Autonomous Vehicles

[Hoofhearted]

I was reading an article around autonomous vehicles, and couldn't help but wonder why the potential for this is? I know there's been problems with it so far (believe Uber had a fatal accident), but I can't help but think of the possibilities.

According to one recent study, "By 2030, within 10 years of regulatory approval of fully autonomous vehicles, 95% of all U.S. passenger miles will be served by transport-as-a-service (TaaS) providers who will own and operate fleets of autonomous electric vehicles providing passengers with higher levels of service, faster rides and vastly increased safety at a cost up to 10 times cheaper than today’s individually owned (IO) vehicles."

There's 5 levels of autonomous vehicles



How long do you think it will be before we get to these levels? What are some of your biggest concerns? Personally, I think we're still quite a ways away. I'll say another 20 years before we can get to level 4. My biggest concern, being a certified ethical hacker, is the security around it. Having the possibility for someone (or something) be able to control millions and millions of cars is a scary, scary thought.

What do you think?

[Hoofhearted]

There's an excellent overview of it as well as the internet in general on Netflix. it's called "Lo and behold". Very good documentary.

[fredtoast]

It is coming. That is for sure.

[CKwi88]

I believe that you either have to have roads that are 100% autonomous vehicles or none. And it will take a long time for the general population to give up their vehicles.

I would take autonomous cars in a heartbeat. The technology needs to be perfected, but I have little reason to doubt that autonomous cars will be safer in the long run. 

Maybe it's the fact I have an hour plus commute to work (I must sound like a whiney New Englander to some of our board members in larger states) but I would love being able to just hop in my car, press a button, go to sleep, and wake up at work. Or do just about anything other than driving during that hour.

[Hoofhearted]

(04-18-2018, 04:09 PM)fredtoast Wrote: It is coming. That is for sure.

What problems do you foresee there being and how long before they become level 4 or 5?

[Hoofhearted]

(04-18-2018, 04:38 PM)CKwi88 Wrote: I believe that you either have to have roads that are 100% autonomous vehicles or none. And it will take a long time for the general population to give up their vehicles.

I would take autonomous cars in a heartbeat. The technology needs to be perfected, but I have little reason to doubt that autonomous cars will be safer in the long run. 

Maybe it's the fact I have an hour plus commute to work (I must sound like a whiney New Englander to some of our board members in larger states) but I would love being able to just hop in my car, press a button, go to sleep, and wake up at work. Or do just about anything other than driving during that hour.

I don’t think there’s any way at all they’re able to make roads just for autonomous vehicles. It would take an absolute unfathomable amount of work to do that. I’m guessing they focus on making this technology more advanced vs all that hard labor for making new roads.

I love the fact these vehicles are AI. So they talk with all the other cars about potholes, traffic, other humans, etc. I’m starting to try and wrap my head around the repercussions of this... no more mechanics, taxi drivers, semi truck... basically the entire transportation industry will be turned upside down

[Hoofhearted]

(04-18-2018, 04:38 PM)CKwi88 Wrote: I believe that you either have to have roads that are 100% autonomous vehicles or none. And it will take a long time for the general population to give up their vehicles.

I would take autonomous cars in a heartbeat. The technology needs to be perfected, but I have little reason to doubt that autonomous cars will be safer in the long run. 

Maybe it's the fact I have an hour plus commute to work (I must sound like a whiney New Englander to some of our board members in larger states) but I would love being able to just hop in my car, press a button, go to sleep, and wake up at work. Or do just about anything other than driving during that hour.

Thinking they probably are wanting to build that sort of infrastructure with the hyper speed transportation I’ve read they’re working on.

[CKwi88]

(04-18-2018, 05:15 PM)Hoofhearted Wrote: Thinking they probably are wanting to build that sort of infrastructure with the hyper speed transportation I’ve read they’re working on.

I have to say, after going on Japan's rail system I was left with the question of "why the hell don't we have these things stateside?"

[jfkbengals]

(04-18-2018, 05:37 PM)CKwi88 Wrote: I have to say, after going on Japan's rail system I was left with the question of "why the hell don't we have these things stateside?"

I have family who has said the same thing about France's bullet train.

