Autopilot Alarmist Claims: Is It Really Crashing More Than Humans?

Long time Autopilot alarmist Faiz Siddiqui has a new article out claiming that Autopilot has been involved in 736 crashes.

But Autopilot is the most widely used driving automation tech in the world. Is it really crashing more than humans? Let's do some math in this thread. pic.twitter.com/Mmc24Rc9U1

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Now, this isn't just me objecting to how the data is presented as an Autopilot user. NHTSA is crystal clear about the limitations of this data in the very report WaPo is mentioning.

Summary incident report data are *NOT NORMALIZED*. pic.twitter.com/SOXoaVoxMp

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

NHTSA lists many other limitations of the data, such as the fact that Tesla reports data more quickly than others and that incident reports may be unverified.

But let's focus on mileage and why it's so important for a minute. pic.twitter.com/G3KXb5T0fI

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Imagine I come to you and say hey i've got the world's safest automated driving system! nobody has ever crashed while using it and nobody has ever died!

Sounds impressive, right? Until you realize 0 miles have been driven with my system. Of course, 0 crashes is easy with 0 miles

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Similarly if system A has 2 crashes and system B has 1 crash it sounds like system B is safer.

Until you realize system A has driven 100 billion miles and system B has only driven 10. Now system A looks much better

Without normalization, this data is meaningless.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

So are people really crashing and dying more often on Autopilot than if they drove themselves, as WaPo seems to imply?

Well, let's do some basic math that apparently they don't teach in journalism school. Tesla recently said they drove 100 billion mileshttps://t.co/2LMPU507xb

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

This is all Tesla miles traveled, Autopilot on or off.

In the US we have 1.35 fatalities per 100 million miles traveled. Drive 100 billion miles and you'd be expected to have 1,350 deaths.

If 17 of them had Autopilot on, that's ~1.25% of total fatalities. pic.twitter.com/l6wC4J7iVJ

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

In other words 98.75% of fatalities occurred when humans were in control of the car.

This is the real problem. Human caused car crashes and fatalities. Autopilot & tesla vision active safety were designed to address this problem, but media pretends they caused it!

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

If you believe clowns like Faiz and Dan O'Dowd, 99% of crashes are caused by Autopilot. But when you actually look at the data it says the exact opposite: 99% of crashes are caused by humans, and could be eliminated with AI in every car.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Now let's turn to crashes. In the US we have 1.53 crashes per million miles traveled. So you would expect 153,000 crashes.

Teslas, even with Autopilot & active safety off crash less. So you'd expect 68,000 crashes in a Tesla with those features off. pic.twitter.com/WkzknQW6Jk

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

736 crashes might sound like a lot, but compared to 68,000 total crashes you find again that 99% of crashes happened when Autopilot was off.

Autopilot was on for less than 1% of all crashes.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

So are people crashing more often on Autopilot than they would if they drove themselves?

That depends on what percentage of miles traveled are on Autopilot. If Autopilot accounts for 20% of miles traveled, but only 1% of crashes, that would suggest people crash less not more

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Conversely, if Autopilot only accounts for 0.1% of miles traveled but 1% of crashes that would suggest it crashes much more often than when people are driving manually

Reports from users suggest some are using it for 96%+ of their drivinghttps://t.co/EFJWSgFvLv

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

I find it very unlikely that Autopilot is used for less than 1% of all miles traveled in a Tesla. In fact I wouldn't be surprised if it is more than 50% of all miles traveled.

What the data suggests is that people are crashing less with Autopilot, not more!

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

The implication of this article is that if you simply banned Autopilot, nobody would crash or die. But that's flat out wrong. People still need to drive, even if you take away their Autopilot.

If Autopilot were banned, you'd see MORE crashes and deaths not more.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Autopilot critics may roll their eyes at this, but it makes perfect sense to anyone who has used AP.

In manual driving mode, humans recklessly zip around in a Tesla that goes 0 - 60 in a snap. Autopilot is cautious and conservative by comparison and monitors driver attention

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

A lot of critics like to pretend that distracted driving is a problem caused by Autopilot when the reality is distracted driving is a huge problem in every car, mostly the ones that don't have Autopilot.

Autopilot & FSD now monitor drivers phone use, making them much safer.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

When I was driving across California on FSD yesterday I noticed how many people were using their phone in the car.

Meanwhile I'm not allowed to take my phone out or the car beeps at me. This is a tremendous safety advantage, to force driver attention.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Journalists like this make me sick. They go and write a sensational story devoid of any context that suggests that we need to ban this technology to save lives

The reality is, if people listened to them more would die. Thankfully, regulators don't because they see the data.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Take a look at this guy's track record. All shitting on Tesla and Autopilot all the time, never taking in new information.

Reminds me of Jonah Jameson. Everything has to be Spider Man is a menace! No regards to what actually happened in reality

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Finally I'll say just because Autopilot & FSD were on during a crash does not mean it is the fault of Autopilot.

Today, the human is always responsible and must maintain control and pay attention to the road, just like when driving manually.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Let's say I'm driving on the highway on Autopilot and an out of control Semi truck barreling down the hill rear ends me sending me into another car.

That would be reported as a crash with Autopilot on, even though nobody would blame Autopilot.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

When you look at many of these cases where there was a fatal car crash, you often find that the people were playing a game on their phone, distracted, or intoxicated.

Today's FSD driver monitoring system makes using your phone almost impossible. This is a thing of the past.

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Oh another thing. He said the number is increasing and skyrocketing. Again, you have to normalize it.

If you look at all Teslas, you could say wow more Teslas are crashing every year! Horrible! 1 million% increase from 10 years ago!

Yeah cuz there were no Teslas 10 years ago

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023

Obviously more Teslas are going to crash every year if you're making more. That doesn't mean there's an issue. You want to look at crashes per car produced or miles traveled to normalize it.

When you look at it that way, crashes per car are going down not up!

— Whole Mars Catalog (@WholeMarsBlog) June 11, 2023