"Jcarr" (jcarr)
06/21/2016 at 10:06 • Filed to: Autonomous Cars | 0 | 16 |
Interesting piece in the !!!error: Indecipherable SUB-paragraph formatting!!! . What does Oppo think?
How We See It: Autonomy and Self-Driving Cars
The fully automated car will never happen. We don’t say this because the word “Driver” is in our name, or because we see the automated car as a threat to our magazine’s existence. We say it for one very simple reason: Automated cars will require closed (yet vulnerable) networks of vehicles in constant communication, and as long as Main Street is a place where there are motorcycles, cyclists, people, and dogs, it will never be a closed network. There will never be an automated dog, much as we find that notion intriguing, and dogs will never be able to communicate with cars. Barking doesn’t count.
There are other reasons why full automation will never happen, but they only serve to underscore the point above: Unless every object in the streetscape is fully automated, nothing can be. Sure, the insurance lobby will fear cars that promise not to crash and will fight their advancement. Carmakers will be loath to assume all responsibility for a car’s actions. The ethical issues, such as whether a skidding car should hit a single mother or a father of five, have been well documented and will be significant, of course. As will carsickness, the psychological loss of control, the absence of self-determination. But nothing will keep the car from being fully, completely automated like a dog with free will.
According to SAE International (the Society of Automotive Engineers), there are six stages of automotive automation. We are in the third , wherein the many features loaded into our cars seek to keep us in our intended lane and at the appropriate distance from the car ahead. The fourth is close, but not without issues. The sixth stage is the complete absence of driver involvement. It is our assertion that we will never get to the sixth stage. The unfettered movement of fully automated cars among free agents will require a system that contains some element of risk, and risk is antithetical to automation.
No, we’re about where we’ll stay forever, with incremental improvements, in the same way planes have long been capable of automation yet still require pilots. Maybe it’s because nobody wants to board an Airbus A380 to Melbourne and see an empty cockpit. Or maybe it’s because an animal could get through the runway fence.
THE SIX STAGES:
SAE standard J3016
—
NO AUTOMATION
Old-school motoring with a human driver, a steering wheel, and, ideally, a clutch pedal, and an attractive passenger in the right-hand seat.
—
DRIVER ASSISTANCE
Steering or braking/acceleration can be performed by the vehicle, based on the environment and the situation. The driver is in control the rest of the time.
—
PARTIAL AUTOMATION
Coordinated steering and braking/acceleration are performed automatically by the vehicle, based on the environment and the situation. The driver does the rest of the driving.
—
CONDITIONAL AUTOMATION
The vehicle can assume all aspects of the driving task, but the driver will intervene as necessary.
—
HIGH AUTOMATION
The vehicle can assume all aspects of the driving task and keeps on doing so if the driver does not respond to intervention prompts.
—
FULL AUTOMATION
The Robots are in control, driving all the time. Nothing to worry about. But they will be needing your liver.
415s30 W123TSXWaggoIIIIIIo ( •_•))°)
> Jcarr
06/21/2016 at 10:21 | 2 |
The baby stroller VS shopping cart problem with AI is huge. A vehicle will have to make decisions based on the situation. Swerve into the shopping cart and not into the obvious human. We’ll tests show they pick a baby stroller confusing it with a meaningless cart. You can list many examples where a split second human decision would save a life. Of course humans get tired etc... and do stupid things. But they accept highway crashes, people don’t accept babies being crushed.
BReLp7dzHM3ytYsE
> Jcarr
06/21/2016 at 10:29 | 2 |
You can pry my steering wheel from my cold dead hands. Screw autonomy.
This is a good article that ran about four years ago, also from C&D. http://www.caranddriver.com/columns/aaron-…
BReLp7dzHM3ytYsE
> Jcarr
06/21/2016 at 10:32 | 1 |
http://www.caranddriver.com/columns/aaron-…
xyzabc
> Jcarr
06/21/2016 at 10:33 | 2 |
...and because they’d have to change their name to “ CAR & PASSENGER ”.
Nibbles
> Jcarr
06/21/2016 at 10:40 | 4 |
I don’t agree. While closed networks, as they put it, could be an ideal, it’s not a necessity for full automation. A car’s sensor networks have 360°, up-to-the-millisecond data on all three axes and can make a decision, plot a course and act on said course quicker than a human can sense danger and check two of their three mirrors. C&D makes the statement up there about “whether a skidding car should hit a single mother or a father of five” let me posit this: Why should we hold the car responsible for making that snap-decision and judgement when we, as human beings, cannot make that decision ourselves? We don’t know if Johnny Three-Piece over there is a family man or a bachelor day trader, all we know is we just rammed him in the shins.
It’s also stated in the article above that Carmakers will be loath to assume all responsibility for a car’s actions. At this point that is an untruth. Three of the major players in automotive automation - Google, Volvo, Mercedes Benz - have all stated that if their vehicle causes damage while in autonomous mode they (the manufacturer) will accept all liability.
