Posted on 10/01/2016 11:35:32 AM PDT by nickcarraway
This might sound like a really terrible idea, but lets give Nvidia the benefit of the doubt for now. Theyre letting their AI-driven car learn solely by watching humans drive.
Nvidia calls their prototype vehicle BB8, and while that name might be more befitting, say, a Volkswagen Beetle, its actually a Lincoln MKZ. Apparently even AI drivers want to feel like Matthew McConaughey diving backward into a rooftop pool when theyre out cruising around.
While you can pick up a Lincoln thats loaded with near-autonomous driving features, Nvidia shut them all down in their test vehicle. They wanted the car to use its neural network to figure out how to drive.
That meant observing human drivers on the road. Sounds like theyre setting the poor AI up to fail, right? Our streets are, after all, teeming with drivers who run red lights, stop in crosswalks, turn without signaling, go left of center, and text the entire time theyre driving.
Fortunately, Nvidia had a more sane plan in mind. They mounted cameras to other test vehicles and then took them out on the road. The video footage they recorded was then handed over to BB8 so that the AI could analyze it and figure out how to reproduce the behaviors it observed.
Neural action shot showing how BB8 analyzes its surroundings to figure out where to drive
The results are pretty impressive. Sure, BB8 mows down a traffic cone at the start of the video, but it handles plenty of real world situations including driving on dirt roads and driving in the dark with ease. Again, its doing this all without ever being explicitly told anything like how to identify the edge of the road or what another car looks like.
That was a cultural prediction, which was silly because flying cars are useless. Technically quite possible, but utterly useless, the wing span would make them undrivable, and the mechanism to bring in the wings would make it so heavy that they’d have no load capacity as a plane. Mine is a technical prediction. This is a technology we want, we know it’s wanted. People have been working towards it for ages, once it’s possible it will take off in the market place. And we’ve now put together the technology to show it is VERY possible. When you get right down to it self driving cars are nothing more than data crunching, they need to “see” the world around them and react to it, the same thing we do. Data crunching is the area where technology grows the most predictably and consistently. We crunch more data faster every day. The smartphone in your pocket as 10 times as much data processing as Deep Blue. It’s not if but when, we WILL have enough computing power available small enough and cheaply enough to do this. Probably we already do, we just need to write the logic.
You’re living in the future, don’t be so grumpy. They’re amazing times if you’re willing to be amazed.
“Theyll turn your car into a cab that you own, get in, sit down, shout an address and relax. “
Cool. I can pimp my car out as a taxi cab and it won’t care.
You may call me grumpy and maybe I am. I call it being a realist. I am also one of the most willing to embrace technological advances. For most of my life I was engaged in a field where each new year brought new tech to machines I would operate.
As those around me griped the machines were already good enough I became the early adopter of the new tech and soon I was able to do amazing things and reached the top of my field. The manufactures would seek my input on the latest changes they had made and bring me prototypes for me to use and critique.
This was all in the field of heavy equipment operation and a segment that would lend itself easily to self driving or autonomous operation. Through my nearly forty years in the field the hope was always to automate the operation. The furthest along this technology has come is high resolution GPS tracking.
I’m a professional driver and heavy equipment operator with vast and varied experience. The challenges to automating off road equipment operation are exponentially less than on road self driving challenges. One of the greatest challenges to on road operation are as Don Rumsfeld may say “the known unknowns, and the unknown unknowns”. Factor in blending self driving and human controlled vehicles the unknown unknowns become vastly more complex.
You made a statement that automating rail operations in the NE corridor was being stymied by union resistance, while certainly that is a component it is also a simplistic justification for looking no further into why block protection has not bee fully implemented as of yet. If the can’t even make that happen how can “they” assure safety of self driving vehicles on public ways?
That in a nutshell is why I don’t see implementation of autonomous driving in your rosy time line. The first time a self driving vehicle takes out a family in their mini van or god forbid a school bus the lawyers will descend like locusts. There is going to be much testing, development, prototyping and further proving before acceptance will occur in a majority of the motoring public.
You may have already seen the story just this morning of polling showing 80% of those polled not ready to accept this technology an 64% saying even after adoption human control to be essential.
We will see autonomous control of vehicles, of that I have no doubt. It will just not be fast tracked in a way you propose. Just “keepin’ it real”.
