AI Says, Self driving cars will make insurance rates for Human Drivers cost prohibitive within 10 years.

[KJRaven]

Just last week the California DMV opted to start allowing Police to ticket self driving vehicle companies. Up until now... I guess they would just go "oh well" when it blocked an intersection or something.

https://abc7news.com/post/californi...s-allowing-driverless-truck-testing/19000544/

California is making some good progress on how to deal with the Autonomous Vehicles. Texas and other states are watching closely and will probably copy paste a lot of regulations. I just hope that CA can do it right.

 

[KJRaven]

Left turn from the center lane on red. If there is no driver, who gets the ticket? Just ask'n.

to be fair, there are 2 turning lanes, and the vehicle navigated the intersection safely... highly illegally, but safely.
In a world where the vast majority of autonomous vehicles are on the road, there will not be a need for traffic lights and all traffic flow will look like this.

Its an old video, but still applies here...

 

[E90400K]

to be fair, there are 2 turning lanes, and the vehicle navigated the intersection safely... highly illegally, but safely.
In a world where the vast majority of autonomous vehicles are on the road, there will not be a need for traffic lights and all traffic flow will look like this.

Its an old video, but still applies here...

It's good in theory. In practice it may not be as simple. To coordinate cross traffic at intersections as depicted in the video, there will need to be a central ground traffic control system that controls the movements of each individual automobile, which means the ground traffic control system will need to know the destination of each vehicle so it knows the direction and timing to move each car in relation to the other cars so as to keep pathways for each vehicle clear. A (pier-to-pier) mesh network (as brought up earlier) will be more difficult to safely operate vehicles for numerous reasons, but one is hacking, another is on-board processing power and redundancy. Moving the traffic flow to a centralized management system brings up privacy rights issues among others, hacking another. Then handoff between one local (municipal) traffic management system to another might not be as easy as thought. Lots of engineering and societal issues to be worked out. The ATC system is very safe because of mostly traffic separation (route planning) and system redundancies.

And to my earlier point, what entity is going to implement the engineering solutions? Get the US DOT involved (initiable IMO), and a well-developed solution (with standards) is decades away. Leave it to Sates, then there is the handoff issues. It may come across as I am pissing on the idea (what's new - lol), but this is not a simple and inexpensive task.

Raising insurance rates on human drivers in 10 years (because humans suck)? Not a chance.

 

[KJRaven]

It's good in theory. In practice it may not be as simple. To coordinate cross traffic at intersections as depicted in the video, there will need to be a central ground traffic control system that controls the movements of each individual automobile, which means the ground traffic control system will need to know the destination of each vehicle so it knows the direction and timing to move each car in relation to the other cars so as to keep pathways for each vehicle clear. A (pier-to-pier) mesh network (as brought up earlier) will be more difficult to safely operate vehicles for numerous reasons, but one is hacking, another is on-board processing power and redundancy. Moving the traffic flow to a centralized management system brings up privacy rights issues among others, hacking another. Then handoff between one local (municipal) traffic management system to another might not be as easy as thought. Lots of engineering and societal issues to be worked out. The ATC system is very safe because of mostly traffic separation (route planning) and system redundancies.

And to my earlier point, what entity is going to implement the engineering solutions? Get the US DOT involved (initiable IMO), and a well-developed solution (with standards) is decades away. Leave it to Sates, then there is the handoff issues. It may come across as I am pissing on the idea (what's new - lol), but this is not a simple and inexpensive task.

Raising insurance rates on human drivers in 10 years (because humans suck)? Not a chance.

I agree that there is a significant amount of work ahead to reach that theoretical point, but the momentum is already there, and I suspect it will happen faster than many expect.

The comparison to air traffic is weak at best. Ground traffic is fundamentally different and doesn't require the same level of rigid, centralized regulation. A car crash is typically a localized event with limited impact; a plane crash, by contrast, can be a global news event involving hundreds of lives and massive property damage.

Modern autonomous tech is being built to navigate roads locally with minimal outside control. In fact, it can run on hardware comparable to a smartphone from a few years ago. Comma.ai, for example, uses open-source software and basic dash-mounted cameras to achieve near-Level 3 autonomy on compatible vehicles. From what I’ve seen, it actually handles lane-keeping and adaptive cruise control better than non-FSD Teslas.

As more vehicles adopt this tech, the risk of collision drops. An "eyes-on" semi-autonomous approach is a perfect bridge—it handles the bulk of the commute, reduces human error, and provides the data needed to eventually take the "training wheels" off entirely.

Time will tell, but think autonomous vehicles will be a huge part of our lives in the not too distant future.

