• Welcome! The TrekBBS is the number one place to chat about Star Trek with like-minded fans.
    If you are not already a member then please register an account and join in the discussion!

Driverless Hype Collides With Merciless Reality

Gary7

Vice Admiral
Admiral
Driverless Hype Collides With Merciless Reality

Insatiable need for tech progress sometimes pushes things sooner than they're ready. I do agree with the article. With roads as they are and the precedent of established driving law enforcement (a bit too lax these days), driverless cars are faced with too many challenges. It's all about the unpredictable nature of the human driver. If everyone obeys the laws, everything is fine. But as we know, from the huge success of auto insurance companies and repair shops, a notable portion of people screw up. And then there's road rage. Drivers who make highly unpredictable moves that driverless cars have a tough time managing.

I do think there's merit to the concept... only I think we have to wait a bit longer for it to be more practical. The first phase would be highway driving. Because highway roads are much straighter and more predictable than local roads, with just a few minor enhancements driverless cars could use them with little incident. The only problem is the unpredictable driver who recklessly speeds or cuts off. There would need to be some kind of deterrent to that kind of driving. Introduce the AMA - Auto Monitoring Assistant. For any driverless car, it's by nature keeping track of traffic around the vehicle. If it detects highly unusual or dangerous movement of other vehicles, it captures that info and is immediately relayed to the state authority. Once people who drive like that start to get caught, they'll realize the days of reckless driving with impunity are over.

For the local level, we need to achieve a point where more vehicles have automatic accident avoidance technology: a) quick braking to avoid front collision, and b) warning signal to drivers too close in the rear for the speed. As this becomes more widespread, local road traffic becomes more predictable. At some point, all vehicles would need to have accident avoidance systems, much like cars must have airbags. And then automated driving has a better chance of being a success.
 
There has to be a Massless Drive that is possible.

Think about a dream that is based on the intact of data that is converted into the massless dream. From mass of reality the mass is converted into the massless reality of the dream.
 
. At some point, all vehicles would need to have accident avoidance systems, much like cars must have airbags.

All new cars, but you'll never get them in all the cars on the road, at least not for a century. Why do I think this? My daily driver doesn't have airbags or ABS. Hell, I'm not even legally required to have seat belts, and it's perfectly legal because it was built in 1962.
 
All new cars, but you'll never get them in all the cars on the road, at least not for a century. Why do I think this? My daily driver doesn't have airbags or ABS. Hell, I'm not even legally required to have seat belts, and it's perfectly legal because it was built in 1962.
Of course, not all cars will. There is the 25 year law, whereby cars older than that are exempt from all kinds of requirements. So it'll have to be some kind of majority percentage.

I hope your daily driving commute isn't very far or in dense traffic. So many lives have been saved by airbags.
 
Recent research done by Dutch insurance companies have shown that the more so called driving assistants are fitted in the car the more likely it is that the vehicle will be in an accident.. the reason is A: all that crap IS NOT WORKING as well as it should and B: people's are even bigger morons in cars with all that crap because they think the car will magically save them.
 
Color me surprised by all this...

Being in the computer industry, this is exactly the outcome I was expecting. It's going to take at least one (more likely several) generations of improvements in processors, sensors and programming before anything like a fully self-driving car is even remotely practical - let alone safe.
 
Color me surprised by all this...

Being in the computer industry, this is exactly the outcome I was expecting. It's going to take at least one (more likely several) generations of improvements in processors, sensors and programming before anything like a fully self-driving car is even remotely practical - let alone safe.
And to be honest, there needs to be some governmental oversight at some point in the near future of all AI systems.
We have the ability to order a system, in this case a multi-ton easily weaponized land vehicle to break the law and drive at whatever speed we feel sufficient, subject to safeguards put in place by the manufacturer.

who needs a crazed militant behind the wheel of a truck killing innocents when the same person, or organization behind that person can order the vehicle to do what they want, and have whoever was behind the wheel leave before momentum kicks in. Sure there are failsafes that could prevent that, but leaving that protection entirely up to individual manufacturers is sketchy and naive. At best AI systems should be forced to obey the laws barring harm to living occupants at all times. Since many of the legislators in my own country think the world is 6000 years old and take scientists more as a threat than a boon, I doubt anything much will be done about it, until of course, they feel threatened and overreact, as usual.
 
