Slashdot Mirror
Alphabet Is Finally Taking the Driver Out of Some of Its Driverless Cars (recode.net)
An anonymous reader shares a report: After almost a decade, Google's parent company Alphabet is getting closer to fulfilling its promise of rolling out cars that can take anyone anywhere without a driver behind the wheel. Alphabet's self-driving car company, Waymo, is introducing truly driverless cars to public roads for the first time, the company's CEO John Krafcik announced today at the Web Summit conference. That means there won't have to be a person sitting in the driver's seat, waiting to take over, and that the car's computer system will complete all parts of the driving task -- though for now, only in some of the company's cars in Phoenix, Ariz. While this move is still geographically limited, it marks the beginnings of Alphabet's driverless future finally becoming a reality. No other company has succeeded in operating a fleet of fully driverless cars on public roads.
I've seen other posts on Slashdot before that were dubious we'd see self driving cars in the next 20 years... but it's not even going to be five before they are in use with real people in all sorts of areas, as this article indicates.
There is just too much demand, too much benefit, and SO much effort being put into making self driving cars work. People seem concerned these cars may make mistakes but the benefits are so huge mistakes will be overlooked, because in the end even now they are probably safer than most human drivers, much less after a few more years of effort.
The largest obstacle I see really is how to deal with snow, which can really block up pretty much any kind of sensor. Otherwise the technology to drive correctly has advanced and will continue to advance at a very rapid clip...
"There is more worth loving than we have strength to love." - Brian Jay Stanley
that driving for a living is going away soon but you'd be amazed how many don't believe that. It's gonna be like when computers decimated junior accountants but without all the new jobs working on computers.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
Quoting from the article:
That means there won't have to be a person sitting in the driver's seat, waiting to take over, and that the car's computer system will complete all parts of the driving task -- though for now, only in some of the company's cars in Phoenix, Ariz.
Now: Phoenix, Arizona. Probably one of the driest spots in the USA, and one with nice, straight roads. Hmmm... Is it possible that the Waymo / Alphabet / Googleplex cars are not that good at self-driving?
I mean this seriously: the more I think about it, and the harder it is for me to take the idea of a self-driving car seriously in anything that is not in the southwestern United States.
A self driving car in some parts of Europe would simply be very, very difficult: anyone who has navigated the beautiful little streets of, say, Granada in Spain knows what I am talking about (hint: very narrow). Anyone who has driven in Norway, or any other country in Scandinavia, knows that the weather can be grueling there (Alaska or North Dakota, some parts of Illinois or Wisconsin also come to mind).
All of this to say, a decade into this slef-driving car project, has Waymo been blowing smoke all along? Is the self-driving car vaporware? Discuss.
The right to offend is far more important than the right not to be offended. (Rowan Atkinson)
Self driving cars that can handle the roads in my area may just be impossible.
Wrong, they are driven by an amoeba brain, just like regular.
...there will be at least ONE employee in the vehicle at all times. More hype.
When I can't buy a 6-speed.
Buses are filling up with folks from Los Angeles, fresh from their classes on Hood Jumping and Curb Tripping. This message brought to you by Dewey, Cheatham, and Howe.
I can't wait until they teach them to make left turns at 200mph.
Then the lower-cost solution is obviously to fix the roads first, then recoup the costs by eliminating the driver expenses after self-driving cars are in operation.
Ezekiel 23:20
This will never happen because self driving cars are impossible, right luddites?
As opposed to the future of cars being driven by post-millennial twerps who can't take their attention off their cell phone for two seconds?
Then the lower-cost solution is obviously to fix the roads first...
If the solution was that easy, the state wouldn't constantly be asking for more money to fix the roads.
An enigma, wrapped in a riddle, shrouded in bacon and cheese
liability both civil and criminal need to be worked out.
Maybe they can hide under eula and other bs + drag out trails so that victims take a low settlement to get on top of there mounting bills.
But in a criminal cases if they try NDA's / destroying evidence it can get very bad for them.
now there may be a push to have doctored evidence to cover for poor code and other system faults.
Who says the technology has advanced at a rapid clip?
I've spent some of my spare time studying self driving car tech (including the Udacity course). It's not like I'm going to be building my own self driving car anytime soon, but I've learned enough to see that self driving car tech is very real and not hype. There are a LOT of prototype cars on real roads today from a large number of companies,, not just in California, not just Google, but others cities and companies now as well. We are very close to the cars operating in limited areas with no employees - think trucks on highways or cars to transport people in retirement communities.
