Google’s self-driving car team prepares to spin out from Alphabet’s X
recode.net
recode.net
This space has become super competitive and it will be fun to watch what happens. I still think it will be many years before we see a fully self-driving car (where no driver is needed at all). I think the best way to do it is to redesign cities to specifically accommodate self-driving cars. Honestly though that might be something other countries such as China and Singapore are much more reticent to pursue.
That will never happen. The car has to drive itself on existing infrastructure or not at all.
Once self driving cars are widespread, changes are certain even if gradual.
Gradual examples include increasing density by replacing parking lots with buildings, relaxing zoning requirements for parking with new residential units, etc
Considering in most cases > 80% of cars are on the road at the same time during rush hours, you can't save much with self-driving cars — not more than with a well-working transit system already.
So you'd need huge parking areas somewhere still.
And the "increade density, reduce parking requirements" — if you do that, you can just improve transit, and get the same advantage.
But that doesn't happen.
People want to own a car, have it always available, and don't want to share.
Self-Driving cars provide no advantage there.
They will be driving around and not be parked..
> Considering in most cases > 80% of cars are on the road at the same time during rush hours, you can't save much with self-driving cars — not more than with a well-working transit system already.
Maybe self driving cars can be the "last/first mile" in transit so that more people actually will use public transit.
> People want to own a car, have it always available, and don't want to share.
I don't want to own a car and no-one that I know of in my surroundings want to own a car. Car companies even do private leasing now because people do not want to own a car.
> Self-Driving cars provide no advantage there.
I think you have to think bigger. Why would you want to own a car if you could call one at any moment with your phone with an app?
They could also go to park in more remote areas, that are uncomfortable for people to walk to, and come back when asked to.
That said, I didn't say we'll have cars driving around with no people in them. I said that cars will be able to drive to remote parking areas by themselves.
So the Park+Ride concept. That’s already possible today, yet completely underutilized. A criticism often brought against transit is "I’d have to switch trains at some point", do you think these people would start using that now?
> They will be driving around and not be parked..
All the time? That’d be quite wasteful, considering most cars are only used for about 2h a day, during the rush hours, for the commute.
> I think you have to think bigger. Why would you want to own a car if you could call one at any moment with your phone with an app?
Why would you want to own a car if you could call a taxi at any moment with your phone with an app?
Why would you want to own a car if busses and trains come every 2-5 minutes at every stop, and you’re never more than 400m from a stop?
Reality is, as European cities show, all these "advantages" can already be gotten with public transit today. I’m not sure what self-driving cars actually improve compared to existing public transit methods. And especially because they’re using space more inefficient than existing transit methods, these advantages would have to be huge to make any tradeoff worth it.
Otherwise you could just invest in better transit, not self-driving cars.
Also, robocab+ride could work on both ends of the commute. Where I live, "office suburbs" tend to have transit connections just as bad as actual suburbs, or worse.
That's a historical accident, not some fundamental of the human condition.
Already young people buy less cars than people their age 1-2 generations ago, and car-pooling is increasingly more popular.
Heck, suddenly a sizable percentage of the population became part-time taxi-drivers with Uber and Lyft.
There seems to also be a cultural element, and I doubt that can be solved with merely technological means.
Making the choice not to own gives you a leg up for other purchases.
Housing, I suspect is bennefitting from similar post-hoc analysis. "Millennials don't want to own homes" which I've seen in headlines should also be weighed against the fact the houses today cost many multiples more than a typical annual salary. I believe when my father was buying his first home the rule of thumb was "roughly 5* annual salary" nowadays in Toronto homes are easily 10*
This stuff (microhousing, ultra-mobility) isn't necessarily about "enlightenment amongst the youth" it's the crushing cost.
Was it about some fundamental inherent desire in the first place, or increased income then too, plus ample advertising and touting it as the "american way"?
Other regions, even as rich, don't place as much emphasis on car ownership.
