Slashdot Mirror
Mercedes-Benz's Self-Driving Concept Car Is Here
cartechboy writes: Mercedes-Benz has finally taken the wraps off its autonomous concept car, dubbed the F015 Luxury in Motion Concept. Shown at the 2015 Consumer Electronics Show in Vegas last night, the concept is a self-driving, fully-connected, hydrogen-electric plug-in hybrid that touts a vision of driving in the future. Mercedes says this concept is not only a means for getting someone from one point to another, but also a usable space for entertainment or work as well as a platform for communication and interaction. The hydrogen-electric plug-in hybrid system is unique in that it produces zero emissions at all times. It consists of a hydrogen fuel cell stack, a lithium-ion battery, and two electric motors. The F015 has a driving range of 124 miles with a fully charged battery, and up to 684 miles with a full tank of hydrogen. While not intended for production, Mercedes shows us that it has the technology today to produce a zero-emission vehicle that can drive itself. In related news, Audi has just shown off an A7 that drove itself 550 miles from San Francisco to Las Vegas for CES.
Did your driving test cover that?
If I have been able to see further than others, it is because I bought a pair of binoculars.
None of today's self-driving car is ready for production and will NOT be until we make real AI. Too many variables for that kind of system, pattern recognition and prediction systems that is unreachable for current algorithms and electronics. Maybe in 30 years from now.
I would like to see mandatory personal vehicle ownership go die in a fire. I hate having to own a car just to go to work. If I can get a reasonable alternative, say a company has a fleet of these things to pick people up and drop them off I would be so happy.
And if that is the response, that is why autonomous cars will NEVER work on public roads.
Either the car drives itself 100% of the time, or I drive it 100% of the time.
If someone thinks that you're going to be driving along in your car not paying attention to the road, and suddenly the computer is doing to say "fuck it, I don't know, you do it" they're complete morons.
Human reaction time will not allow a driver who is disengaged to suddenly be in control of the car.
There is no "sometimes the computer drives and sometimes you do", unless it's a complete transition. But when the computer is driving, I should be able to climb in the back and sleep, or read the paper.
Otherwise the entire system is doomed to fail, because it simply won't work in the real world.
I see zero value in an "autonomous" car which is periodically going to decide it isn't responsible. And I'm sure neither Benz nor Google plan on indemnifying you from legal responsibility.
In which case this will always simply be a toy.
Lost at C:>. Found at C.
The nuclei were formed in the first few seconds after the big bang. Molecular hydrogen didn't form until about 380,000 years later, after the photon epoch.
Man will you be surprised in 3-5 years.
Just look at voice recognition's progress within the last two years alone. It had reached a plateau of about 80% accuracy, even when speaking slowly, clearly, and without regional accent. Now? I can mutter drunkenly and it'll get it more often than not. I can say very local business words like Phydeaux (upscale pet store) and get the spelling right!
It's called machine learning. The more experience/scenarios etc these self driving cars get, the better they'll be.
I'd say I'm as confident as you (but in the opposite direction) that within 10 years computers will out-drive humans in all scenarios. It won't even be close. 360 degree sub-millisecond informational input versus our meatbag eyes and reflexes??
Meanwhile, back in the real world, my car can't even figure out when I say 'redial', and gets about four of of ten numbers wrong when I try to dial a number directly.
You want to convince me you are SERIOUS about getting into the driverless car? Then build a Concept Bus - or Concept Garbage Truck.
Those are large vehicles that honestly do not need drivers. They are expect to drive slow, not fast and usually travel set routes. Small cities can easily afford to self-insure them, and they won't have to worry quite so much about the stupid technology ignorant laws, as they will be purchased by the people that enforce, if not write the laws. Finally they are already expensive and the cities pay large salaries to people to drive them.
They will in all probability be the very first driverless vehicles we actually see on the road [as soon as we 1) convince the unions to let us and 2) actually get them to work.]
So forget about concept 'cars' and show me a concept bus or concept garbage truck.
excitingthingstodo.blogspot.com
We don't need real AI. We need good-enough-to-be-safe AI, with human intervention available when the AI calls for it.
That was in 1997 when I worked at what later became the KIT.
Back then they tested an early artificial neural net controller under real life conditions on the Autobahn A8. The driver just sat with his arms folded behind the wheel.
This technology has been a long time coming and still lawmakers haven't caught on to it.