Unless every object in the streetscape is fully automated, nothing can be. I do take exception to this because it’s already being proven false. As well, as an automation engineer, this statement goes directly against everything I work for. If the above statement is true then it would be true not only for the streetscape, but other scenarios where automation is pursued (and gained). Automation isn’t about removing the human (or mammalian, or avian, or what have you) element(s) entirely; it’s about minimizing, mitigating, or separating the possibility of a living being’s intervention and, by proxy, error.
...Risk is antithetical to automation. That is incorrect as well. Risk is the driving factor of automation. Automation itself exists because of risk . We know the risks inherent in piloting an automobile will never be fully removed; that’s no reason to stop mitigating. We are passing the 3rd stage of automation and heading clearly into stage 4. There are production vehicles that one can purchase today that nearly achieve full stage 4 automation. Tesla’s latest generation Autopilot is in the stage 4 category on the SAE scale, and 3.5 in NHTSA’s.
We haven’t reached peak automation yet; there’s still plenty to do. I doubt we’ll reach complete level 6 on the SAE scale (as there will always be people who drive themselves, keeping complete level 6 out of reach), but 4 is achievable in the very near future, 5 is maybe 10-15 years out.
Ash78, voting early and often
> Jcarr
06/21/2016 at 10:49 | 0 |
I do agree and will put some words in their mouth (and SAE's) that each stage is an exponentially greater leap than the one before. What do you nerds call that, a logarithmic scale? That. It will take a couple decades for all new cars to reach the fourth stage, then (barring some serious and overbearing legislation) we'll probably never hit stage 5 en masse. Some cars might touch on 5-6, but I also agree that unless we have uniformity across all cars and predictability for all unknowns, it's not going to resemble a sci-fi movie in any of our lifetimes.
Urambo Tauro
> Jcarr
06/21/2016 at 10:56 | 0 |
Fully-automated cars will inspire assholes of all kinds to test their limits.
While that may sound like progress at first, it will quickly show the car to be highly vulnerable. Take jaywalking for example. People will cross the street whenever the hell they feel like it because it will have been proven that the car will stop for them. YouTube will fill up with more prank videos, too.
My bird IS the word
> Jcarr
06/21/2016 at 11:33 | 0 |
I feel that it is one of those sci-fi technologies that
seem
plausible, until implementation is considered. Similar to how our cell phones cant act as super universal remotes ( turn our lights on, cars, TVs, ect.) and similar because there are too many companies with different software and coordination. These cars don’t exist in a bubble, and computers by their nature can only do what they are programmed, no real “thinking” exists.
My speed3 is happy
> 415s30 W123TSXWaggoIIIIIIo ( •_•))°)
06/21/2016 at 11:41 | 0 |
We’ll tests show they pick a baby stroller confusing it with a meaningless cart.
I am pretty sure it will correctly identify the cart. Google photos can already recognize different breeds of dogs and types of pasta right now .
Skip to 34:54
Math is powerful.
My speed3 is happy
> Jcarr
06/21/2016 at 11:55 | 3 |
This will probably be one of the articles that will be remembered 100 years for now for how bad we are at predicting the future or how short sighted people from the past are. It will be the same as all those articles written about how “man will never fly”
In the September 1901 issue of McClure’s Magazine Simon Newcomb who was a distinguished astronomer and professor at John Hopkins dismissed flight as myth and asked what useful purpose could it serve? “The first successful flyer will be the handiwork of a watchmaker and will carry nothing heavier than an insect.”
There were more articles like this at the time written by very respectable engineers.
Keep in mind by 1901 the Wright brothers were laying out plans to travel to Kitty Hawk to test their 3rd glider.
Nibbles
> My speed3 is happy
06/21/2016 at 11:58 | 0 |
Not to mention the sensor arrays can pick up on heat signatures. My Xbox knows the difference between me, the dog, and the couch; I’m pretty certain the autocars can too.
My speed3 is happy
> Nibbles
06/21/2016 at 12:06 | 1 |
Exactly, an autonomous car has full 360 degree awareness whereas humans are blocked by the a-pillar and limited by not having eyes in the back of their heads.
Another video!
Sorry, as a software engineer this stuff is fascinating to me.
Nibbles
> My speed3 is happy
06/21/2016 at 12:09 | 0 |
As an automation engineer, it’s interesting to me too :)
davesaddiction @ opposite-lock.com
> My speed3 is happy
06/21/2016 at 13:48 | 0 |
He said it “helps you express yourself and keep your conversations going”. I’m pretty sure he meant to say that it “helps you to communicate with other people in your life without really communicating with them”...
davesaddiction @ opposite-lock.com
> My speed3 is happy
06/21/2016 at 13:53 | 0 |
It is fascinating, for sure.
415s30 W123TSXWaggoIIIIIIo ( •_•))°)
> My speed3 is happy
06/21/2016 at 15:39 | 0 |
Well maybe if it has time, they tried situations where you hit one or the other and that microsecond you need to form a decision didn’t look good enough with the machine.