The problem is not the lack of government wanting to do it, it's the people crying "not in my back yard!".

There is a high speed rail project here in the Mid-Atlantic that is on hold because of a town 15 minutes from me fighting it coming through.  This town was LITERALLY BUILT FOR THE RAILROADS!


As for autonomous vehicles, I agree with the sentiment that they will need their own roads free from human intervention.  Reason being, humans are too unpredictable in their reactions and will screw it up for the automated vehicles...

[*Go Cards*]

It is coming but still feel it is a long way off from being perfectted and accepted

[Au165]

(04-18-2018, 08:53 PM)jfkbengals Wrote: As for autonomous vehicles, I agree with the sentiment that they will need their own roads free from human intervention.  Reason being, humans are too unpredictable in their reactions and will screw it up for the automated vehicles...

A computer with high speed processing powers and 20+ sensors is more likely to react correctly to humans being unpredictable than other humans are for a variety of reasons. Computer don't fatigue as humans do, they don't get distracted, they have more information to make decisions with at any given time. Let's say you are coming over a hill at 45 MPH and there is something in the road. At 45 MPH you'll need about 196 feet to stop if you make the decision when you see it, however a computer would only need roughly 130 feet to stop if the correct course of action is in fact to stop. 

Let's change the scenario and say the correct course of action is to swerve because the object is coming at you now. A human within a second has to decide left or right and how hard to swerve to miss the object as well as maintain control without hitting anyone else. The issue is there is no time to check both left and right of you and to calculate this, so people's general reaction is to swerve to the right. An autonomous car with it's complete 360 degree sensing network can know all obstacles around it and then choose the most favorable action based on having all info infinitely quicker than a human could compile it.

I think mixing autonomous cars into our current roadway system as soon as possible is the best possible solution. By mixing in rolling computers that can make better, more calculated, decisions while avoiding fatigue and distractions that lead to most vehicular accidents I think we can see the number of roadway fatalities drop quickly.

[Hoofhearted]

(04-19-2018, 08:27 AM)Au165 Wrote: A computer with high speed processing powers and 20+ sensors is more likely to react correctly to humans being unpredictable than other humans are for a variety of reasons. Computer don't fatigue as humans do, they don't get distracted, they have more information to make decisions with at any given time. Let's say you are coming over a hill at 45 MPH and there is something in the road. At 45 MPH you'll need about 196 feet to stop if you make the decision when you see it, however a computer would only need roughly 130 feet to stop if the correct course of action is in fact to stop. 

Let's change the scenario and say the correct course of action is to swerve because the object is coming at you now. A human within a second has to decide left or right and how hard to swerve to miss the object as well as maintain control without hitting anyone else. The issue is there is no time to check both left and right of you and to calculate this, so people's general reaction is to swerve to the right. An autonomous car with it's complete 360 degree sensing network can know all obstacles around it and then choose the most favorable action based on having all info infinitely quicker than a human could compile it.

I think mixing autonomous cars into our current roadway system as soon as possible is the best possible solution. By mixing in rolling computers that can make better, more calculated, decisions while avoiding fatigue and distractions that lead to most vehicular accidents I think we can see the number of roadway fatalities drop quickly.

I what this technologys ability to value human life over anything else. If it's going down a side street, does it (or can it) know the difference between human life and, say, debri flying down the road? Maybe instead of new infrastructure, anyone who buys this car gets a chip implanted in them lol

[Au165]

(04-19-2018, 08:57 AM)Hoofhearted Wrote: I what this technologys ability to value human life over anything else. If it's going down a side street, does it (or can it) know the difference between human life and, say, debri flying down the road? Maybe instead of new infrastructure, anyone who buys this car gets a chip implanted in them lol

Yes it can recognize human over things. Standard home PIR sensors can tell the difference, these vehicles are using super high end versions along with LIDAR and other sensors. Like everything they aren't perfect yet, and some are better than others, but everything on the road currently tests at "as good or better" than average drivers on the road today. Obviously things like the Trolley Problem are ethical questions that humans struggle with and would have to decide within a second who to take out, but we can't ask AI to solve ethical dilemma's that even we have no answer for.