10k a year only sounds reasonable if you're making payments, and I suspect the amortized cost once you (eventually) sell the vehicle will be much lower.
Granted, this is only my own experience, but considering it's so far off from 10k makes me wonder how accurate that statement is for the general population.
Lots of people are eating a few grand in depreciation each year, plus higher fuel costs, plus higher insurance rates. And don't forget financing costs, licencing, taxes. Many people are clearly eating more on depreciation than you are spending, all in.
10k may be a little on the high side, but 5 or 6k isn't at all (you're already at, what 3 to 3.5 on the cheap end).
So my napkin calculation was roughly $30k for the base auto, financed at $500/month for 5 years. $1000/year on insurance and $50/week on gas.
$500 * 12 = $6000 $50 * 52 = $2600 +$1000
total = $9600
In the above calculation there's $0 spent on service & $0 on financing interest. The theory is that as age goes up, the annual repair cost increases as well.
I think there's also a bit of a confound in the fact that I'm operating in Canadian dollars where my numbers might be inflated... as well, I haven't owned a car myself since somewhere around 2005.
My payment at the time was $450 on a Mazda 5.
All said and done though $5k to $10k is not a wildly unreasonable range. Which I think we were generally agreeing on.
Because cities are expensive infrastructure that takes many decades to change after it has been built. Unless an earthquake or other natural disaster or war takes a city down, change at the level of changing the roads, moving whole neighborhoods, etc is very slow.
They are also incredibly popular (if you take housing prices as a proxy for desirability).
Which is exactly the argument I wanted to make ;)
> It would be wonderful for me to go to bed and wake up 1000 miles away.
Trains do that already :)
Not-so-commercial robot car pools might even start getting viable in rural areas, because some countries mandate a minimum level of public transport coverage and basing a few robocabs in remote villages could be many orders of magnitude cheaper (and certainly more useful) than repeatedly driving an empty bus around just to fulfill a schedule so sparse that only the most desperate will ever bother using it. Another angle that might lead to pockets of rural adoption much earlier than expected: many rural populations already have a lot of experience with pooling agricultural vehicles, this could occasionally be a seed for adopting local pool e.g. to support the elderly and doubling as a fallback for personal vehicles.
Well, tough luck then. We have to make some sacrifices for the environment.
Our ancestors not that long ago, merely 50-70 years or so, walked on foot, for miles, to get to the same grocery store.
Yet a common criticism when that is suggested is that only the poor use public transit, and that everyone in public transit smells bad.
Which is what I’m questioning here: Self-Driving cars provide no real technological advantage over existing transit methods, and the real issues are social problems.
And we can’t solve social problems with technological solutions.
Providing dense parking garages gets a lot easier if it doesn't need to be omfortable for people to get to the cars and you don't need to be able to open the doors when parked etc.
But the overall need for parking will also drop if more cars are shared.
> Considering in most cases > 80% of cars are on the road at the same time during rush hours, you can't save much with self-driving cars — not more than with a well-working transit system already.
But self-driving cars can be used to augment the transit system too. E.g. complement buses with mini-buses and smaller cars to handle overflow. Imagine an Uber-like service with an option to "call a bus" to any number of nearby "bus stops" at a lower price than if you call a car. If a full size bus is within 5 minutes, with space free, it'll make you wait for that, otherwise it'll dispatch a regular car at the same discounted price.
You could do away with normal bus schedules in favour of having the buses follow the busiest stretches at any time of day, and use cheaper/smaller vehicles to take other passengers.
E.g. near me the main bus route is used by two very distinct groups of people who either intend to go quite far, or who intend to go to the nearby train stations or high streets in either direction from me. During certain parts of the day the riders are often dominated by one or the other, and you'd get far more efficient use if you could shuttle people between the train stations/high streets and the longer distances separately. Knowing the start/end of each journey would allow you to "sort" the riders accordingly and dispatch vehicles much closer to optimal.