 

[RedJoker]

We don't have that central control / mesh network structure now with human drivers. I'm not convinced we would need that for AI either. For instance, if my car had an AI that made the exact same decisions as I would, based on what it senses, it would still be safer than me simply because it could react faster, monitor many more sensors in all directions, and never get tired or distracted. It certainly wouldn't be perfect but it would be an improvement.

I honestly think this would be why the insurance companies would drive this adoption. For many of us, AI driven cars have to be perfect in order for us to trust it. For an insurance company, 10% better might be good enough. With 40,000+ fatalities a year, that saves 4,000 people. That's significant, in my opinion.

Don't let the perfect be the enemy of the good... or something like that...

 

[KJRaven]

We don't have that central control / mesh network structure now with human drivers. I'm not convinced we would need that for AI either. For instance, if my car had an AI that made the exact same decisions as I would, based on what it senses, it would still be safer than me simply because it could react faster, monitor many more sensors in all directions, and never get tired or distracted. It certainly wouldn't be perfect but it would be an improvement.

I honestly think this would be why the insurance companies would drive this adoption. For many of us, AI driven cars have to be perfect in order for us to trust it. For an insurance company, 10% better might be good enough. With 40,000+ fatalities a year, that saves 4,000 people. That's significant, in my opinion.

Don't let the perfect be the enemy of the good... or something like that...

I’m with you on that. Even small, incremental improvements save lives every day. I’ve seen countless crash videos on Reddit that could have been avoided entirely with just basic sensors, a camera, and a little bit of compute.

Regulation is already moving in that direction. The NHTSA is pushing for driver monitoring systems as early as next year, and while it’s controversial, it highlights a hard truth: driving is a privilege, not a right. If we can't solve the issues of drunk, distracted, or exhausted driving ourselves, regulation has to step in to protect the public.

We often forget that seatbelts, windshield wipers, and even headlights were once regulated additions that faced pushback, yet they fundamentally changed roadway safety. Autonomy is just the next logical step in ensuring we can get from A to B without it being a life-or-death gamble.

 

[E90400K]

I agree that there is a significant amount of work ahead to reach that theoretical point, but the momentum is already there, and I suspect it will happen faster than many expect.

The comparison to air traffic is weak at best. Ground traffic is fundamentally different and doesn't require the same level of rigid, centralized regulation. A car crash is typically a localized event with limited impact; a plane crash, by contrast, can be a global news event involving hundreds of lives and massive property damage.

Modern autonomous tech is being built to navigate roads locally with minimal outside control. In fact, it can run on hardware comparable to a smartphone from a few years ago. Comma.ai, for example, uses open-source software and basic dash-mounted cameras to achieve near-Level 3 autonomy on compatible vehicles. From what I’ve seen, it actually handles lane-keeping and adaptive cruise control better than non-FSD Teslas.

As more vehicles adopt this tech, the risk of collision drops. An "eyes-on" semi-autonomous approach is a perfect bridge—it handles the bulk of the commute, reduces human error, and provides the data needed to eventually take the "training wheels" off entirely.

Time will tell, but think autonomous vehicles will be a huge part of our lives in the not too distant future.

This is where I disagree. I think the comparison to ATC is very pertinent to the implementation of autonomous ground traffic control. Being I worked on ATC surveillance for 15 years finishing in Dec 2021, I see the ground traffic management situation differently and with some experience (just trying to give my creds). Air traffic safety is extremely safe because how it is implemented. It is so safe because of a very well-controlled air space via vehicle separation regulations, near complete control over traffic routing, constant monitoring, redundancy, and quite importantly, air traffic has 3-axis of freedom to avoid collisions.

Ground traffic has just 2-axis of freedom where the y-axis (horizontal) is very constrained by permanent obstructions; obviously the Z-axis is zero (0). Concerning the X-axis, for example, using a 500 MPH airspeed in enroute air space traffic is separated by 5 miles and several thousand feet of altitude. The equivalent 5-mile separation, (which is about 36 seconds in time) at 60 MPH for ground traffic would be about 3,200 feet. Using a just a 5-second cushion for traffic at 60 MPH translates into 440 feet separation between cars travelling in the same direction. No one keeps 440 feet off the car in front of them. At 440 feet, less cars will fit on a given section of road, which the autonomously-driven public may not find acceptable. Additionally, there are no controlled intersections in air traffic at the same altitude (on purpose), that is where air space collisions happen when the system (or pilots) fail. Ground traffic is full of zero altitude change intersections (and where mostly the problem is) and very limited avoidance paths.