There has been and continues to be a huge volume of data gathered from current sensors, especially Tesla, that makes getting closer to true self driving tech much more rapid than it might be otherwise. Like I said, at this point self driving car tech can handle most situations better than human drivers - as long as teh sensors are not blocked.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Illinois fails if they can just do the speed limit
Let's see. Didn't a self driving Tesla recently fail to determine that a tractor/trailer was making a wide turn around a corner it front of it and thereby chopped the Tesla occupant's head off?
Then the lower-cost solution is obviously to fix the roads first, then recoup the costs by eliminating the driver expenses after self-driving cars are in operation.
Exactly how would that be lower cost? Lower cost than what?
What expense are you eliminating or recouping? This isn't a closed circuit race track where eliminating the driver could save you the expense of the driver. *I* still need to get somewhere, so I'll be in the car either way. The driverless cars are LESS efficient (the computational power to operate them is quite significant at this time), so there is actually a LOSS of money on energy/gas per mile.
If you're hoping to save money on parking, then excuse my language, but fuck you for even thinking of it (not necessarily "you" as in the parent, but whomever is thinking this, cause it always comes up). The options here all suck:
* have it drive around all day without you instead of paying the high price to park downtown? Great... more traffic, more gas/energy used, more miles driven (rubber/roads wasted), all so you can avoid paying for the most efficient option of just plopping it into a spot. Oh, and if it's electric, how the fuck is it going to recharge? Oh yeah, park it and let it charge. This is a dead end.
* have it drive home and come back to pick you up? Again, more traffic (fully doubling the time and miles spent on the road). If you're trying to avoid parking downtown, this is surely going to lead to congestion pricing and/or normal tolls, plus it'll use more gas/energy/etc.
* have it car share and go pick up other people? We already have great solutions for this that serve us better - taxi's/uber/lyft/etc. Sure, self driving can assist in those industries, but let's not pretend this is going to be how everyone will use their cars.
You would/might be correct that it could save significant money by being far safer and dramatically reducing accidents. However, that equally applies to assisted driving cars. Any safety related savings that would come from driverless cars could (and will) also be present in the assisted driving cars.
This post may be a bit harsh, but I am honestly curious if there is some real benefit I'm missing. I've seen a bunch of stories/threads on this, and the reasons I've ran into all seem really dumb (such as those above). Driverless might be cool, and I'd enjoy kicking back while my car takes me somewhere safely, but I don't see the point in baseless claims like those above, unless those are all just trolls. Am I wrong?
The word 'finally' indicates an editorial opinion. You're entitled to have one, dear msmash (what a surprise), but your really ought not place it the articles, let along in the headline. When you do, you come off as trying to tell your readers what to think. That's unprofessional.
Right. Because your roads are special snowflakes.
Look, I know you all believe everyone are rational actors, behaving rationally, and that bad stuff doesn't happen.
But it does.
There are people who will use automatic driverless vehicles. In ways you never "intended".
"Oh, but it will be hackproof."
Yeah, I heard that one before.
"Oh, but it will have a police kill switch."
Heard that one before too.
You're all very very very naive.
-- Tigger warning: This post may contain tiggers! --
If you live in the places they're 'testing' this, there's going to be accidents, and maybe injuries and deaths.
Just because Chicago drivers are morons that thinks taking Lake Shore Dr at 55+ mph is a smart idea doesn't make it a smart idea or Chicago drivers any less stupid.
LSD at 55+ is on thing yes 60 is pushing that. But 55 on the tollways is a joke.
You probably think "AI" is right around the corner
If you don't realize it's already here in a big way, I think you need to research modern AI systems.
But not AC (Artificial Consciousness) which is what a lot of people think of when they say AI.
along with a Mars rocket.
Not quite as soon as Musk thinks but probably not more than six years behind his states schedule. Do you truly disagree? It's not had to see at this point how he gets to there from where we are now.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Only 60mph?
Hell...down here we do that through neighborhoods with small children.....
[scoffs]
ON the Hwy I keep it usually pegged between 90-100mph if I'm not in too big a hurry.
Light travels faster than sound. This is why some people appear bright until you hear them speak.........
