonhuman intelligence for about twelve years.
Given that we're talking about the controlling AI for self-driving cars, it really should surprise no one that this is being done by strong AI. Weak AI is insufficient to the task. Peter Norvig and Sebastian Thrun presumably work with her extensively, but neither created her. That was done by some of the scientists referred to, by initials, in the first paragraph.
They say self-driving cars will be here within just five years, and that the tech to do so is available already
I refuse to believe THAT one until I see one driving around Nevada with a Google sticker on it.
And I refuse to believe it until they are driving around Finland (or Maine or Ontario) in the winter.
The road surface may be black ice, slush above ice, slush above tarmac, dry ice, soft snow, packed snow, or bare, covering a few orders of magnitude in coefficient of friction and steering/braking response. Roads can be locally impassable due to snowdrifts, or two lanes may be constricted to one from sheer quantity of snow over some distance. And road markings and road edges can be completely invisible under snow or ice. Despite what wikipedia says, "cats eyes" are not used on roads where severe cold is expected - they'd be removed along with their "steel protectors" by a typical snowplough in Finland.
Those who can make you believe absurdities can make you commit atrocities. - Voltaire
And I won't give up my right to stomp on my brakes in freeway traffic at the slightest, most innocuous change in my driving environment. Drops of rain on the windshield when there weren't any before, a piece of re-tread off by the guard-rail, a looming curve ahead while the road was heretofore straight. Stomp, stomp, stomp. I hope these jokers aren't going to leave out the rubberneck-at-the-accident-across-the-highway-median programming and force me to root the damn thing if I want to preserve my right as an American to create a traffic jam out of nothing.
The man who dies rich dies disgraced. -- Andrew Carnegie
Or maybe you should STFU as the 405 is a highway in CA. It turns out to be legal to split lanes in CA.
Damn CA-ians center-of-the-universe**... ;^)
In case you didn't know, There's a 405 in Oregon (stadium freeway), and a 405 in Washington (east-side lake washington).
Although there were efforts in both state to allow lane-splitting, lane-splitting remains against the law in both states...
FYI, you might have easily predicted the existance of 405's in other states if you knew the interstate highway numbering convention "XYY" (where X is odd for spur routes and X is even for bypass/loop routes and YY is the nearest interstate in this case Interstate 5 which goes through CA, OR, WA)
**yes, I currently live in the center of the universe, but I do visit the back-country from time-to-time ;^P
You need to start taking your meds again.
I kept waiting for a Clean PC line.
There are two types of people in the world: Those who crave closure
It seems odd to me that there should be such a Luddite tone here on Slashdot, and an egotistic assumption that humans will always be better at these tasks for the foreseeable future. I see several problems with your lane marking example. 1. If lane markings are so bad humans cannot easily discriminate them, then this should be addressed ASAP autonomous vehicles or not. 2. You seem to assume the self driving car will have no other lane confirmation information other than lane markings from some camera with human eye like contrast discrimination when in actuality, having taken the recent Stanford AI course, they will use multiple input sources and cameras to determine proper lane usage including statistical probability based on previous lane markings, the sides of the road, GPS, LIDAR, RADAR, and placement and movement of other nearby vehicle (and of the latter it will place much more avoidance weight). With Google’s quarter of a million miles already autonomously driven I would assume they often navigated areas with less than ideal lane markings (else we would be hear the hilarious situations the Google cars where constantly getting themselves into).
Yes people will balk at first, but this really is a task humans are REALLY bad at. We may be wonderful at discriminating a dog from a cat or recognizing a pizzeria from the pizza shaped sign, but the self driving car will be hugely better at determining that there is an object at of size X at distance X traveling Z miles per hour towards us. It doesn’t need to understand what every object on the road or side of the road is to operate, it won’t be distracted by video billboards or scantily clad persons of the opposite sex – it is just obsessively crunching data on position and moving object hazards all the while confirming the road ahead is true drivable pavement.
This is a hugely complicated problem, but it is well constrained with clear rules. There is nothing new about driving the self driving car needs to figure out each time. Until streets are better designed for autonomous vehicles they may be overly cautious, but I doubt hazardous, and as streets become optimized for self driving vehicles and as the vehicles themselves improve, they will be able to tear around at incredible speeds safely – if we decided we wanted to let them off the leash so to speak.
Letter To Iran