Musk Admits Autopilot Not Perfect, Plus First Look At Model 3 Production


As promised yesterday, CBS This Morning posted video from the Fremont factory floor where the Tesla Model 3 is produced.

Additionally, Tesla CEO Elon Musk took CBS This Morning co-host Gayle King for a ride in the Tesla Model 3 so that she could experience Autopilot first hand.

Video description (above):

Elon Musk says Tesla’s autopilot system will “never be perfect”

Federal investigators forced Tesla out of a probe of a deadly crash in California last month involving the electric carmaker’s autopilot system. The driver died after crashing into a highway barrier. “CBS This Morning” co-host Gayle King experiences the autopilot system in the Tesla Model 3 and asks CEO Elon Musk about the safety concerns surrounding the technology.

And here’s the video featuring the Model 3 production line, where Musk says over 2,000 were produced over the past 7 days:

Video description:

Tesla CEO Elon Musk offers rare look inside Model 3 factory

The pioneering electric car company Tesla has suffered a series of very public challenges since the beginning of this year. Its high-profile CEO Elon Musk called this a period of “production hell.” Most of the troubles revolve around the company’s Model 3 sedan, its first mid-priced, mass-produced electric car. Musk takes “CBS This Morning” co-host Gayle King on a tour of his Silicon Valley factory. It is the first time network cameras were allowed inside the production line.

Video description:

Tesla CEO Elon Musk calls Model 3 assembly line “‘Westworld’ for cars”

Tesla’s Model 3 sedan is its first mid-priced, mass-produced electric car, but production of the vehicle has experienced significant delays. The company’s CEO, Elon Musk, takes “CBS This Morning” co-host Gayle King through the Model 3 assembly line at his Silicon Valley factory, the technology of the car and the parts that go into the Model 3.

*Some people have noted that they are unable to watch these videos in various countries. However, they’ve been able to watch them from the original source. Follow the link to watch on CBS This Morning.

Categories: Crashed EVs, Tesla, Videos

Tags: , , ,

Leave a Reply

29 Comments on "Musk Admits Autopilot Not Perfect, Plus First Look At Model 3 Production"

newest oldest most voted

“Autopilot Not Perfect”

If perfection were the standard for any safety system before it is put in cars, we would still be waiting for ABS, Air bags, Seat Belts, etc to be 100% perfect before being installed in any cars. And we would be seeing an order of magnitude more deaths while waiting for perfection.

You just don’t have good news reporters in the USA do you.

No Miggy, we really don’t.

Elon Musk says Tesla’s autopilot system will “never be perfect”

Good for Elon for stating this as fact. I get very tired of reading all the comments from people who don’t understand this; people who say we should wait until self-driving cars are “perfect” before they are allowed out on public roads.

As Nix already said, we can’t demand that air bags work perfectly every time; neither can we demand self-driving cars never have an accident. The realistic goal is to reduce the accident rate, not to eliminate it completely, just as with air bags, the goal is to reduce injuries and deaths, even though we know air bags can never prevent every injury or fatality.

“The thing to keep in mind is that self-driving cars don’t have to be perfect to change the world. They just have to be better than human beings.” — Deepak Ahuja, CFO of Tesla Inc.

(⌐■_■) Trollnonymous

There Aint nothing perfect!

You’re darn tootin !

Relaying too much on robots, you will always need the human touch to build things

An interesting place to learn about vehicle safety statistics is Insurance Institute for Highway Safety (IIHS). They have stats by make, model, and year.

The stats are a little different than fatalities per million miles. They are “driver deaths per million vehicle-years of registration”. I think we can get at least the driver death rate if we assume about 13,000 miles per vehicle-year.

The Chevy Volt and the Nissan Leaf were the two most-safe vehicles in the 2014 year. 7 driver deaths per million vehicle-years each. The average for 2014 was 30 deaths, so the Volt and Leaf were less than 1/4th of the average. No stats for the Model S. Seeing that the Model S had a total death rate about 1/4 of the average,, though, it would appear that the Volt and Leaf, without any “Auto Pilot” are just as safe as a Model S with Auto Pilot. So I don’t know that anything can be concluded. I suspect the bigger factor than AP is 5-star impact safety ratings with crumple zones, air bags, belts, ABS, etc, etc.