Stop being so extreme... I already have a car that will not automatically stop without hitting the car in front of me. Yes, in rare cases (extreme stop or someone cutting me off so close the car doesn't see it), I have to respond. But unless I see one of those two scenarios, I just cruise home while chatting hands-free with my wife. And it's wonderful.
Peter predicted that you would "deliberately forget" creation 2000 years ago...
I disagree. It doesn't have to be a "sudden" transition... the transition can be in planned scenarios (like transitioning from freeway to non-freeway driving). The car could also pull to the side of the road when weather conditions become too adverse.
"Driver must be prepared to take control at any time" is a reasonable disclaimer for early self-driving cars, and may even become the law in many states.
Just because you won't buy such a car doesn't mean there are not plenty of people who would.
All my liberal friends think I'm a conservative, all my conservative friends think I'm a liberal.
I disagree with your assertion, and it just so happens that I have a software development background that includes voice recognition, visual interpretation, and vehicle communication systems.
Voice recognition is a really cool subject. There are a finite (and surprisingly small) number of sounds that make up the English language. But for each of those sounds there are nearly an infinite number of dialects. This is combated by looking at the context of each individual sound to find known sound patterns. This can still result in thousands of different words, so we look at the context of the sentence. And through out this effort you have to deal with people talking quickly/slowly, while emotional or out of breath, with a southern drawl or a New York accent, or ESL folks that may say put words in an order that makes the context obvious. Heck, we did it for early onset dementia patients. The amount of work it took to dial in the VRE even when we knew the specific user and their inflection/dialect was massive.
When it comes to visual recognition, the AI for a car doesn't need to understand what a dog is, it just has to recognize that there is an IR source on the current trajectory. The AI doesn't need to comprehend what a boulder is, it just needs to recognize that there is an obstruction in the road.
The AI doesn't need to be able to identify that there is a lady wearing a yellow spotted sun dress doing interpretive dance while high on mescaline in the middle of the road, it just needs to identify that there is something in the road and respond accordingly (slow down, swerve, stop, etc...)
The exact same thing is true for humans. For example: the other day I was driving my wife's car. I was pulling out from the gas station and something caught my eye after I thought traffic was clear. So I slammed on the brakes thinking I missed something. Turns out it was a reflection of a dash light on the side window.
Yeah, there will be false positives. A vehicle may decide that a pothole is actually an obstruction, or that the railroad track is the end of a road. But in the vast majority of cases these are targets that the vehicle can identify from a significant distance away. It's not like you're going to be driving on the open road then come screeching to a halt for no reason.
And each of these false positives is something that will be handled through refining the AI system. Through IR, RADAR, real-time 3-d surface mapping, V2I communications, V2V communications, etc...
It's coming, and quickly. And many municipalities and states are looking to leverage the V2I systems. If you have a V2I enabled vehicle in Las Vegas, you can actually get the system to tell you how fast you should drive to avoid hitting red lights. Most major metros already have systems in place picking up tire pressure and blue tooth signals to determine traffic volume and speed, which is how those fancy "12 minutes to exit 123" signs get populated.
As the 2017 cars start rolling out and more data becomes available, we'll see technology leaping ahead. For example, in Wisconsin, our 511 site has a public facing developer API, so even if the state can't invest in some cool new apps and vehicle information systems, individual developers and manufacturers can: http://www.511wi.gov/Web/extra...
-Rick
"Most people in the U.S. wouldn't know they live in a tyrannical state if it walked up and grabbed their junk." - MyFirs
> There was a time when Slashdot was full of people who'd think more than two seconds about their brave new economic ideas, rather than just demanding a pony.
Yep, that time was October 5th, 1997. Then the second user joined the site, and the idiocy began.
"And the award for most contrived example goes to...." I kid, I kid.
But seriously, how many human drivers would pass your test? I'd imagine they're do worse that the automated cars by a long shot. Here's why: a human driver in your example is going to to be overwhelmed with information as soon as the snowboarder appears, in this confusion they are likely to forget the dangers of swerving right or left.
In contrast, the machine does not become overwhelmed. The world passes very, very slowly to it. What does it see? Everything in your example, but in a simplified context. It sees the guardrail. The guardrail is a known item to it, it signifies danger. It cannot cross the guardrail. It sees the truck in the other lane, it is a large object and moving quickly, it cannot move into the path of this object. There is a new object in the road, an object that maybe be a person or animal. It may have to collide with this object. It begins to apply the brakes in a controlled fashion, and lightly impacts the snowboarder. All because it wasn't driving 60mph on an icy mountain road like a goddamn idiot.