I'm 100% convinced that driverless cars will eventually become a reality. What I can't say is whether it will take 10 years or 50.

Driverless cars are just the first stage. Once the cars can self-drive, you move on with transponders that talk to an AI control system so traffic can be deconflicted to a large degree. What will follow that is 'bee-hive' style AI in the cars themselves, which will coordinate as a hive mind to reduce congestion and keep things flowing more smoothly. When we get to the point where vehicles are driving themselves more than humans are, getting a driver's license will become more difficult- the idea will be to LET the computers do the driving, and a human wanting to drive in that mix will require special training and certification akin to a CDL nowadays.

It's coming. It's just a matter of how long it all takes.
 
Interesting would be to discuss who or what would be responsible in the case of accidents. Owner? Shotgun rider? Manufacturer? the Machine itself? Do all road accidents then become industrial accidents?
 
I know! Replace roads with tracks. Have one lead car, and have all other cars attached to it and each other. Set aside locations where people can board and exit cars, and create a schedule of arrivals and destinations at key places along the route. That should enhance safety and mitigate accident liability.
 
The earlier real-world solution was to have the roadway include embedded guidance signals to help cars stay in their lanes and inform them of upcoming entrances/exits etc.

The cost to overhaul all of the roads in the US alone would be off the charts... Worldwide? Are ya nuts?!?
 
I'm having the opposite read on the whole situation. Yeah, driverless car may not be coming as fast as some people expected, but that's hardly the failure the title implies. On the contrary, driverless car technology is still as promising as it was before, and to me is grown only more promising as time has passed.

It's obviously not ready, which is exemplified by how there still isn't a single fully autonomous solution that's unrestricted by road conditions in some form. However, aside from the obviously reckless attempt by Uber to push a driverless killing machine (ugh, fuck Uber), the experimental driverless cars have performed well. So well in fact I'm often asking if the humans are ready to be driving cars on the roads where the AI does fine, because so far it seems to have been performing better than us. (Although that's subject to more independent evaluation.)

I believe we're close to having a rollout of fully autonomous vehicles soon, and they would be driving safer than we do. I don't know how soon ‘soon’ is – as an alleged hype-person I'm giving it a couple of years, but it may be a decade or two. This can happen really fast or really slow, and you can't really tell from the publicly known issues the autonomous systems are facing. With software, seemingly enormous problems can be rather easy to fix, and sometimes seemingly trivial ones may need an enormous amount of work. If the car has a tendency to make blatantly dangerous moves when overtaking, that may be fixed in a couple of lines of code, but if it experience a minor confusion during some specific lighting conditions, that may be next to impossible to pin down, and require full retraining of the models or worse. And we can't have a guessing game on the specific there, but I'm still betting on the shorter timescale.

Part of what gives me assurance is that even if these cars are performing much worse than reported, and the people developing them are hiding some huge problems with them, those are things that end up on a list of problems that get resolved for good one and by one, and sooner or later we're getting there regardless.

I believe that even Tesla's approach, which I initially judged to reckless and a weird dangerous middle ground between autonomous and human-driven, may deliver an autonomous vehicle not much later than they are bragging they will. And they are second to Uber in my disappointment. They will be behind Waymo on that one, and if they aren't, that would be kind of reckless, IMO


There are three generic problems that I'm worried about:
1. Even with the promising technology, corporations' disregard for safety is troubling: Both Uber and Tesla have gone down this road, with Tesla being saved by the technology not messing up as bad as it could have. Actually, the fact that things aren't much worse with companies trying to rush this to the streets sooner than it is ready is a testament to how promising autonomous vehicles are, but it still leaves me worried about corporate decisions in the future.

2. These cars would sooner or later get remotely exploited, and people will take control over them. Self-driving capability in itself doesn't mean someone can remotely get in, but current solutions are getting updates remotely, and future solutions will have those updates mandated by law to ensure the cars are driving with the safest driving software version. Those remote channels can and will be exploited to inject malicious code alongside, or in the place of, the self-driving one, and scenarios that were until recently only the subject of science-fiction will start happening. Someone sitting in the nearby woods would lock you in your car and drive you somewhere.

