Google to Make Driverless Cars an Alphabet Company in 2016
bloomberg.com
bloomberg.com
It'll merely spare consumers from a trillion plus dollars - every decade - in wasted costs on buying cars and then hardly ever utilizing them. While likely saving a quarter of a million lives every decade, and further sparing consumers a large percentage of the vast associated medical, insurance and lost-asset costs that go with those injuries and deaths.
The 'dumb pipe' of the Internet is going to eat the dumb car - as it'll try to do with everything it can touch - sucking out the complexity and inefficiencies with smart, hyper-available software.
A few decades from now, people will consider buying a 'dumb' $30,000 gasoline engine car to stick in your driveway and rarely use, to be comically absurd.
I would take the hacker culture of Google or Apple over drunk brosefs driving from one bar to another on a Saturday night every time.
In fact I hope for the day that driving on public roads would become illegal by humans.
My big worry is that we'll have autonomous-car-AI monoculture. If that happens, we'd lose the advantage of the "compensation" other drivers in traffic can do. We could end up seeing car-behavioral "prions": glitches in car software that cause a strange behavior, which in turn triggers the same glitch in the cars attempting to compensate for that behavior. You could see, for example, every car getting into a state where it thinks it needs to swerve "more right than other cars" to avoid an emergency, and thus an entire "pack" of cars driving off the road and into a fast clockwise-torquing circle.
Or, less eyecatching but more fatal in the near-term, you could see a monoculture-wide decision that works alright for single driverless cars, but actively harms drivered cars (and other driverless cars with other-model AIs) when the majority of a "pack" of cars is from the monoculture. For example, I could see a "pack" of driverless cars all deciding to jam on the brakes at once (because, say, a bunch of spike-strips suddenly popped out of the road in between them or somesuch)—leaving the one drivered car boxed in by them to keep going and crash into the vanguard of the pack.
Sure there will be some accidents, but nowhere near the same frequency as today. Autonomous cars are much better/faster at "compensating" than humans are.
A by-product of this might be an airline level of safety and after the accident investigation. There certainly would be the financial benefit for it. Once the autonomous car comes it will be safer than humans to start with and only become safer and safer with time. It might get to the point where we regularly travel at 90+ mph on the highway with barely a care in the world.
No, but they have enough smart people, and engineers of every stripe that if your life did depend on it, they'd probably make it work. Google's driver-less cars seem to be doing pretty good at the moment.
Plus, when you have more driver-less cars on the roads, they can communicate with each other to adjust to similar speeds based on congestion, accelerate/decelerate in the most efficient manner (so as to not slow down the flow of traffic), and in the case of road hazards, they'd communicate together and avoid them.
Based on most of the drivers I see on the roads, I'd absolutely feel safer if most of them didn't control their vehicles, and I'd definitely trust a lot of smart people developing algorithms more than the average driver...
On top of that: "It'll merely spare consumers from a trillion plus dollars" -- seems a bit of an exaggeration, does it not? Transportation will still be needed, and it's going to have to be paid for somehow or another. If it's not directly towards paying for a car it's in taxes for infrastructure, or it's to companies who have to invest the capital and make a profit on it. It changes how the money is spent, I guess, and the hope is that it's more efficiently spent, but I suspect that the costs will just shift around and not necessarily save "trillions of dollars every decade".
Right now, many people have cars sitting idle for a large portion of the day because they need them for an hour a day, or sometimes an hour every couple of days, or a couple of hours a week. While it's true that there are surges in demand in the normal morning and evening commutes, there is still a lot of excess driving capacity that is needed for trips other than those surges, and so you only have to have enough cars to cover those peaks.
In addition, there may be other people who could take public transit, walk, or bike to work, but don't because there are other things that they need to do at various points that require a car, and so it's more convenient to just always have it available.
