Ask HN: Do you think self-driving cars are feasable at all?
There are hardly self-driving trains. No self-driving planes, or ships (they always have a watching human at hand, who also performs any non trivial operations). Self-driving is not there yet for every other transportation options which have way better procedure standardization and more advanced technology (with the exception of rockets perhaps).
How is it realistic to think society will embrace self-driving cars in 10-15 years and that they will prove safe if we haven't taken that step with aviation or most subway systems?
(
30 comments
| 8 points
| on HN
)
On the other hand, no self-respecting country boy is going to want to give up the freedom and independence that he gets from driving his own vehicle. Separating yourself from the herd is a value proposition, and self-driving cars can't compete with it.
What's left?
1. In some cases suburbanites might use self-driving cars to get from the suburbs to the edge of the car-free downtown zone. It's possible the technology would work for them, but, do any of them perceive an unmet need? Is there anybody out there really frustrated by having to look at the road during their commute? I would think there's a lot more incentive to find ways to shorten the commute, than to simply make it hands free.
2. Interstate trucking is a possible use case, if people would tolerate it. You'd need a plurality of states to get on board with the idea of a very potentially deadly machine going on autopilot, and you'd need to sort out who's liable when people are killed by it. And you need to figure out the "last mile" -- who drives the truck after it takes the exit off the highway. There's a real shortage of truck drivers in the world, so this is probably the one market where there's a commercial incentive to develop the tech, though.
And why do the low hanging fruit first when we have the technology and manpower to go for the juicer stuff?
A self-driving car is a canonical example of lots of software, sensors and actuators working seamlessly and robustly at all times.
The recent issues with VW, Jeep, Toyota clearly demonstrates that car manufacturers can't even get human controlled cars to be 100% trustworthy. What chance is there that they'll be any better with a self-driving car?
If a self-driving car is better than a human who is drunk, eating, and texting all at the same time, then it's already a safety improvement.
I disagree with this. There are many routine driving situations where if someone else does something stupid, you're probably dead. Given how many cars are on the roads, I'm amazed that there are as few accidents as there are. (I submit as anecdotal evidence: if you're reading this, then you're not dead.)
Indeed, it is relatively uncommon for people to limit or reduce their driving on account of the risk of an accident. Accidents do happen, but for practical purposes most people round down the risk to zero.
That said, I agree with your main point. If humnas are 99.99% trustworthy and automatic cars are 99.995% so, then it seems clear that we should allow them on the roads.
I don't think they cope that well. If you're under 35, a car accident is your main worry (other than suicide):
Motor vehicle crashes are a leading cause of death in the U.S. More than 33,000 people died from motor vehicle crashes in 2013 alone
http://www.cdc.gov/injury/wisqars/overview/key_data.html
http://www.ons.gov.uk/ons/rel/vsob1/mortality-statistics--de...
The thing is, can you imagine sitting on a self-driving car minding your business in a busy road shared with human drivers?
Since not all cars will be automated, the success depends on convincing everyone that they're still safer in a world dominated by all the other shitty drivers.
I read somewhere that Google cars accidents had always been caused by the other person. Question is, would a human have "seen it coming"? Can a human judge that someone is driving agressively, or shyly better than a computer? Can a human infer from a slow-driving car that it is probably trying to park despite not using the blinkers? And which such informations, make better decisions to avoid others being imprudent?
Still, probably most accidents are to blame on the drivers themselves and their own mistakes and self-driving cars would fix that, but what probably is really going to count if you want to sell such car is probably the "sense of security" you can get rather than the hard stats.
That's why I don't see fully self-driving cars being adopted by anyone anytime soon as some pretend, and it is much more realistic to think of forms of "assisted driving" (where you still have to drive and pay attention) and still need driving controls in the vehicle.
Self-driving cars will succeed, but only when relegated to dedicated, communal, self-driving equivalents of HOV lanes. In a controlled environment, when it's only necessary to deal with exceptions, self-driving cars will rule. Outside the HOV lane, self-driving cars will be more like GPS-Assist++. Human required, but the broadstrokes navagation off-loaded to the computer. The last mile problem of figuring out where to park, last minute adjustments, etc., will still be human-led.
However I think we're going to see more and more 'driver assistance' like Tesla's autopilot mode where the driver remains responsible for the car. This kind of tech will get more reliable and trickle down to cheaper models, to the point where most highway / motorway driving is done in semi-autonomous mode. When this happens (and accident rates go down), people and governments will be much more ready to embrace fully autonomous cars.
Has anyone here experienced the Tesla autopilot? There are some fun videos on YouTube. It seems impressive but still a little unreliable.
How is it materially different to give ownership of that to your local municipality (who already knows where you live and probably how much water/electricity you use)? Municipal government is probably the one you have the most control over.
For example, cargo vehicles not carrying any passengers or a driver would represent a huge cost savings in logistics.
There will be some class of autonomous vehicles, perhaps remotely supervised, regularly on the road within a decade. They may not be consumer passenger vehicles, but they don't need to be to make a huge impact. To focus on just whether your car will dive itself is static thinking - what new opportunities could autonomous transport create?
Public adoption & trust is another story.
http://www.autoblog.com/2015/11/04/october-2015-the-better-t...
Almost all of London's tube network is self-driving. The human operator controls opening and closing of the doors and that's about it.
And many others, but not so many compared to the whole list: https://en.wikipedia.org/wiki/List_of_metro_systems
At least, I'd like to think this is a success, but it's taken how many years to get here, and it's a closed system with so many variables under control compared to a car.
People had the same qualms about automated elevators when those were new.
Trains and elevators are different - they run on tracks/guides and do not have any unforeseeable obstacles, operating conditions thrown at them. Furthermore, there are several mechanical safety interlocks that will stop any out of control situations. Planes and cars have to operate under conditions with far more uncontrollable variables.
I don't think it's really true that they "do not have any unforeseen obstacles". A lot of random shit can happen on a train line, even when it's got limited access lines like the underground driverless trains linking the terminals at Hong Kong airport. I think its more that they "do not have any possible choices except go, and (try to) stop".
If a Train is "automatic" it has three tasks: start when loaded. stop at pre-determined locations. try to stop if something is detected in the way.
As everyone else is saying they need to be much more thoroughly tested and approved.