Self driving cars only need to be better than people, they don’t have to be perfect. More importantly they only have to be better than AMERICANS, some of the least trained least competent drivers in the civilized world.
There’s basically ZERO unknown unknowns. Driving is simply object identification. And you don’t really even need to identify most object, just identify how their movement could impact your intended movement. That’s it. We like to pretend it’s the horribly complicated thing, but let’s face it, 14 year old kids drive, not well, but they do.
There is no other reason block protection hasn’t been fully implemented yet. The unions are fighting it, and they’re fighting it in the most effective place: spending. Block protection is an expensive program, and very man hour intensive. By keeping the spending low, and the number of people on the job low, they keep the project from being completed, thus keeping their members in jobs. The system works, and the unions know it, so they fight it.
Automatic driving is being implemented RIGHT NOW. It’s not a “rosy timeline” it’s bothering to pay attention to current events. Google has put MILLION of self driven miles on the books. Dozens of companies are working on various solutions. This article is about a company that makes VIDEO CARDS entering the market. The fact that so many companies see a clear revenue opportunity here is proof that this is happening NOW.
That polling data actually supports me. It shows that 20% of the market is ready to buy a self driving car the minute one hits the market. Desktop PCs didn’t have anywhere near that level of buy in in 1980 and look how that progressed in the following 10 years.
You’re not keeping it real, you’re failing to learn from history. Look at how desktop PCs progressed. Cellphones. Smartphones. VCRs. DVDs. Streaming video. From introduction to the market place to having changed a basic element of society happens very quickly.
.
The best any driver-less car will ever do is imitate the least competent drivers that are already clogging up our roads.
They are a ‘solution’ to a non-existent problem, and will constitute an insoluble problem themselves.
.
6000 years of falling IQs is not a pretty thing to behold.
.
Incorrect on every count. Computers are faster than us, never get tired, never get drunk, never get angry, never get distracted, never get sick. And there is very much an existent problem: 30,000 dead people every year in this country almost entirely because of human error.
And anybody that thinks IQs have been falling for 6000 years is a moron and should stop hassling people.
.
Speed is irrelevant to the problem.
Processors are incapable of original thought, and can only use cataloged responses.
If you are offering these idiot cars as a solution to drugged drivers, better solutions are already being ignored.
They will relentlessly advance the dysfunction of our road system by many orders of magnitude in a short period of time.
.
>> “And anybody that thinks IQs have been falling for 6000 years is a moron...” <<
I offer your posting history.
.
BWAHQAHAHAHAHAHAHA Speed is irrelevant to the problem?!?!? BWAHAHAHAHA Right, because when you’re traveling at 60MPH the least important part of avoiding an accident is reacting quickly. You’re hilarious.
They don’t need original thought, avoiding accidents is conceptually very easy, you change speed and or you change direction. It’s ain’t rocket science.
The cars aren’t idiots, they’re tools. They’re a solution to ALL forms of bad drivers. Including the under trained, the guy who had a fight with his wife, the guy who’s 18 hour hours days, and every other thing that stacks up to make people accidentally kill each other.
No, they won’t. The road systems are fine, it’s the drivers that suck.
How many years have you had that tag? And yet you hang out here. You do nothing but insult people and post patently ridiculous statements, often times appearing drunk, and just generally being useless. Your the one not as intelligent as you’d hoped. You’re worthless, and getting more pathetic with every post. Talk about a bad posting history, most of your posts don’t even make sense.
OK you win, I’m an idiot that knows nothing and has no experience, also I’m a grouch with no vision for the future. As I said before, “see ya in 2020”. If you are right I’ll happily humble myself, as you are so much smarter and more worldly than I, I’m sure you will not lower yourself for the same.
Didn’t say you’re an idiot, said you’re not recognizing how far this disruptive technology has already come, nor the adoption curve on disruptive technology. It moves fast.
I put a lot of work into not being wrong. I started predicting the first mass market self driving car would hit in 2020 a few years ago. And the tech is actually moving faster than I expected.
“Would it stop and ask for directions?”
It would have to argue with itself first and then give itself the silent treatment after it lost.
Public transportation will never be safe in America (and now Europe, too) because of the ferals.
LOL
The only automated aspect of a van I drove a long time ago was an automatic breaking system. It caused me to rear end a car.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.