> And the "increade density, reduce parking requirements" — if you do that, you can just improve transit, and get the same advantage.
See above - automated vehicles will change transit too, but there are lots of situations where people still want to go door to door and are willing to pay accordingly. E.g. I take the bus from the local train station and home most of the time, but if I'm in a rush I might not be able to wait. Sometimes currently that's frustrating as getting a taxi her normally will take me 5 minutes. If I'm in a rush and there's a bus <10 minutes away, there's certainly a non-zero chance that the taxi will be late and the bus fast, and the bus will arrive first and I have to consider if I want to cancel the taxi or wait and be later than if I'd just taken the bus in the first place.
With a "combo service" with guaranteed pickup from specific locations in <5m by either bus/min-bus/car at the same price depending on what's closest, I think you'd see a lot more people here opting for something like that than people who currently use the bus, which could very well have the effect of making it viable to use far more buses.
A lot of the car users I see here (London suburb) mostly use public transport anyway, and have the car as "just in case" that will be increasingly hard to justify with solutions like the above.
Couple that with increasing pressure to reduce the number of allowed parking spaces in new developments further and it will keep getting less desirable to own cars many places.
> People want to own a car, have it always available, and don't want to share.
Do they? A lot of people feel they need a car, but I think far fewer want a car if they felt they had better options. Personally I don't even have a drivers license. Sometimes think about getting one, but it's an annoying time commitment for little benefit, when I can have a car here with a driver in 5 minutes so I can relax and read or play a game while I get to my destination. Ten years ago it was more annoying, but now it's gotten to a point where I don't think I'll get a drivers license in my lifetime (I'm 40), as the inconveniences keep dropping. The one remaining thing that makes me sometimes wish I had a car is for road trips on holidays to places outside of bus routes where it's expensive to hire a driver - self driving cars will take away that inconvenience too.
A lot of people do want to own a car still, but I suspect that car ownership will start trending downwards very quickly once self-driving cars drive the cost of car sharing services down further and make it easier to provide better guaranteed availability and shorter waits and more flexibility (e.g. get whatever type of car you happen to need)
A lot of people also have conflicting wants. E.g. they may actually want to own a car, but want that next holiday more, and so opt out of car ownership once a shared one ends up sufficiently cheaper.
You mean like they do in Manhattan today?
>But self-driving cars can be used to augment the transit system too.
Here's the thing. All these scenarios are possible with the addition of a $10-ish/hr driver today. Yet they don't exist. Maybe no one's made the right pitch to VCs or maybe, just maybe, there really isn't demand for this sort of thing (in the US). Variants of smaller ad hoc minibus services do exist in other countries.
No. You don't want to burden human drivers with, for example, the mental stress of a constantly-shifting route that would be driven by a computer-optimized shared taxi system. You also increase efficiency by 33% when you free up the driver's seat.
I don't know, do they have parking garages with floors where an adult can't stand upright and can't actually get out of the cars? I know you can automate it and have robotic systems to lift cars in place etc., but that adds complexity and cost that will largely disappear if the cars can park in spaces humans can get out of all by themselves.
Maybe the robotic solutions will be the end game, but painting narrower bays for self-parking cars will be the much cheaper short term solution, and e.g. modifications to add messanine floors etc. the medium term solution.
> Here's the thing. All these scenarios are possible with the addition of a $10-ish/hr driver today. Yet they don't exist.
A $10-ish/hr driver is enough to make a big difference in cost. Let's assume some downtime for maintenance. An average 20 hour utilization over 360/days year for a car would add $72k/year in costs.
Now add in that models like Uber's are new and most places aren't anywhere close to optimizing pickup times, and most cab companies are technologically just leaving the stone age, and why should we be surprised that nobody has done exactly this yet?
There are additional differences. E.g. Tesla has indicated they want people to be able to opt-in to having your own vehicle used. The availability of cars can go through the roof without a massive increase in cost once you don't depend on a supply of drivers.
