Tesla "Full Self Driving" not capable of autonomous driving

[Irace86.2.0]

Originally Posted by Spuds

-The difficulty was obvious from the get-go. His early "demo" wasn't a demonstration of the actual algorithm driving the car, he just had a bunch of stuff hard coded in. As an engineer with some experience in autonomy, I was certain Musk/Tesla was full of shit back in 2017 when he said it would only take a year.
-I would need to look closer at the data to discuss this further and I just don't have the time or energy. I still have a high amount of suspicion though.
-He's all about personal freedoms until the "free speech absolutionist" finds somebody saying something he doesn't like. Also a Hellcat does have a limiter, it's called traction control. You can defeat it if you want to wind up wrapped around a tree but that's a bit like falling asleep at the FSD wheel. The difference is you don't have the CEO saying this car will drive itself.

--I don't think he was being disingenuous, but I don't really know. The demo could have been manipulated, but it is more likely the car had just driven that path multiple times like being familiar with the way home. The original graphs in his presentation predicted the growth curve of vehicles on the road, the adoption rate of FSD, the data from the fleet feeding the algorithm, and it all presented itself as a feed forward process with a trajectory in the exponential growth stage of an S curve. The problem was that learning isn't linear or exponential, and so just because the abilities of the system were moving forward in a predictable way at one point, didn't mean it was going to continue. But since he admitted that it wasn't working out as he thought, then is it "fine from a critical perspective", as you said?

--You can read my last response to Capt Spaulding above. Unless the data from Tesla is an outright lie, as it is presented, it would be impossible for it to be what it claims and just a be a fluke of bias, such that the real truth is that Autopilot is more dangerous like the media and people here suggest.

--This is true. He definitely has shamefully silenced his critics on Twitter, while preaching free speech. Again, Tesla/Musk doesn't say the car drives itself, such that someone can be asleep behind the wheel or not be behind the wheel. And I said the Hellcat doesn't have an accelerator limiter, which it doesn't (or a speed limiter like a 155 limiter), and I said if someone can turn off traction control or stability control on the Hellcat then someone should be able to turn off FSD nannies on the Tesla, but then you mentioned traction control like I didn't mention it myself. Did you miss reading that sentence? Again, people are drawing conclusions about the situation being dangerous, but where is the evidence for this being the case? It is objectively more dangerous to have a button to remove traction control on a 707hp car, but Dodge puts it in the car, so why are we making a special case for Tesla's vehicles? Why are we special pleading without any real evidence?

[Dadhawk]

A question on the Tesla data/reporting, out of genuine curiosity if anyone knows.

Is Tesla saying FSD is safer on its own, or is it saying the combination of FSD and a fully aware driver is safer than the driver alone?

Frankly, I can see where FSD+Driver is safer than just Driver. However, I can't see that FSD alone is safer than a fully aware Driver without FSD.

[Capt Spaulding]

Originally Posted by Dadhawk

A question on the Tesla data/reporting, out of genuine curiosity if anyone knows.

Is Tesla saying FSD is safer on its own, or is it saying the combination of FSD and a fully aware driver is safer than the driver alone?

Frankly, I can see where FSD+Driver is safer than just Driver. However, I can't see that FSD alone is safer than a fully aware Driver without FSD.

Here is what Tesla’s literature says.

https://www.tesla.com/en_eu/support/autopilot

[Irace86.2.0]

Originally Posted by Dadhawk

A question on the Tesla data/reporting, out of genuine curiosity if anyone knows.

Is Tesla saying FSD is safer on its own, or is it saying the combination of FSD and a fully aware driver is safer than the driver alone?

Frankly, I can see where FSD+Driver is safer than just Driver. However, I can't see that FSD alone is safer than a fully aware Driver without FSD.

The statistics on Tesla's website says it is an aggregate of all incidents including accidents that happened within five seconds after Autopilot was disengaged, just in a case a long swerve was involved is my take. The data doesn't include FSD beta data. The FSD data would be nice to see, but ultimately it is a beta program and may be more biased because they can and have opted-out people who were abusing the system, but would probably need to include people who were outright abusing the system; if Dodge sent out a rollover recall or crash report for Hellcats because people doing donuts were rolling their cars, that would be bad data. I'm sure Tesla will post data once it is not in beta.

Considering Autopilot requires periodic inputs into the steering wheel and is camera monitored, we have to assume there is some minimal amount of driver engagement. There has been a number of studies, which has shown that driver's are on average less engaged with automated systems than without, but the MIT study found that this disengagement behavior doesn't modify the behavior all the time; when Autopilot is disengaged, the driver resumes "normal" engagement levels, so the behavior isn't carried over, which is good. With that said, the difference in driver engagement may not lead to a significant drop in performance/accident-avoidance irregardless of whether Autopilot was there or not; we don't know how this drop in engagement would affect the incidence of accidents; would it raise it 1%, 10%, but add Autopilot and that appears to compensate and then some.

https://electrek.co/2022/10/11/super...propilot-iihs/

https://electrek.co/2021/09/13/tesla...ion-new-study/

Again, the thing I come back to is that Autopilot is mostly used on the highway and appears to do well. FSD beta is used for both, but I'm guessing there is more city driving than rural driving, and interestingly, only 20% live in rural areas, but 45% of accidents/fatalities happen in rural areas. This could throw off data.