The issue is more acceptance of a robot that may kill someone during complex events. For instance the Arizona issue where the woman crossed into the street right in front of the car. In the video you can easily see based on when the person even ended up in the camera frame that no normal human driver in those conditions avoids that person because you can barely see them. There would be little outrage if a human hit her, but because it was an autonomous vehicle we lose our mind. The difference is the programmers can continue to improve the AI and sensors to get better, statistics show human drivers are getting worse so the outlook should still be positive. 

It's weird because a lot of the issue with them is this distrust of technology for important tasks. In most situations, even beyond driving, it is proven that AI handles high stress situations better than humans. We have to get passed our own preconceived belief we are better at mundane tasks than AI and accept our inferiority in certain areas. 

[Hoofhearted]

(04-19-2018, 09:32 AM)Au165 Wrote: Yes it can recognize human over things. Standard home PIR sensors can tell the difference, these vehicles are using super high end versions along with LIDAR and other sensors. Like everything they aren't perfect yet, and some are better than others, but everything on the road currently tests at "as good or better" than average drivers on the road today. Obviously things like the Trolley Problem are ethical questions that humans struggle with and would have to decide within a second who to take out, but we can't ask AI to solve ethical dilemma's that even we have no answer for.

The issue is more acceptance of a robot that may kill someone during complex events. For instance the Arizona issue where the woman crossed into the street right in front of the car. In the video you can easily see based on when the person even ended up in the camera frame that no normal human driver in those conditions avoids that person because you can barely see them. There would be little outrage if a human hit her, but because it was an autonomous vehicle we lose our mind. The difference is the programmers can continue to improve the AI and sensors to get better, statistics show human drivers are getting worse so the outlook should still be positive. 

It's weird because a lot of the issue with them is this distrust of technology for important tasks. In most situations, even beyond driving, it is proven that AI handles high stress situations better than humans. We have to get passed our own preconceived belief we are better at mundane tasks than AI and accept our inferiority in certain areas. 

Yeah, but in what situations can it do that? if it's -30 out, blinding snow, can it still tell that? What if a kid or someone was behind an object? I have my hesitations about it's full adaptability. 

And yeah, the issue of the robot and the security behind it is a huge cause for concern. I think people don't entirely trust it yet because it doesn't have a sense of perception and reasoning; they go off what a person is programming it. 

[mallorian69]

Unless they can come up with a 100% glitch/crash free completely unhackable system they shouldn't be allowed on the roads. Lives should never be 100% in the hands of computers.

If you want separate roads and use them strictly for cargo I'm good with that.

[Au165]

(04-19-2018, 10:14 AM)Hoofhearted Wrote: Yeah, but in what situations can it do that? if it's -30 out, blinding snow, can it still tell that? What if a kid or someone was behind an object? I have my hesitations about it's full adaptability. 

And yeah, the issue of the robot and the security behind it is a huge cause for concern. I think people don't entirely trust it yet because it doesn't have a sense of perception and reasoning; they go off what a person is programming it. 

Blinding snow gives it a better chance than the human eye, as I said it is using LIDAR and other sensors that don't require sight to know someone is there. A human can't see behind something, but with a combination of sensors the car will have better chances to do so.

Security is a fair hesitation, but I'm not sure how much of the critical systems of the car are networked. I would assume the very little is outside of GPS, but that is satellite to car which is a much more secure connection than simple networked connections. As to the perception and reason, Adaptive AI does have a sense of reasoning. Adaptive AI is the idea that it gets information plugged into it to start but gets ran through millions of tests and is given feedback of pass and fail and over time it figures out how to correctly navigate situations through learned behavior. 

....and again I come back to people are actually horrible drivers who kill 40k people in the U.S. each year. The perception that people are proficient drivers is just that only a perception, not actually a reality.

[Au165]

(04-19-2018, 10:53 AM)mallorian69 Wrote: Unless they can come up with a 100% glitch/crash free completely unhackable system they shouldn't be allowed on the roads. Lives should never be 100% in the hands of computers.

If you want separate roads and use them strictly for cargo I'm good with that.