> Variants of smaller ad hoc minibus services do exist in other countries.
Actually variants of ad hoc minibus services do exist in the US too. Shuttle services from airports, for example. But they're nowhere near what I suggested.
To do this for bus lines you run into the same problem as above: Drivers is a big limiting factor. The bus companies near me constantly struggle to get enough drivers (and yes, they raise salaries, but then they'd be unable to survive under the constraints put on them - take away those salaries, and they'd have a lot more flexibility in what they'd be able to do)
My understanding is that there are more sophisticated robotic garages elsewhere (probably in Japan). But attended parking lots with cars up on lifts is very common in Manhattan.
This is an extreme example but you get the idea: http://www.smalltowndjs.com/images/nyc-parking-garages-2598-...
Which easily multiplies cost by whole numbers if you allow multiple years for the hardware to break even.
The fixed hourly cost for the driver is also why cabs are orbiting for hails (unless regulated not to) instead of waiting cold, further increasing cost per use.
With passive minutes basically free, the economic threshold between passive waiting and active orbiting would be considerably shifted, resulting in additional efficiency gains.
Self-driving cars reduce the costs of cars quite substantially: risks go down substantially, expands the pool of people who can use it (young, elderly, drunk), the ability to multitask makes the time a commute takes much less expensive, the ability to call a car is hugely convenient.
And of course the fact that you don't need a driver means commercial driving gets a lot cheaper.
I expect self-driving cars to substantially increase the number of cars on the road.
Rentable cars in most major areas and high-density transit system already fulfill exactly these requirements.
Take any city in Europe. ~15-20% of the people commute with cars.
Do you think these people, who could already use transit, suddenly will stop driving? What advantage do self-driving cars provide to them that transit and rentable cars don’t?
So why do 15-30% of people in these cities still drive? What will self-driving cars provide that transit and rentable cars didn’t yet?
It's likely that even at equal or higher density, self-driving cars will be able to have higher throughput with comparable speed limits.
Initially, though, it might be sufficient to "encourage" them by allowing them in commuter lanes. [this should even be trivial, because it should be pretty easy to find passengers uber-pool style if adding passengers would improve the trip]
We will se first one to two seat fully self-flying vehicles than cars.
Elon Musk's predictions are consistently unreliable.
Somewhere in the middle there's probably a point where self driving is demonstrably safer than human, even before they're "ready" for everywhere.
It's just a question of more training(altough a recent NHTSA? regulation says the planning stage cannot use black box techniques like DNNs).
And once self-driving is a business, i suspect we'll find ways to get said training data for cheap enough and fast enough. For example Tesla/Uber drivers as trainers .
The way I see it (as a software engineer myself), is that each wreck that the AP did not prevent gets added as a test case in the AP unit tests that all have to pass. Also, each time a human wrecks, it could/will send the data to Tesla to continue simulations and ensure AP would not wreck in the same conditions.
How many of the wrecks, even fatal ones, essentially a different driver making the same mistake over and over? Likely millions. Real fleet learning would simply solve that problem once, and prevent thousands of fatalities as a result. As more and more AP capable cars hit the road, this simply will exponentially increase. Thirty years is laughable seeing how much computing has improved in merely the past 20. Maybe 5 tops assuming the legal hurdles are actually surmountable.
http://www.theverge.com/2016/10/19/13341194/tesla-autopilot-...
They also have every single human driver wreck captured with AP capable hardware to make fully autonomous vehicles safer. The way they are doing it right now is brilliant.
Edit: As Elon has previously said, not releasing something that saves lives is actually unethical. If a single life is saved by AP software that would have died if the vehicle was driven by human meatware, not releasing it is the unethical thing.
[1] http://www.cnn.com/2015/11/13/us/google-self-driving-car-pul...
[1] http://fortune.com/2016/12/09/michigan-self-driving-cars/
The basic problem with VOTL is power/weight ratio of current batteries the maximum fly time you get is 15 minutes with no reserve.