Does an autonomous car slow down to let another car into traffic
Yes, even a Tesla autopilot (the most conservative system on the road today by far) can see that is happening today, currently it tells the driver to engage but it could slow down itself as it does for cars braking. The more advanced systems already recognize and handle this because they are making constant predictions about what cars around them will do and accommodating those possibilities.
There are tons of situations where humans together decide how to interact
And in all of those situations you have a TON of jerks on the road that do not obey those unwritten (or even written) rules, so again no worse than a human driver - with the potential to be better because a computer can be more patient than most people would be.
The autonomous vehicle needs to be able to work with other drivers just as a human driver does all the time
Why do you use the word "needs" here when humans do not do that today and they are fully licensed drivers.
It only "needs" to be mildly better than humans in some respects to be better, and self driving car tech is already there in terms of behavior around other humans driving or walking or biking.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
You've apparently misread my comment.
Ezekiel 23:20
Yes there's liability issues to work out.
However the thing about self driving cars that insurance companies will LOVE is the amount of data.
Right now maybe you can work out what really happened from a black box, but you are generally getting things like accelerometer data and car inputs (steering/gas/brakes).
With a self driving car you are getting a vast amount of data from well before any accident actually happens, so it will be much easier to determine fault.
Furthermore insurance companies will love the lack of variability of self driving cars. Even the best drivers may have days where they are sleepy or sick or whatever, that affect the chances of being in an accident.
With a self driving car, the probabilities are much easier to predict because the cars reactions are more predictable and frankly better than a humans in all sorts of ways - they don't get mad, they would be more cautious (again in predictable manner), they can really see 360 where a human cannot, they can react faster than humans to very sudden issues and also anticipate issues better based on subtle behaviors of drivers nearby that many humans will not pay attention to.
So to insure the self driving car will be much easier on insurance companies - there will be accidents to be sure, but the insurance companies can have a much better handle on the actual risk they are insuring.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Exactly how would that be lower cost? Lower cost than what?
Better roads plus less expensive AI hardware times a few millions could potentially be cheaper than bad roads plus significantly more expensive and power-hungry AI hardware times a few millions. (We often simplify robots by tweaking their environment, notably in factories, for example.) And taken further, if "self driving cars that can handle the roads in [his] area may just be impossible" is taken as a hard assumption, then the cost differential is not a one-time hardware expense but the operating cost of the human driver.
Ezekiel 23:20
From the article: "Alphabet's self-driving car company, Waymo, is introducing truly driverless cars to public roads for the first time, the company's CEO John Krafcik announced today at the Web Summit conference."
Alphabet better scrounge up "waymo" money once those self-driving cars get loose and the legal actions begin to roll in!
Circle the wagons and fire inward. Entropy increases without bounds.
If you're driving a car, you can't do anything else productive (at least safely or legally). Self-driving cars would free up most of the time you spend commuting for getting work done or entertainment, either of which would be worth significant money.
A group of cooperating self-driving cars could probably use about half of the parking space as human-driven cars by arranging themselves as densely as possible:
You could probably have one driving aisle for every four rows of cars; cars could move themselves as needed to free up access to packed-in cars.
Cars could pack within inches of each other because there would be no need to open doors, and cars could sort themselves by similar sizes.
I once had to open my door and follow the white line on the road outside the door
People who have driven in the mountains or in other real blizzards know the deal that I am talking about when I say snow is the biggest issue.
It doesn't matter how slow you go, or what kind of wipers or cleaning fluids you may have, sometimes the snow just overwhelms everything. Like you said, I've been in some situations where the only thing that could be done was driving with my head out the car to see anything at all, either the road side or reflectors.
Although a GPS could tell you when you had gone generally super wrong ("You are now on a lake") I sure wouldn't trust it to keep me on a road, there are places in Utah where weird signal issues have causes the GPS to think I was about 500 felt off the road when I was on it. In a snowstorm the already weak GPS signals get a ton of interference. Maybe if cities have radio transmitters in each streetlight you could at least keep between them.
There probably is some clearing tech that could keep sensors having visibility, but after seeing an inch or two of ice coating my whole car in some situations, I am unsure how this can be solved. I think for a while yet truly series weather will simply halt self driving car traffic (to affect trucking most). Maybe this is where human truckers make a living for a while, getting self-driving trucks through major snowstorms.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
Automated cars can, after you arrive at a destination, buzz away to park in a nearby garage rather than right in front of your stop. When they let you off at work, they can head for the recharge plugs in a distant corner of the parking area to spend the day.