National Highway Traffic Safety Administration (NHTSA) surprisingly show that a greater number of traffic-related injuries and fatalities occur on rural roads. According to the data from the 2010 United States Census, 19 percent of people in the United States lived in areas that were classified as “rural.” Despite the rural population being the distinct minority, NHTSA reported that in 2019 alone, 16,340 people died in motor vehicle accidents that occurred on rural roads, as opposed to 19,595 reported traffic fatalities in urban locations.

-Quoted here: https://www.dolmanlaw.com/blog/auto-...l-urban-areas/

[Dadhawk]

Originally Posted by Irace86.2.0

Again, the thing I come back to is that Autopilot is mostly used on the highway and appears to do well. FSD beta is used for both, but I'm guessing there is more city driving than rural driving, and interestingly, only 20% live in rural areas, but 45% of accidents/fatalities happen in rural areas. This could throw off data.


-Quoted here: https://www.dolmanlaw.com/blog/auto-...l-urban-areas/

I don't really see a significant correlation between number of persons living in rural areas, and the number of accidents. It would only make sense to me if you tracked the drivers involved in the accidents correlated to where they live. I don't "live" in a rural area, but I drive thousands of miles every year in rural areas. Most drivers do, at least in my experience.

[Irace86.2.0]

Originally Posted by Dadhawk

I don't really see a significant correlation between number of persons living in rural areas, and the number of accidents. It would only make sense to me if you tracked the drivers involved in the accidents correlated to where they live. I don't "live" in a rural area, but I drive thousands of miles every year in rural areas. Most drivers do, at least in my experience.

That is a good point, but then I thought I remembered something about the majority of accidents happening close to home, and I found this (below). While five to fifteen minutes might put many people out of town/city into a rural area, it seems more likely that rural accidents happen to rural people and urban accidents happen to urban people.

In 2001, Progressive Insurance found in a survey that 52% of wrecks occurred within five miles of the drivers’ home and 77% occurred within 15 miles.

An electrical engineering company called Dolphin Technologies discovered that 25% of all car accidents occur within three minutes of a trip and 14% occur within six minutes. Additionally, the National Highway Traffic Safety Administration found similar findings that corroborate Progressive Insurance and Dolphin Technologies’ research. They found that 52% of accidents occur within five miles of a driver’s home and 69% occur within ten miles.

Quoted from: https://www.crossenlawfirm.com/blog/...pen-near-home/


Then I found this (below), which says deaths in urban areas were more likely to occur on collector and local roads and not on interstates and freeways, so even if someone lives in an urban area and uses interstate highways and freeways to travel outside the city, they don't crash on these urban freeways, which isn't to say urban people aren't venturing off into rural areas, but I'm just saying it seems most probable that urban people are having urban accidents.

I'm thinking this happens because roads are in poorer condition, visibility is worse, more wildlife and debris in the roads, smaller road sizes with no shoulders, more likely to drive fast or drunk, less police, more curves, less warning signs, people might be less engaged because there isn't traffic, etc.

Part of the point is that, yes, the data from Tesla might be an apples to oranges comparison because maybe Autopilot and FSD are used in lower risk areas, but also, they are used in lower risk areas, so there might just be less danger, and the systems don't have to be that good. Just a thought.

In 2020, crash deaths in rural areas were less likely to occur on interstates and freeways (13 percent compared with 20 percent) and on other arterial roads (22 percent compared with 58 percent) than crash deaths in urban areas and more likely to occur on collector roads (43 percent compared with 10 percent) and local roads (21 percent compared with 12 percent).

Quoted from: https://www.iihs.org/topics/fatality...ral-comparison

[Irace86.2.0]

This also talks about FSD development/progress.

[Dadhawk]

Tesla also opened up its charging network in the last two days. Charging spots popping up all over the US. Great stuff.

Still don't want a Tesla, but they will be my charging network of choice.

[Capt Spaulding]

Originally Posted by Irace86.2.0

This also talks about FSD development/progress.

If only I could bring myself to believe anything they say.





But I can’t.


But, like DH, if I live long enough to buy an eV I may well use their chargers.

[Irace86.2.0]

Originally Posted by Dadhawk

Tesla also opened up its charging network in the last two days. Charging spots popping up all over the US. Great stuff.

Still don't want a Tesla, but they will be my charging network of choice.

Here is a good review of the experience. Seems like a 10' J1772 EV extension cord would fix this problem. This one is 20'.

https://ev-lectron.com/en-ca/collect...05747645&_ss=c

[Dadhawk]

Originally Posted by Irace86.2.0

Here is a good review of the experience. Seems like a 10' J1772 EV extension cord would fix this problem. This one is 20'.

Doesn't work for DCFS but would work in a pinch.

[Irace86.2.0]

Originally Posted by Dadhawk

Doesn't work for DCFS but would work in a pinch.

Yeah more in a pinch. I guess the level 3 cables are water cooled, so that would make it hard for an aftermarket cable to work. I was unaware until I researched that.

[Capt Spaulding]

Ezra Dyer (Car and Driver) has an opinion piece in the NYT today suggesting that Ford may be the new Tesla.

https://www.nytimes.com/2023/03/04/o...mote-work.html

[soundman98]

Originally Posted by Capt Spaulding

Ezra Dyer (Car and Driver) has an opinion piece in the NYT today suggesting that Ford may be the new Tesla.

https://www.nytimes.com/2023/03/04/o...mote-work.html

hard to disagree. one of the points i was expecting he was going to make but didn't was how old the model s and y are getting. i mean, they've been making the model s in mostly the same form with minimal updates since 2014.