The driver can "take over" in some form for parking, though even that may operate in a way that is completely different from today's vehicles. I expect to see a lot of things like this over the next 10 years.
However if the car is not able to find an alternate route, it may require the driver to take over, so it wouldn't be fully autonomous.
In either case, the car has to be able to reliably identify conditions in which it is necessary for the person to take over, which is definitely far away from being solved at a level that would be required to meet consumer levels of safety.
However, there still needs to be a driver who can take over which means it can't function as an autonomous taxi.
Are you sure "reticent" is the word you want here?
https://en.wiktionary.org/wiki/reticent : "Hesitant or not wanting to take some action; reluctant"
Add a hard firmware reset button as well.
- If the hack/virus was already installed, a network shutoff won't help much.
- If a car is 10 years old with 10 years of safety updates/patches, resetting to the original firmware might not be a viable solution. But there needs to be a failsafe way of knowing the running software has not been tampered with.
Master battery disconnect which auto locks all brakes
> - If a car is 10 years old with 10 years of safety updates/patches, resetting to the original firmware might not be a viable solution. But there needs to be a failsafe way of knowing the running software has not been tampered with.
Plug and play ROM modules?
probably want to separate those. Kill the engine at 80 mph, you can coast to stop. If you lock the brakes at 80 mph, you immediately lose most of your control, and, since you killed the battery, you lost traction control and probably ABS. And you risk blowing the tires.
Rather than a firmware 'reset', a 'firmware safe mode' control.
This loads a super simple and infrequently updated image that lets the car just be a dumb car, and nothing else.
Also, I guess one would press this 'I want to take control' button in an emergency. That means the persons suddenly finding themselves driving this vehicle have to be above-average drivers.
As for the inter-vehicle communication, if it proves to be effective in improving security, I guess it will eventually be mandatory or at least "encouraged" by making your insurance more expensive if you disable them.
Assuming that devices don't have cryptographic issues, IoT security is still a mess. So yes firmwares are probably full of BO vulnerabilities. As for more recent techniques, I'm aware of some of them but I'm no expert so I can't comment properly :)
In addition, societal issues of unemployed drivers aside, are the economics of having human drivers really such an impediment to shipping goods economically today? At least in the US, rail is mostly used for long distance shipments which is effectively automated--to a first approximation.
Highway driving takes its toll on the driver. You're usually on a single lane, maintain distance from car before and follow speed limit.
Just lane following and adaptive cruise control and make this journey a lot less tiring.
(And even the sort of features you can buy today in luxury cars are pretty nice from what I see.)
I imagine automating limited access highway travel is much easier than automating the "final mile". I actually can see this sort of thing happening well before cars are automated to be honest.
Have one person remotely control 100 self driving trucks simultanously
Most of the automation efforts revolve around "drive on the highway from point A to point B" scenario. If you have a container that's that simple, rail is much more cost-effective.
Trucking has fairly complex loading/unloading requirements, with scenarios such as "haul 8 vehicles, unload them at 8 different addresses", "haul a load of tomatoes and lettuce, do 5 drops at In'n'Out locations, 1 at Safeway, 1 at Albertsons" fairly common. There's an understandable lack of trust in the system, where the trucker is also responsible for ensuring the staff on the recipient side is not unloading (i.e. stealing) too much or helping themselves into another customer's load.
There's also "stop and wait your turn at weight stations", "stop to refuel" as well as "stop at random checkpoints when crossing state borders" scenarios that seem to require just way too much of human involvement.
I agree any sort of driver assistance tech at level 3 and below is beneficial for public safety, I am not sure it's an industry game changer.
You're describing a traveling salesman problem - which there are plenty of working solutions (even if we don't know the best solution yet) And then an autonomous forklift problem - there are a few companies already making them. And packaging + placement can simplify any issues if they exist. Maybe throw in a robotic arm in the back of the truck for good measure.