I see implementation of ground traffic automation as more difficult than air traffic control (which BTW, is quite automated right now). The key to ATC working is the underlying communication system the data and information is transmitted over; we've all heard of what happens when the aged physical comms infrastructure fails - i.e. Newark, NJ airport. The DOT is now dumping $12B to improve the situation over the entire national air space. Ground traffic is going to be highly dependent on the comm systems used to support it (there is a whole lot there to unspool to make the comms side 5-nines reliable).

I'm not trying to argue with anyone about it. I just trying to explain how I see it.

 

[KJRaven]

This is where I disagree. I think the comparison to ATC is very pertinent to the implementation of autonomous ground traffic control. Being I worked on ATC surveillance for 15 years finishing in Dec 2021, I see the ground traffic management situation differently and with some experience (just trying to give my creds). Air traffic safety is extremely safe because how it is implemented. It is so safe because of a very well-controlled air space via vehicle separation regulations, near complete control over traffic routing, constant monitoring, redundancy, and quite importantly, air traffic has 3-axis of freedom to avoid collisions.

Ground traffic has just 2-axis of freedom where the y-axis (horizontal) is very constrained by permanent obstructions; obviously the Z-axis is zero (0). Concerning the X-axis, for example, using a 500 MPH airspeed in enroute air space traffic is separated by 5 miles and several thousand feet of altitude. The equivalent 5-mile separation, (which is about 36 seconds in time) at 60 MPH for ground traffic would be about 3,200 feet. Using a just a 5-second cushion for traffic at 60 MPH translates into 440 feet separation between cars travelling in the same direction. No one keeps 440 feet off the car in front of them. At 440 feet, less cars will fit on a given section of road, which the autonomously-driven public may not find acceptable. Additionally, there are no controlled intersections in air traffic at the same altitude (on purpose), that is where air space collisions happen when the system (or pilots) fail. Ground traffic is full of zero altitude change intersections (and where mostly the problem is) and very limited avoidance paths.

I see implementation of ground traffic automation as more difficult than air traffic control (which BTW, is quite automated right now). The key to ATC working is the underlying communication system the data and information is transmitted over; we've all heard of what happens when the aged physical comms infrastructure fails - i.e. Newark, NJ airport. The DOT is now dumping $12B to improve the situation over the entire national air space. Ground traffic is going to be highly dependent on the comm systems used to support it (there is a whole lot there to unspool to make the comms side 5-nines reliable).

I hear what you're saying, and I appreciate your experience and expertise. However, I don't think it needs to be that complicated, at least not yet, if ever.
Cars have the advantage of being able to pull over, slow down, stop, or change speed much faster than planes. This offers more options if an anomaly arises. Yes, they might be in the way of an emergency vehicle now, but those instances are rare and will improve as they train for edge cases. They will continue to evolve as we teach them how to handle situations, eventually knowing what to do without outside help.

 

[The Dude]

Anytime I correct Ai. It always responds with a "Im glad that you caught that." So Im not putting myself in those hands.

 

[KJRaven]

Anytime I correct Ai. It always responds with a "Im glad that you caught that." So Im not putting myself in those hands.

I hate that response from ChatGPT and Gemini as well; they are large language models programmed to provide an answer, even if fabricated, and they are so often entirely not factual and should be questioned and reviewed for accuracy.
This is a different kind of "AI" than what should be used in a vehicle, which should have more rigid decision-making and safeguards to prevent hallucinations.

 

[TPL]

Really? You need to be an emeritus professor from Carnegie to figure that out?

To figure it out, no. To get government funding for figuring it out, yes.

 

[KJRaven]

To figure it out, no. To get government funding for figuring it out, yes.

☝This guy bureaucrats!

 

[E90400K]

I hate that response from ChatGPT and Gemini as well; they are large language models programmed to provide an answer, even if fabricated, and they are so often entirely not factual and should be questioned and reviewed for accuracy.
This is a different kind of "AI" than what should be used in a vehicle, which should have more rigid decision-making and safeguards to prevent hallucinations.

Just saying, but if I have to rely on "should have" from an AI driving app, I'll stick with what I know, me. I'm way more important to myself than I am to an AI driving app. I have skin in the game to keep myself alive.

Again, just saying.

 

[NMNeil]

Perhaps a key to autonomous vehicle acceptance will be lack od windshield preventing occupants from seeing what’s in front, more like a subway car.

What a wonderful opportunity to have a captive audience who get swamped with more puerile ads that can't be turned off

 

[Tom Sawyer]

The comparison to air traffic is weak at best.

Gotta disagree. There's lots of parallels to be found.

We don't have that central control / mesh network structure now with human drivers.

Isn't that what traffic lights & stop signs do?