I just had the distinct displeasure of driving for two hours on west virginia’s interstate highways in the pouring rain in a rental car with windshield wipers that barely worked. It was an incredibly stressful experience; I drove 40 MPH most of the way as I was being passed at 70 MPH by huge trucks and small cars alike.
I was definitely not at “100%” on that journey. If someone had run across the road in front of me, I’d definitely have killed them.
A system, even an imperfect one, to guide my driving would have been incredibly helpful and would no doubt made the highway much safer.
It seems clear that a database of changes to roads should be a government function. It’s not like I can just put up a stop sign wherever I feel like it without local government approval.
Are local governments in your part of the world flush with cash and renowned for their infallibility? Because in my neck of the woods, councils would be the last piece of governmental bureaucracy I’d want being responsible for whether my car (and everyone else’s) ran a red light!
Something a few people reading this article seem to miss is the new or temporary traffic signal is just an example of how difficult keeping up to date mapping gets. There are many more hazards that can pop up easily without anyone having any chance to map them, government or otherwise.
They can’t be expected to see a new stop light, but -can- see a kid, cat, pothole, etc.? Just getting out on the road with workable collision avoidance they’d be able to “see” new lights and signs. The problem is knowing it IS a light or sign— and that (it seems to me) could be taken care of by equipping the emergency lights and temporary stop signs with transmitters that the cars could detect that tell them there is a new traffic-control object ahead to watch for? Or “slow speed” condition ahead? I would think there would need to be something like that in place anyway–an emergency way to say “detour” or “bridge out”. Yeah, the transmitter could fail, but just by having that system in place you’d limit the potential problem by a huge margin and probably bring it well within the accident level of human drivers in such a condition. Goddammit, there HAS to be. I’ve been counting on a Google chauffeur to get me to the Bingo parlor in my old age.
Don’t worry about liability. We’ll cross that bridge when we come to it. I think there are much more difficult ethical issues involved than the question of who needs to pay for damages.
Who gets to decide on the risk tradeoffs inherently involved in a dangerous activity like driving?
An unexpected obstacle appears on the road. You were going at an “acceptable” speed.
For humans, that just means you couldn’t forsee it and you were going the same speed any sane judge would have gone in the same situation. That child appeared out of nowhere and it wasn’t your fault.
For a machine driver, that means that someone had to decide that the given parameter setting would cause about 1000 children to be run over worldwide every year (which is probably a big improvement), and reducing that number further was not worth the delays and customer dissatisfaction caused by it.
Politicians have to make the same tradeoffs today, but today they legitimately don’t know what they are doing. They don’t know if a speed limit will save any additional lives. Once a few million Google cars are in service, Google will have no excuse not to do data mining and A/B testing and know how exactly what tradeoff between dead children and people late for work they are making.
Should it be legal to set your car to a higher-risk driving style than Google decided? Is it moral?
Should an automated car swerve to avoid the child and thereby risk hitting oncoming traffic instead? Does the passenger of the car have a say in this? Do the passengers of the oncoming car?
I’ve commuted by bike on two continents, and I never really had a problem. After all, at least during the day, hand signals are actually more obvious than indicator lights.
But to be fair, I kept to a part of North America where they actually already had one or two bike lanes. I felt like all the other people using them where immigrants or exchange students from Europe, but motorists had already been introduced to the idea that there might be cyclists on the road.
Airplane autopilots are the closest thing to self-driving cars we have. The air traffic system is orders of magnitude simpler that road traffic–no pedestrians, no hydroplaning, minutes of separation between vehicles instead of seconds, smart crash-avoidance beacons, etc. etc. ad inf.
For all of that we still don’t have an airplane autopilot that doesn’t go “Shit! I Dunno! Your Airplane!” whenever it gets inputs it can’t resolve.
That’s bad enough in airliners, when pilots who get single-digit hours a year of “stick and rudder” time suddenly have to resolve a plane in crisis. It’s pretty clear this is causing crashes.
In cars? Where even the primitive “lane following” and “smart cruise” features we have today convince drivers to leave the controls entirely and climb into the back seat? Any car that claims to be “smart” needs to be able to do without the driver altogether, in all circumstances.
Even the much simpler problem of automating the far-more-regulated transportation system isn’t anywhere near that threshold. I’ll remain skeptical of self-driving cars until truly self-flying planes become a reality.
Agreed! The article makes some interesting points I hadn’t considered before, but between the headline and slant of the article it feels like Slate is looking to get a few extra clicks by going after google.
In NYC the use blinkers is nothing but a sign of weakness…I think we can get machines to not only understand but to actually respect signals a lot better than humans. Perhaps bikes need singalling blinkers, or attaching a signalling light to your hand. I think the extra effort will be worth it in the long run.
I have an iPhone app that can see and recognize a foreign language and translate it on the fly. I just saw another one that can recognize mathematical equations among text and solve them. My camera recognizes which part of the picture is a face and knows to focus on it.
So how hard can it be for a program to recognize street lights and signage? I understand it’s not easy in and of itself, but when we have those similar technologies so well developed, it doesn’t seem like much of a stretch.
Automated cars are highlighting just how slap-dash our transportation system is, just as it is highlighting how horrible of drivers we are. The added requirements that come with automation, I think, will actually strike people as more of a “holy shit, we don’t already do that? how are we not all dead?!?” rather than, “sorry automated cars, you are just too high maintenance.”
The car in automated car is just one piece of the puzzle. The rest of the system has to meet it half way, and I think we can do that. We need to think of automated cars as pat of an expansion of public transit, not an extension of human-driven cars. They still need to be in a system where the system gives back, just in a more flexible way than a train.
In rural areas, switch off the self-driving until the systems have caught up. If your car doesn’t get sufficient “input” from the surrounding infrastructure, it quietly pulls over to the side of the road, gently wakes you up, and tells you to take over.
There were some lengthy articles about the complexity of the mapping some months ago. The sensors required to create the maps are more advanced than the ones standard google mapping cars use, so each car can’t be expected to carry them. I think lydar is involved?
Reading those articles and now this makes me all the more pessimistic about driverless cars on the open road. Maybe instead, they’ll be hybrid vehicles that switch to ‘drive assist’ when you are on a ‘smart road’ that communicates with your car and is closed to pedestrian and bike traffic… Ideally this would be a freeway where everyone has the same driverless system.
I think another issue embedded in these stories is the assumption that the problems of road inefficiency will be solved by private users investing in new tech. It’s clear that it would be incredibly useful for all the driverless cars to be openly sharing information with each other- destination, velocity, ect. This would ease the burden that any one vehicle has to bear… Now, at a certain point this just begins to look like a massively individualized (and inefficient) public transit system… Why not instead focus on creating an automated public transit system rather than the individualistic one-car-per-person model? I talking minority report cars.
I imagine that whatever tech comes out of googles endeavors can be put to this use eventually. But it just seems like they’re starting a drawing and focusing entirely on the eyeball, when there’s a whole body that need to be fleshed out. I guess that’s how you solve some problems though.