Autonomously refilling a tank of gas could be a fun problem to solve. But full service stations already exist. Those should be able to stand in until there is a better solution, such as Tesla's battery swap stations but for trucks.
Now, without a human driver the trucks can be working 24/7. No need for sleep. Which might be worth it even if a human accompanies the truck and just sleeps in a back section. Also who wants to pay a person to drive a truck, when the truck can drive itself for free?
This is already in-progress and isn't that transformative. To most people the delivery or long-distance truck drivers pretty much are invisible driver less machines.
(i asked this in a previous submission https://news.ycombinator.com/item?id=13130017 , asking again as the topic hit the front page)
Maybe this is related to the new laws in Michigan that allow self driving cars - http://fortune.com/2016/12/09/michigan-self-driving-cars/ . Now they have a potential market - doesn't matter how large, they count on it that other states/countries will follow.
- What a "Orwellian" dream life we gonna have. Cannot wait google to go on medicine also. Then we all can use pills from google. But what will happend to your car, your medicine and your home, if you dare to be opinionated away from mainstream ? - what a nightmare!
Google has been known to lock users out of all their services if they are believed to have violated the ToS of even one service.
Having self-driving cars from Google integrated into the same ecosystem could have serious repercussions for users, so is this wise? Do we need legal changes, or should Google just change their model for locking out users?
Forcing companies to serve users isn't really an option either, but Taxis have to do that — how can we solve these issues in a realistic way?
It is then considered the "property" of the university, which is sold as IPR to the highest bidder to make money for the university. It is also summarily locked behind academic paywalls but there are some solutions to that (icanhazpdf , scihub).
We just fail on everything else in that sentence: make results free, and increase competition.
Things might change if those laws were revoked, turning public universities back to the public, and then funding them appropriately. 30% tax funding is not "funding"... Let the privates do what they want.
Break them up, and force the patents to be openly available for a nominal fee (like the patent on the transistor once was). Refuse to license a self-driving car for street use if a centralized firm, rather than the car's owner, can decide how it's used.
https://backchannel.com/self-driving-cars-will-improve-our-c...
Tesla and other self-driving developers are working without the luxury of a worldwide mapping database, and while not perfect, their products more or less could work on any road, because they're based on input from the sensors. But Google cars are dependent on input from Google servers, which means, where that data is lacking, the cars can't drive.
If they're looking to run some sort of business to start making money, I could maybe see like an Uber-like service across very small geographic areas they have mapped. But the narrow fence around the area you can practically drive with one makes it useless to sell as a consumer vehicle.
As of late 2014[0], Google engineers admitted the mapping required for their cars to work wasn't feasible at a nationwide scale, and that if they didn't know to look for say, a stoplight, on their maps, they'd run right through it. (Looking for arbitrary red lights is probably a hard problem too.)
That's two years ago, yes, but they've announced nothing, technologically, in this field in this time (please comment if you have an additional source relating to the mapping of the roadway for a Google car), despite tons of PR about the number of miles they've driven or what cities they were driving in, with the same technology they've already had. Also bear in mind, almost all of Google's senior talent in self-driving cars has left Google X. So if they were having trouble with this problem in 2014, their newer, less experienced people now, probably are having more difficulty in 2016.
[0]http://www.slate.com/articles/technology/technology/2014/10/...
[1] https://en.wikipedia.org/wiki/Traffic_sign_recognition [2] http://blog.caranddriver.com/how-high-definition-maps-are-pl...
But there are a number of reasons the need pre-mapping are not a huge issue. A) When autonomous cars start to be introduced people will still know how to drive, and can pilot the car through unmapped areas. By the time everyone has an autonomous car, and no one knows how to drive the world will be mapped. B) Google has already shown with Streetview that they can create detailed, up-to-date maps of most roads. C) This problem will likely be solved in time.
I also don't think precisely mapping each road is going to be a difficult problem if you enlist a huge network of cars contributing data. The problem is highly parallelizable.