If you're in the car going somewhere then why not drive it yourself?
I love driving. I've driven across the US a few times, have done straight 19 hour shifts before. I've driven in other countries, driven offload.
But there are a lot of times where I could be doing something better with my time. If nothing else with a self driving car I could enjoy the view around me a lot more instead of focusing so much on the road.
It also opens up a vast amount of possibilities for side trips. You could be looking ahead to see what fun things might lie just a bit off the interstate to stop and have a look at, instead of flying past them.
It also would also vastly increase the range of day trips for most people. Like I said I've done a 19 hour drive before, but I sure would not do that now - I would have to stop and spend the night in the middle. I would once again not mind a 19 hour trip in a car to get to someplace in a day, if I wasn't having to drive most of it and pay attention the whole time.
It also makes longer commutes feasible. For a while I was doing 30-40 minutes each way, and a know a lot of people do much longer commutes - it was driving me kind of crazy just from the daily annoyance of the whole thing. But if I could read or work during that time, I wouldn't mind a longer commute at all - and because I could work while in transit, I could arrive at work later and leave earlier to essentially lose no time at all.
And that is the fundamental reason why I really am excited about self driving cars, because of how much time and mobility it gives back to everyone. It's why even though I love driving I don't mind ceding control, because the benefits so vastly outweigh the loss of some fun I have currently when driving. I would probably opt to be able to take full manual control at times, but at this point I don't see it as a deal-breaker if I cannot.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
With Uber facing numerous issues trying to get places like London to allow its style of contractor-based taxis, I have a feeling this same taxi industry that beat down Uber will also restrict driverless taxis as they are a threat to their livelihoods and careers...
There are many roads where driverless cars are going to be a long time coming. Single track roads with a few passing places where cars going on the opposite direction can pass. When cars meet one has to reverse, on one road I have driven with a cliff wall on one side and a drop on the other. Roads where there is a mix of rickshaws, cycles, motor scooters and cars, and if you wait for reasonable clearance you would never move. Roads with broken surface and mud, where you have to zigzag to avoid holes. Ice-covered passes where the correct move is to turn back and find a different route. Bring able to drive on phoenix doesn't mean it will cope with any road.
Will their Wall Street market crash too?
Don't let reality cloud your vision of a perfect future! That'll scare the investors and it doesn't make for very good headlines.
So then how is the operating system supposed to interact with the devices?
"fix the roads" may involve destroying a beloved landform like a mountain, or bulldozing a historic site, or taking land / traffic right through people's yards and possibly houses. Consider why a lot of cities have some really narrow roads, or why roads wrap around a mountain instead of plowing through, etc. Transportation, to me, isn't the most important thing in the world. Would I rather folks take a sharp turn and spend an extra 7 seconds than demolish a waterfall and reroute the stream which is irrigating thousands of acres of farmland and trees downstream? Of course.
No, Tesla autopilot does not do what I described, and Tesla will likely tell you you should take over in those situations.
Hi, that's exactly what I said in my post. Would you kindly read more carefully?
It doesn't "likely" do that i those situations, as I described that IS WHAT IT DOES DO. So it already knows the situation is there, the software just doesn't want to have to decide what to do about it and asks you to take over so you can make the choice.
They's be judged by their most annoying interactions with humans.
Just like BMW's are today, but people still buy and drive BWM's. People know what to expect.
Do they autonomously drive faster than the speed limit?
Yes of course, to match traffic speeds.
Will they choose to stop back from an intersection so an oncoming left turner can take it faster?
Far more often than a human since they can predict how an oncoming car might need to turn t make a corner in a way most humans cannot.
Will they move over enough for a right on red driver to eek by on the right?
Much sooner than a human would since they can see to the rear far better than any human.
Many times a rule is broken at the benefit of all involved.
The advanced self driving cars systems of today take ALL of those things you mentioned along with millions of things you have not into account already.
Its far from simple.
I never claimed it was simple. I said the demand was vast and so too are the resources being applied to solve these challenging issues, to the point where the more advanced software is better than most humans in the kind of tricky interactions you mentioned, and in the long run no human can compete with cars that have vastly better visibility. If I could drive in a wonder-woman style all transparent car I sure would choose to do so - that's what it's like for a self driving car all the time.
There are way too many situations that they will not be able to deal with human interaction as well as humans.
