Tesla Autopilot failed to notice a tractor trailer

[Dadhawk]

Originally Posted by soundman98

you being a pilot, this comment surprised me.

i made a similar comment a few times on jalopnik, and had close to 30 comments from claimed pilots about how autopilot is not automatic in an airplane, and everyone knows this, making the name of the system on a ground vehicle perfectly reasonable...

IMHO they were wrong.

Yes, an autopilot in an airplane has different stages from basic wing levelers to fully automated landings, and some really advanced ones have traffic avoidance. However, there is a big difference between flying in "big sky" and down a crowded highway.

Even without autopilot it is not unusual for a pilot to have his head in the cockpit and not looking outside for seconds, or minutes while reading a map, setting radios, etc. Heck, you even fly blind and never look out the cockpit.

You cannot drive a car, in a heavy fog (which is just a cloud) with your head inside looking at the instrument panel. Pilots do this all the time and most with the help of an autopilot.

Also, the general public really does believe autopilot means you don't have to fly hands on or look outside.

[soundman98]

Originally Posted by Dadhawk

Also, the general public really does believe autopilot means you don't have to fly hands on or look outside.

that was exactly what i was trying to convince these people of.

on the bright side, i now know more about how airplanes work than i will ever need.

[Irace86.2.0]

Originally Posted by Dadhawk

The name of their product and the rantings of their leader doesn't help that situation much.

The name really doesn’t matter. Autopilot vs Pilot Assist vs Drive Pilot vs Drive Assistant Plus vs Super Cruise—whatever—no difference. Is the argument really that Tesla owners are so clueless as to spend thousands of dollars on an extra purchase option only to use it without knowing what it is about? Moreover, it would be immediately apparent that Autopilot is not Level 5 the second the system disengaged or beeped its warnings, so unless everyone who is crashing has done so during their first use of the system, they probably learned that the system requires the driver’s attention.

[86MLR]

Auto-pilot

[JD001]

Originally Posted by 86MLR

Auto-pilot

Is this thread on 'auto-bump'??

[86MLR]

Originally Posted by JD001

Is this thread on 'auto-bump'??

No, the other one

LETS DANCE

[soundman98]

Originally Posted by Irace86.2.0

The name really doesn’t matter. Autopilot vs Pilot Assist vs Drive Pilot vs Drive Assistant Plus vs Super Cruise—whatever—no difference. Is the argument really that Tesla owners are so clueless as to spend thousands of dollars on an extra purchase option only to use it without knowing what it is about? Moreover, it would be immediately apparent that Autopilot is not Level 5 the second the system disengaged or beeped its warnings, so unless everyone who is crashing has done so during their first use of the system, they probably learned that the system requires the driver’s attention.

no, the argument is moreso that the majority of drivers are that stupid.

it's not a vehicle brand issue. IMO, drive assistant plus, or super cruise are both much more preferable names, as they both indicate increased capability without indicating that the car can do things itself. anything with 'auto', or 'pilot' in it tends to indicate being hands-free in some respect.

the real problem is that everyone keeps trying to differentiate their iteration of the same tech. if they just made marketing materials with "our advanced level 2 cruise control" it would eliminate a lot of these problems.

but as they say:

[Irace86.2.0]

Originally Posted by soundman98

no, the argument is moreso that the majority of drivers are that stupid.

I think you're using the phrase 'majority of drivers' quite liberally. Most drivers aren't stupid. Accidents in general are very rare--especially serious ones. I work in the ER of the trauma center for the city, so I would know. While some of the crashes could be the direct fault of Autopilot (it veers abruptly--unlikely), the vast majority are most likely due to driver negligence, which has nothing to do with Autopilot. It isn't the case that they believed the car was fully autonomous. These same drivers may have crashed while texting on their phone or doing something else negligent. Autopilot just becomes an escape goat to blame someone else and to collect a settlement.

