EVNow
Well-Known Member
Thats why you shouldn't use NHTSA data. Don't use known bad data !
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Elon's Thanksgiving gift: Beta 5 coming in a few days
- Thread starter DanCar
- Start date
EVNow
Well-Known Member
I posted some in this thread. I've posted the links to those white papers in the past.
Basically Virginia Tech (?) conducts several studies to get a better handle on accident rates, types of accidents etc. That is what Waymo , Cruise etc use.
This just in - human driver ignores traffic light, kills two people. We need legislation to ban this dangerous practice!
Man arrested in fatal 6-car crash on Thanksgiving | WTOP
Man arrested in fatal 6-car crash on Thanksgiving | WTOP
Daniel in SD
(supervised)
EXACTLY!!! Thank you!
It honestly blows my mind that folks are so up in arms about the potential of FSD. Two systems are always better than one. And the current state of FSD requires that the human driver is ALWAYS in control of the vehicle.
Folks in these forums act as if FSD is stating that the driver can take a nap while the car does all the work. We KNOW that the technology isn't there yet.
My wife and three kids were involved in a horrific accident. Someone t-boned them in an intersection and flipped their car. She called me hanging upside down in the car with my 1 year old, 3 year old, and 5 year old kids SCREAMING in the background. Thankfully only the 3 year old suffered any injuries (she broke both bones in her left arm). A traumatic accident to say the least.
My wife insists on us taking our Model 3 on family trips because she likes the redundancy of Auto Pilot and ME driving the car.
I don't know if we will ever get to full L5 in my lifetime. But I'm 99% certain that a mixture of autonomy and human attentiveness is 100% better than one single point of failure (humans).
Daniel in SD
(supervised)
That depends on what percentage of red lights the computer would miss and what percentage of the time the human operator would catch it.
It seems like it would be much better for the car to prevent the human from running a red light than the other way around. A human might be fooled into thinking that the car is far better at seeing red lights than it is and let their guard down. A computer may have a high error rate but it will never let its guard down.
I'm not surprised.
Tesla already makes the safest vehicles from crash testing. Adding FSD or computer monitoring to improve safety triggers all sorts of concern trolling.
I bought FSD and I use autopilot every time I drive. It's a game changer for the freeways and stop-and-go urban traffic.
Tesla already makes the safest vehicles from crash testing. Adding FSD or computer monitoring to improve safety triggers all sorts of concern trolling.
I bought FSD and I use autopilot every time I drive. It's a game changer for the freeways and stop-and-go urban traffic.
The human driver is always expected to be in control. I'm not sure why everyone keeps skipping over this.
There’s definitely a push to discredit what Tesla is achieving here. Completely agree with turnem. I’m not sure you could convince me that the majority of drivers on the road today are safer than a driver WITH an enhanced driver assist package (call it whatever the f*** you want). There are so many variables that come up including actual environmental variables such as weather, road conditions, obstacles, lighting, surrounding bad drivers, not to mention the individual driver variables including sleep deprivation, visual acuity, mental processing speed, etc. The list is really astounding.
So by providing an incredibly advanced computer brain that only improves with time, can only reduce OVERALL harm. It’s never going to be a switch on, switch off scenario where one day human drivers are routinely killing each other and themselves, then the next day computer based drivers don’t kill or injure a single person. But the lives saved and injuries prevented during that transition are still significant and important. To discredit that is ridiculous.
I don’t believe anyone posting here that’s pro FSD-beta is really suggesting that a fleet wide release would be a “L5” release. That’s really not the point at this juncture. I look forward to watching and eventually participating in continued improvements.
So by providing an incredibly advanced computer brain that only improves with time, can only reduce OVERALL harm. It’s never going to be a switch on, switch off scenario where one day human drivers are routinely killing each other and themselves, then the next day computer based drivers don’t kill or injure a single person. But the lives saved and injuries prevented during that transition are still significant and important. To discredit that is ridiculous.
I don’t believe anyone posting here that’s pro FSD-beta is really suggesting that a fleet wide release would be a “L5” release. That’s really not the point at this juncture. I look forward to watching and eventually participating in continued improvements.
Daniel in SD
(supervised)
People do not always behave the way you would like them to and I'm not sure why anyone would rely on that. They are not machines.
Last edited:
Daniel in SD
(supervised)
I agree, I just don't think beta FSD will improve safety. All the advances that Tesla and others are making can and will be used to make cars safer though.
I understand what your saying. I just don’t see FSD beta as a simple party trick. I believe and personally see a much larger picture forming here.
Daniel in SD
(supervised)
It's hard to understate the value of true FSD. I have no disagreement with that.
I just don't agree with the logic:
- Automation can make cars safer.
- Beta FSD is automation.
- Therefore beta FSD makes cars safer.
It is possible for Tesla to design the system so that it has a confidence factor. For example when there are three lights it is harder to make a mistake compared to only one traffic light to look at. If Tesla doesn't have enough nines in certain situations it could demand driver take over.
It is a very interesting road we are on. I think everyone can agree there will be accidents with FSD in control with driver supervision. We probably won't hear about most of those, like running into curbs. What we really care about is where there is serious injury or death.
Will FSD reduce or increase serious injury? This is a big juggling act that Tesla is dealing with. Early on, like now, it will reduce serious injury because no one trusts it and everyone is paranoid while using it. Later on as it gets better and much better, humans have a tendency to be less paranoid, monitor less, and that is when serious injury will occur. There are lots of arguments about this like:
- If you use a defeat device then it is your fault
- It is not the pedestrians fault that got run over. Tesla shouldn't allow this.
- Even if there are a few serious accidents overall accidents are much lower so we should continue the development that is in overall saving lives. (Yes there is no data to support this now)
Last edited:
I bet most owners care about running into curbs, probably all care about causing even minor injuries. I would hope that the first priority of FSD is "don't run into anything".
HighZ
Member
I do believe that the FSD that gets released by Tesla (not today's beta) will be a big improvement in safety. Even today's beta could be safer with a more limited domain to avoid the potential of a big mistake (maybe daylight on well-defined roads as one example).
But we know that there are people even with today's released AP software that like to take videos of their car driving down the road with no one in the driver's seat. All we need is someone reckless or even someone deciding that writing that email is more important than paying attention just at the same time as the car doing the 'worst thing at the worst possible time'. The last thing we all want is heavy government regulation of what software we should be driving with if there were a flashy accident caused by the new software. And yes, those accidents happen every day in all cars and we can't do much about it. But if it were to happen with a Tesla running FSD, they could force that FSD be pulled back.
All that said, perhaps FSD will become much more bullet-proof soon to allow it to get into everyone's hands, or maybe there will be better driver monitoring to allow it to get out to more people sooner. I certainly want it - I am just afraid of what a very small percentage of less-than-average drivers may do to cause a lot of problems for the rest of us.
I think people will get irritated if brakes hard for a snake in the road, a paper bag, etc... But yes, not running into important things should be highest priority. Are there any videos of it avoiding potholes?
Do people have a sense of what FSD beta is really good at? Has it made a mistake of staying in your lane going straight through an intersection? Or following a single curving lane? Those seem to be significantly improved over the unofficial Autopilot usage on city streets right now.
If Tesla "officially" releases a relatively restricted version of "Autosteer on city streets," perhaps people's expectations of the system will be tempered, and Tesla can gradually "unofficially" add more functionality without a big splash that could attract abuse of "Full Self-Driving Capability." This would show significant progress publicly while allowing those who are already using Autopilot on city streets to experience more of the FSD improvements.
Similar threads