So yeah, driverless cars can only get you so far, you're right that the peak demand number is pretty much the minimum number you could have. But if I go an look at my residential street during the day, when commuters are presumably at work, I still see about half of the cars present. Those are all cars that wouldn't be necessary if everyone were riding in self driving cars that could rearrange themselves based on demand. That reduction in number of cars would also reduce the amount of space needed to store all of those cars, which could help bring everything closer together making it more walkable, bikeable, or efficient for public transit, which could further reduce the demand for cars.
Do we really have excess capacity compared to the peak demand? Put another way, how many people own a car and aren't using it at e.g. 8:50AM? At a complete guess I'd say maybe 5%?
(Also, given zoning and city layout, most houses are a lot more than 5 minutes away from most offices. Even if my commute is 8:20-8:50 and yours is 9:15-9:45, we're probably both heading into the city centre from outside, so the same car probably can't serve us both).
If you are willing to share the ride, you can have the same car serve you and someone else who shares a similar route and is willing to travel 10 minutes early (or late - can be incentivize by a discount).
Self-driving cars can make car-pooling easier - it can be a hyperlocal public transport system. Pick up 4 people who are traveling from/to within a 2-mile radius with each other, and peak traffic will be 25% of what it is now (best case). It's even simpler considering that most people commute the same route everyday.
What does the self-driving car change in that case? Wouldn't that be just the same as ordinary carpooling that we can do today?
* ability to add more vehicles on demand
* automation
For carpooling with my own car, that's really too much work to try and hunt down someone, figure out how cost sharing is going to work, ensure that our schedules are in sync often enough for it to work out, and so on.
With an automated system, it can just adapt to changing demands. Maybe some days schedules are too far out of sync, so it just sends dedicated cars. Maybe some days there's someone better to pool with.
You can already do this with systems like UberPool, but with self driving cars, you don't have to pay a dedicated driver, so it becomes even cheaper and more convenient.
As I said, when I look out at my residential street at 8:50, I see about half of the cars still parked there, or maybe a third (very rough estimates, I haven't actually counted). Now, I do live in a city with good public transit options, that's also fairly walkable and bikable. But at least where I live, there's a pretty serious oversupply of capacity, just because people do need cars occasionally and the options for not owning a car are not good enough.
In fact, on the next street over, parking goes way up at around this time since I'm near a major subway stop, so a lot of people park here and take the subway the rest of the way in. Rather than just sitting there, going back up and shuttling more people to the subway would be a more efficient use of cars that you could do with self driving cars. The commutes from house to subway station are likely under 30 minutes, and since rush hour generally actually lasts about two hours from 7 AM to 9 AM, you could fit a couple of round trips in that time period.
> (Also, given zoning and city layout, most houses are a lot more than 5 minutes away from most offices. Even if my commute is 8:20-8:50 and yours is 9:15-9:45, we're probably both heading into the city centre from outside, so the same car probably can't serve us both).
This is increasingly untrue. Mixed use developments, companies moving out to the suburbs for cheaper real estate, people moving back in to the city, and so on mean that reverse commutes are increasingly common.
Well part of the idea here is that self-driving taxis can make public transport a lot more feasible.
The reason people usually can't combine driving and public transport at the moment is that there isn't enough parking. Once you have self-driving taxis, so people don't need to park, it's perfectly feasible to take a car ride to your nearest bus stop.
Plus, once people are able to do this, it makes public transport a lot more viable as a business. Instead of needing to make end-to-end networks, they can focus on improving efficiency on popular routes.
Econ 101: Incentives matter.
And because they're fun to drive.
Google owns maps, video/photo database of almost all roads (streetview) and push for autonomous cars, yet they still didnt even start using all this data to build 3d map of the world a la Photosynth.
Last time I checked Warsaw(capital of EU country) in google earth there was maybe 20 biggest buildings, all hand modelled. Actually even calling them modelled is too much, those were pretty much boxes of approximate size.
1) What do these cars do in ambiguous driving environments: construction zones, poorly/under relined lanes, freshly paved roads, hazards, dark/rainy roads where camera images are useless, etc.