If the following statistics are true then Teslas and/or Telsa drivers are safer vehicles/drivers than average. A person is 4x more likely to get into a crash in a normal vehicle than a person in a Tesla without Autopilot, and a person is 7x more likely to get into a crash in a normal vehicle than a person in a Tesla with Autopilot engaged:

https://futurism.com/the-byte/tesla-...ot-safer-human

According to the National Highway Traffic Safety Administration’s most recent data, there’s an auto crash every 436,000 miles driven in the United States.

In the fourth quarter of 2018, Tesla reported one accident for every 2.91 million miles driven with Autopilot engaged. The first Tesla safety report of 2019 shows that rate increasing slightly, to one accident every 2.87 million miles.
However, both of those figures are better than the statistics for Teslas without Autopilot engaged: one accident for every 1.76 million miles driven in Q4 2018 and one every 1.58 million miles driven in Q1 2019.

[Dadhawk]

Originally Posted by Irace86.2.0

If the following statistics are true then Teslas and/or Telsa drivers are safer vehicles/drivers than average.

First, these are Tesla statistics so they are automatically suspect (and I would say that if it was Toyota releasing Toyota statistics).

Second, at least to me, this is only significant if you are comparing like vehicle and owner classes. If you compare this to ALL cars, that means you are including everything for $2,000 cars driven by teenagers to supercars driven by rich, self-absorbed I own the road you just get in my way types.

I'm guessing, but do not know, that people who buy in the so called "luxury sedan" and other ranges the Tesla are in are most likely less accident prone than the overall population anyway. I'm basing this on the anecdotal evidence of 20 to 50 accidents I see per year while commuting (I drive a lot) and rarely is it this class of car is involved in an accident of substance.

[Irace86.2.0]

Originally Posted by Dadhawk

First, these are Tesla statistics so they are automatically suspect (and I would say that if it was Toyota releasing Toyota statistics).

Second, at least to me, this is only significant if you are comparing like vehicle and owner classes. If you compare this to ALL cars, that means you are including everything for $2,000 cars driven by teenagers to supercars driven by rich, self-absorbed I own the road you just get in my way types.

I'm guessing, but do not know, that people who buy in the so called "luxury sedan" and other ranges the Tesla are in are most likely less accident prone than the overall population anyway. I'm basing this on the anecdotal evidence of 20 to 50 accidents I see per year while commuting (I drive a lot) and rarely is it this class for car is involved in an accident of substance.

This is why I qualified Telsa vehicles or Tesla owners, which gives room for causality being linked to the car or to a subclass of owners being more responsible. I have seen people try to analyze and compare data to other luxury brands with data that really doesn't exist yet (kinda like you said above--it is coming from Tesla), and the conclusions are vague at best. There is definitely a relationship between income and traffic accidents:

https://www.herrmanandherrman.com/bl...car-accidents/

Even if musk is exaggerating the safety of their vehicles related to other luxury cars and owners, do we really believe he is lying about the safety of the Autopilot system being safer than not having it engaged? This is a stretch. The fact is that only he has the data for this statistic. His company is running simulations in every Tesla that is on the roading, pitting its AI against the driver. It is comparing what the AI would do with what the driver chooses to do, so it learns, and when the driver crashes into someone, there is a simulation of what the AI/Autopilot would have done, so Tesla not only has data on how safe the Autopilot system is based on actual crashes, but it also has data on how many crashes the Autopilot avoided or could have avoided. This puts it in a very unique position to make a truth claim that can't be substantiated. That doesn't mean that they should be blindly believed, but it is just something to consider.

What do we know? Tesla's do really well on crash safety tests. In fact, they do so well that their performance is often off the charts, so they may achieve a 5 out of 5 when their likelihood of injury is closer to a 6 out of 5.* There are videos of the Model X doing rollover tests, and it couldn't be rolled over. Of course, Musk being Musk isn't without controversy (see the link below).

https://www.consumerreports.org/car-...-test-results/

*A 5 could be 15% of injury or less; a 4 could be 20% or less; a 3 could be 25% or less; etc. Tesla could have a chance of injury of 4%, but still gets a 5 when it should get a 7 because 5 is the max.