They can and all have been broken down. You are all worried about an infinite number of interactions but it more boils down to, where can I position my car so that I'm not hitting anything and nothing is going to hit me. How can I help other drivers do what they are trying to do based on the information I can obtain watching them. That's all a great simplification of course but it's not an insoluble problem; if it were humans couldn't do it, and plenty of humans that cannot think very well are driving cars today.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
"they can head for the recharge plugs" ...which means employing someone to wait back there all day plugging/unplugging the cars.
Have the alphabet folks really, really dog food. require their loved ones to be driven first: mom and dad, children, nieces nephews ... (I left spouses off)
Wherever they might live. Show me that, for 50% of alphabet employees,. And I'm a believer.
You do understand that bookkeeper was how folks who couldn't afford higher education got a start in something besides ditch digging. And I hope you also understand that the world does not, in fact, need ditch diggers. We have machines for that and they do the work of a thousand ditch diggers for less then the cost to feed them just enough to keep digging. Next you're gonna say something about pulling yourself up by your bootstraps, somehow completely impervious to the irony of using a phrase that is literally psychically impossible to describe succeeding solely through ones own efforts...
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
Automated plugging in is a pretty simple problem to solve compared to automated driving. Can always use inductive charging too.
This space intentionally left blank
Yep, all historic towns should be levelled to make SDCs easier to implement. Who needs all those medieval churches and castles and walls and stuff anyway.
I don't think that *buildings* have anything to do with this. (You don't drive through buildings, do you?)
Ezekiel 23:20
I'm stumped by your choice of examples. I was thinking of such things as machine-readable road markings or communication infrastructure for vehicles, and you suddenly jump to demolishing houses and leveling mountains...why?
Ezekiel 23:20
End of the Automotive era. Longtime GM executive thinks the car industry as we know it is pretty much doomed.
This could be be pessimistic. Or optimistic. Depends on how you feel about driving. I used to like it. But with all the distracted drivers and vile traffic in my area I am ready to kiss it goodbye. Forty thousand dead on US roads. Robots couldn't do worse. Heck. Trained bears couldn't do worse.
"No fear. No envy. No meanness." Liam Clancy
NDAs wouldn't cover the other car and its passengers . . .
It's a matter of working it out and allocating it, but it is "merely" that.
Once the accident rate/severity can be brought significantly below the of human drivers, the total cost of vehicle production and liability becomes less than the total cost of human vehicle and regular liability.
The only "real' issue here is whether to allocate the liability to the owner or the manufacturer--and even so, the total costs remains the same.
One possible solution is to make the manufacturer liable for the first twenty or twenty-five years, and the owner after that. This lets the "insurance" be built into the initial cost, and it can be financed as part of the car.
Wherever the liability is placed, the bottom line for the consumer/driver is going to be about the same.
The one allocation that will *not* work is the current system of liability for negligence by drivers and "product defect" by manufacturers; every accident would see expensive litigation over this. Liability really needs to be placed on one or the other (or in a fixed percentage).
hawk, attorney and displaced economics professor
p.s. Yes, this is an application of Coase' theorem, with the litigation costs being the cost of negotiation
So it's not lower cost at all. You're suggesting that the roads need fixed in order for self driving cars to work, period. To recap, the conversation went:
> > > > Self driving cars that can handle the roads in my area may just be impossible.
> > > Then the lower-cost solution is obviously to fix the roads first, then recoup the costs by eliminating the driver expenses after self-driving cars are in operation.
> > Exactly how would that be lower cost? Lower cost than what?
> (you said something to the effect that better roads meas cheaper AI cars than would be needed if roads were awful (and they may not work at all on awful roads), and that could potentially save money overall)
AFAICT, you have no idea if it'd be cheaper to fix all roads first, or to leave shitty roads and focus on the cars ability to deal with shitty roads.... and neither of those deals with it being cheaper than just continuing to have human drivers.
Is that right? IE. that you're "lower-cost" has nothing to do with a comparison to existing human driven cars?
But that (having the car park itself once you get to work) doesn't save money overall. It might save people in cities who drive into the city a few minutes, but it's not going to save anything for most people. The car is still less efficient, and if it needs a fancy garage with automated plugging or inductive charging, that's gotta cost most than a gravel lot.
Again, sure, that's cool, and we all like cool things, but why try to sell it as more than what it is?
Can always use inductive charging too.
That's a nice way to recoup the green energy savings by using an electric vehicle... throw them away on lossy inductive charging!