2) How do these cars handle system failure while driving? Blown tire, engine failure, etc.
3) Is there "Moral Decision Engine" code in current generation cars? E.g. person runs out in front of an autonomous car, the car chooses to spare the person by veering off and hitting some inanimate object. I can imagine a class of accidents in traditional hands-on cars where a driver hurts himself in order to avoid harming a pedestrian. Is such moral logic hard wired into autonomous cars?
2) Engine failure is not a big deal, it just means coast to a stop on the side of the road with hazard lights on. Blowouts are a little trickier. But what about unusual stuff like hitting a deer? A wheel falling off? Or another car drifting into the side of our car on the expressway? Those are tricky but not truly rare. This is an area where implementors will need to compete yet purchasers will have a very incomplete view of the relative performance of cars on the market (it's not a single number like HP).
3) Hitting a pole instead of a person is an easy choice. Yet one that many humans would not be able to make quickly enough. I think autonomous cars will outperform human drivers in "moral" decision making quite early on.
1) Expecting a human to be on deck within some fraction of a second in challenging times will not fly.
2) Perhaps software in cars can be independently safety tested. Just like cars are collision tested today.
3) Understand about human reaction times. My question is about making moral choices: hit a pedestrian or sacrifice the driver?
If it was a case of hitting a pedestrian versus driving off a cliff I feel like the pedestrian will just be SOL for getting in that situation. Otherwise you could have murder by car relatively easily.
Maybe we could make people register their car settings at the DMV like we do for organ donations. "Would you like to opt-in to the murder-children AI driver, or the martyr driver?"
I expect auto insurance companies will have the final decision on this one. They'll provide lower insurance rates for cars that make the correct decisions minimizing total insurance payouts (whatever those happen to be). With lower insurance costs and therefore a lower effective price, sales of that particular brand of self-driving software should increase. Competitors will then rush to do the same.
I would assume self-driving cars would be programmed similarly.
>Miles driven since start of project in 2009: “Autonomous mode” means the software is driving the vehicle, and test drivers are not touching the manual controls. “Manual mode” means the test drivers are driving the car.
> Autonomous mode: 1,320,755 miles
> Manual mode: 955,771 miles
So it seems the cars are self-driving only half of the time! The answer to 1) seems to be driver takes over all the time.
It may be comparable to landing a plane automatically without visibility, as is commonplace nowadays...
This, at least, isn't an issue...
And introduce more bureaucracy as now Google Car has to negotiate with Google for access to AI tech, computers, maps etc instead of just using whatever they can?
What have they been doing over the last 3 years? I even tried reading their blog, but they only report on building more prototypes or driving more miles. It's a blackout of actual information about their challenges and achievements. When will they be confident enough to launch it, or at least talk about it?
[1] "California: Self-driving cars must have driver behind wheel under DMV proposed rules" (16 December 2015)
http://www.mercurynews.com/business/ci_29262037/california-s...
[2] "Google 'disappointed' by proposed restrictions on driverless cars" (16 December 2015)
http://www.usatoday.com/story/tech/news/2015/12/16/google-di...
Why is this a problem unique to self-driving cars? Is it that it "won't" runover the counterparty? Have you ever even considered the idea of running over a thief? I suspect our legal infrastructure would not look kindly upon you in most jurisdictions.
Kinda sounds like you're inventing problems here and pretending human solutions exist (again, how will any truck run away in the first place?)
I'd imagine a car will hundreds of cameras and sensors in it would be better equipped to deal with crime anyways.
I've seen plenty around constructing some trolley-problem style moral argument about deciding who to kill in a crash. I don't know why it's expected that they'd have to solve, in a fraction of a second, a problem that people have been arguing over for years. The human reaction will probably be at the level of "jerk wheel quickly".
Stick with what works now! upgrade to the new Hermes 3000 typewriter