3. Cars will be vulnerable to road vandalism. Even if they can be made better than humans in average current road conditions, if a human is actively trying to play the AI, the human will win. So placing peculiar objects in peculiar configurations on the road, or defacing road signs may make those cars go crazy. (Now I'm suddenly disappointed Kirk's solution to the Kobayashi Maru didn't involve tricking the AI playing the Klingons, and instead involved injecting subroutines. Then again, the simulation was probably simpler than that and didn't have an AI.)

Those I'm much more worried about than the self-driving cars not working in themselves.
 
I think the biggest step forward will be automated trucking.

Trucks will drive slower, but they'll also not need any rest breaks. They'll get better fuel economy. And they'll have depots nearby a highway exit where they can keep local road driving to a bare minimum. Then there will be a new class of hitch hikers that jump aboard any of these automated trucks to get a free ride. ;)

I absolutely agree that driverless cars are the future, but right now companies are pushing the technology faster than they should. There's a certain technological impatience... to get ahead for greater profit.

I still think the next step towards driverless cars will be the inclusion of automatic braking and adaptive cruise control systems that become standard on every new car. Much like what we've seen happen with airbags to supplement safety belts.

Another thing... we do have to "tame" the populace on their driving habits. There are too many reckless drivers out there (in varying degrees). What will force better compliance will be added cost to the drivers. Discounted rates are now available to drivers who will accept a module plugged into their car that captures and communicates driving habits to the insurance agency. Right now it's fairly primitive, only detecting rapid stops and starts and measuring time vs. distance (to see if you're speeding frequently). But it'll eventually get more sophisticated and tap into distance measurements from other vehicles relative to speed--meaning, it'll know if you're tailgating. The US NHTSA will eventually have to mandate that the sensors used for adaptive cruise control will serve as a warning system to manual drivers of unsafe distances. You get people to keep safe distances from each other and several things happen: 1) fewer accidents, 2) less stressed drivers, and 3) better gas mileage.
 
I don't think we're that far off from driverless cars with a lower crash rate than human drivers.

Problem is, 'Lower than humans' is still not zero, and every crash that does happen will be seen as indictment on the entire idea.
 
Driverless cars are totally and utterly useless, they can't even detect motorcycles so they're not safe on the highway and even more rubbish in the city, they can't handle dense trafic, can't detect mopeds/scooters, they can't handle groups of pedestrians, @Gary7 those driving assistance sensors are crap, like I said before, insurance companies have concluded that the more driver assistance stuff is fitted the more likely the vehicle is to have an accident, these toys don't work reliable at all and humans who own such a car behave like even bigger assholes because they assume the "magic" will protect them...:cardie:
 
I’m terrified by a vehicle going 70mph that doesn’t have human judgement. It scares the crap out of me.

For instance, when there’s an aggressive driver pulling up next to me on the highway super fast, I can almost sense that they’re gonna cut me off—so I slow down a little. I don’t think that’s an instinct that computers will really be able to grasp, but it definitely saved me a couple times.
 
Recent research done by Dutch insurance companies have shown that the more so called driving assistants are fitted in the car the more likely it is that the vehicle will be in an accident.. the reason is A: all that crap IS NOT WORKING as well as it should and B: people's are even bigger morons in cars with all that crap because they think the car will magically save them.
I agree with this where adaptive driving is concerned, because some fool will set cruise control and relax... not remain keenly aware of traffic expecting the car will take care of sudden traffic changes. However, I do think the emergency braking feature is a good idea. You just have to forget it is there.
TribbleFeeder said:
For instance, when there’s an aggressive driver pulling up next to me on the highway super fast, I can almost sense that they’re gonna cut me off—so I slow down a little. I don’t think that’s an instinct that computers will really be able to grasp, but it definitely saved me a couple times.
You're so right. That kind of "data" is something a car's automated system at this point cannot factor. And your choice was spot-on. Let aggressive drivers move on so they're far from you. The hope is that major roads are under surveillance and police would be able to observe reckless drivers and intercept them. But... it doesn't appear to be done consistently, as the rise in aggressive drivers is obvious. People drive recklessly, partially enabled by cars that are so precise in handling and powerful in acceleration compared to cars from decades ago, and until they have an accident they think their driving style is OK, putting so many lives at risk.
 
I wonder what is better, a human watching over and correcting a computer or a computer watching over and correcting a human, I am leaning towards the latter as long the computer is actually smart enough to know what is happening I guess that is where it goes wrong.
 
If you are not already a member then please register an account and join in the discussion!

Sign up / Register


Back
Top