[Dadhawk]

Originally Posted by Irace86.2.0

Even if musk is exaggerating the safety of their vehicles related to other luxury cars and owners, do we really believe he is lying about the safety of the Autopilot system being safer than not having it engaged? ...

I don't know if he's being deceitful or not, although I will say Musk will always spin things in the best light possible for his situation. That's what he's best at, and why I don't but a lot of stock into what he says. In the end he usually accomplishes what he says he will, which I respect and admire, but he's going to say what benefits his company, as he should. He's also going to spin the numbers to his benefit, as pretty much any non-independent study will do.

What i haven't seen is a comparison if the severity of the accidents between their Class Two automation and people driving their cars. For example, how many fatal accidents per mile driven are there between the two, or the cost of repair on average between the two? Maybe I just missed it.

[soundman98]

cost-to-repair on tesla is currently a difficult number, as there's plenty of reports on very minor accidents that take many months to repair due to a lack of replacement parts. it's not a very common metric to look at though.

even the tesla-centric sites acknowledge it.
https://teslatap.com/articles/tesla-...a_parts_delays

[soundman98]

Originally Posted by Irace86.2.0

I think you're using the phrase 'majority of drivers' quite liberally. Most drivers aren't stupid. Accidents in general are very rare--especially serious ones. I work in the ER of the trauma center for the city, so I would know. While some of the crashes could be the direct fault of Autopilot (it veers abruptly--unlikely), the vast majority are most likely due to driver negligence, which has nothing to do with Autopilot. It isn't the case that they believed the car was fully autonomous. These same drivers may have crashed while texting on their phone or doing something else negligent. Autopilot just becomes an escape goat to blame someone else and to collect a settlement.

If the following statistics are true then Teslas and/or Telsa drivers are safer vehicles/drivers than average. A person is 4x more likely to get into a crash in a normal vehicle than a person in a Tesla without Autopilot, and a person is 7x more likely to get into a crash in a normal vehicle than a person in a Tesla with Autopilot engaged:

https://futurism.com/the-byte/tesla-...ot-safer-human

i think we're confusing arguments. i'm not attempting to argue that the system is less safe, i'm attempting to poorly argue that people are making it less safe by expecting more out of the system than it can actually do.

as you guy's started talking about, there is the relative intelligence and relative financial worth of the current ownership groups. i've said before, tesla is in a sort of oddball spot, as they're one of the first ever auto makers to work from the top down instead of the bottom up. every other car company started producing budget-minded vehicles, and then slowly moved products upmarket as their customers aged and/or had more money.

i feel that the user interface issues are going to grow more significantly as they continue to move more into the mainstream, downstream of where they started. with the release of the porsche taycan, porsche has been harping about how they are held to much higher reliability and user standards than tesla, which has a lot to do with them being an established company. i've been curiously watching how this is or will affect tesla in the future as they continue to move more into the mainstream instead of being a more limited high end production like how they started.

google has run into some similar autonomous issues as tesla, where their system is plenty safe, but most issues arise from the times where the google system perfectly follows the law in the scenario, but the human driver does something odd that leads to an incident.

i will say, that due to the fear/expectations/newness of such systems, any and all incidents are very unfairly weighted against the technology. it's not all that different from that first uber safety report. the numbers look dramatic, but some article commentors started putting it into a statistical comparison against airplane safety, and then the uber numbers for each type of incident started to look safer than that plane travel...

[Captain Snooze]

Originally Posted by soundman98

i'm attempting to poorly argue that people are making it less safe by expecting more out of the system than it can actually do.

"Risk compensation is a theory which suggests that people typically adjust their behavior in response to the perceived level of risk, becoming more careful where they sense greater risk and less careful if they feel more protected. Although usually small in comparison to the fundamental benefits of safety interventions, it may result in a lower net benefit than expected.
By way of example, it has been observed that motorists drove faster when wearing seatbelts and closer to the vehicle in front when the vehicles were fitted with anti-lock brake."

Wikipedia