This logic fails because humans already aren't 100% so demanding 100% before you improve from something that is inherently bad is simply stifling progress through fear mongering. The goal should be 100%, but an acceptable level should be, better than humans, which most testing shows it already is in many scenarios due to reaction time and inability to fatigue. I already laid out how reaction times of humans in their ability to process information at high speeds gives AI the advantage from the get go.

[mallorian69]

So you would be ok with a system that can be hacked to where every vehicle can be made to accelerate and make a hard left turn? If you were lucky only a few million people would be killed.

Any computer with any kind of data link can be hacked and anyone who believes that those who have nefarious goals wouldn't work to use that to their advantage is a fool.

I have no doubt that North Korea, Iran, or ISIS would jump at the chance to kill off 10s of millions of imperialist infidel dogs with a press of a button.

[XenoMorph]

(04-18-2018, 03:19 PM)Hoofhearted Wrote: I was reading an article around autonomous vehicles, and couldn't help but wonder why the potential for this is? I know there's been problems with it so far (believe Uber had a fatal accident), but I can't help but think of the possibilities.

According to one recent study, "By 2030, within 10 years of regulatory approval of fully autonomous vehicles, 95% of all U.S. passenger miles will be served by transport-as-a-service (TaaS) providers who will own and operate fleets of autonomous electric vehicles providing passengers with higher levels of service, faster rides and vastly increased safety at a cost up to 10 times cheaper than today’s individually owned (IO) vehicles."

There's 5 levels of autonomous vehicles



How long do you think it will be before we get to these levels? What are some of your biggest concerns? Personally, I think we're still quite a ways away. I'll say another 20 years before we can get to level 4. My biggest concern, being a certified ethical hacker, is the security around it. Having the possibility for someone (or something) be able to control millions and millions of cars is a scary, scary thought.

What do you think?

These will only be as good as their programming...

Could help eliminate bad drivers but will have tons of issues one you called out is a major issue in every aspect of our lives... The more you put in the hands of technology   the easier it is to access everything about you.

If cars drove themselves....  And someone could control them I think we see lots of high powered peoples cars suddenly driving off bridges lol.

Personally I would have to be disabled before I let the car drive me somewhere.


Put driverless cars are pointless unless the cars also can fuel themselves.

[Au165]

(04-19-2018, 11:13 AM)mallorian69 Wrote: So you would be ok with a system that can be hacked to where every vehicle can be made to accelerate and make a hard left turn? If you were lucky only a few million people would be killed.

Any computer with any kind of data link can be hacked and anyone who believes that those who have nefarious goals wouldn't work to use that to their advantage is a fool.

I have no doubt that North Korea, Iran, or ISIS would jump at the chance to kill off 10s of millions of imperialist infidel dogs with a press of a button.

Hacked is a buzz word that gets thrown around a lot, but few understand it. The idea you are going to be able to remotely control someones car is kind of comical. Cars for the most part are self contained systems, some use satellite connections for GPS but satellite to system connections are about the most secure form of connection out there especially if using double blind encryption. Cars are already run almost completely by computers, if you could gain a connection to a car today you could already do this by feeding the car commands, the issue again is the system is a self contained system so only a direct connection to the car allows for this to happen. Now these cars adding WiFi into the car need to be careful that the Wifi system is segmented away from the controls system, if this is done there isn't really a threat there. Even so, fighter jets and military vehicles already used special encryption that have been successful to this point of not killing a bunch of people with their weapons due to takeovers. A lot of this technology is getting ported into staving off threats in autonomous cars.

The bigger issue is people trying to fool the sensors intentionally through devices to jam sensors but this wouldn't be as much of the mass attack dooms day you predict but rather more of a one off guy being a dick/ trying to commit crimes. They are using multiple systems to confirm information being fed to the car to try and minimize this, but it is a reasonable concern and one that constantly needs to be improved. The other way you could "hack" the car is on the component side by embedding malicious code in pieces of the system either in production or through an accessory that gets introduced into the contained environment. The thing about this is you'd need access to some sort of connection to feed it live time commands which as I stated before isn't really part of the current design of this generation of autonomous vehicles.

There are already much more critical networked systems that are prone to "hacking" than cars. As I said trying to take down a large portion of autonomous cars all at once isn't likely, however taking down an electrical grid or blowing up a gas pipe is far more likely yet I don't see people grand standing about it.