That statistic conflates two different stats: 1) How safe is the vehicle itself. 2) How safe is the typical owner’s driving habits. What you have likely discovered is that Volt and Leaf owners likely have less aggressive driving habits than buyers of sub-4 and sub-3 second performance cars. This pattern repeats itself throughout this IIHS data. Since drivers who drive aggressively self-select faster cars, and select these cars in higher numbers than they select 7 or 8 second economy cars, there is selection bias in those statistics. In other words, it is not possible to obtain the relative safety of the cars based on this data, without correcting for the drivers. Individual insurance companies actually work very hard to figure out this specific thing, and how to separate the two out. This way they can structure one premium based upon vehicle safety for the same vehicle, that is more expensive for unsafe drivers, and much less expensive for safe drivers. They spend millions of dollars fine-tuning these formulas to attract as many good drivers of a certain performance car, while pushing poor drivers of the same car to their competitors (or charging an extreme penalty rate). So if you are… Read more »

“That statistic conflates two different stats:

“1) How safe is the vehicle itself.
2) How safe is the typical owner’s driving habits.”

Going by vehicle years instead of vehicle miles certainly skews the accident statistics, but I suggest a different reason than you’ve given here, Nix.

According to the source linked below, the average annual mileage for Leafs is only 9697 miles. That’s considerably lower than the national U.S. average of ~14,000 miles.

What this suggests to me is that owners of Leafs drive them considerably fewer miles than most drivers, or at least most American drivers, drive their cars. Either Leaf owners are self-selecting for those who don’t drive as much, or else — more likely — owners of Leafs are likely to own a second car, probably a gasmobile, which they use for longer trips.

The same source also reports average Volt annual mileage at 12,238, still slightly lower than the average for American cars.

It’s pretty straightforward to reason that cars which are driven significantly fewer miles per year are, on average, going to have fewer accidents per year.

I think you are right. Both factors apply. There really isn’t a way to reverse engineer these statistics back to a pure vehicle to vehicle comparison. There are just too many variables.

Because yet another variable is that the EV’s aren’t distributed evenly across the country, or between cities and rural areas the same as other cars on the list. They also aren’t distributed in poor weather areas the same as other cars on the list, with sales in southern california dominating sales.

The more you examine the underlying confounding variables, the more meaningless these numbers become.

When the NHTSA’s ODI division studied Tesla airbag deployment and mileage data before and after Autopilot rollout, they found that accidents were reduced by about 40 percent.

The NHTSA study did not control for AEB, which was introduced a few months before Autosteer. In fact, the IIHS found a 40% improvement with AEB in other cars:

I don’t understand why people keep repeating this bad meme.

The article you linked to only claims, for the car so equipped, a 40% lower incidence of rear-ending other vehicles.

It does not claim an overall 40% lower accident rate, which is what the NHTSB found with Tesla cars equipped with both Autopilot and AutoSteer (not just Autopilot alone).

That still doesn’t make Tesla’s claims valid, as they don’t control for AEB. If they want to make valid claims, they should perform a better study.

Why do their statistics for AP+AEB have to be broken down to component statistics when the whole works as a sum of the parts? Not breaking down the statistics further doesn’t invalidate the known data.

Two quick thoughts:

1. Tesla charged my card for the $2500 for the M3 yesterday.

2. I have full confidence in Tesla.

3. Would not mind a quick roll with Gayle King.

Ok, that was actually three.

Dude, congrats on getting much closer to owning your Model 3! Please post your first impressions after you take delivery so we can learn what you think of the car.

And I too am a Gayle King fan, although I must admit I would much rather take a quick roll with Norah O’Donnell. 😉

Totally sick of videos that are ‘not available in your country’; please don’t link to videos that aren’t open.
Three videos in this story are not available in Oz.

We have no control over which videos are allowed in various countries, nor are we even aware. For the recent Musk interview videos, many people outside of the U.S. have had a problem. We’re sorry that they aren’t available where you live.


In comments on the previous article including this video, those people who could not access the video from InsideEVS reported that they were able to access it from the source, the CBS This Morning website.

It would be helpful if you’d insert a note to that effect into this article, along with the appropriate link:


Thank you very much! I’ve added the note and the link to the website.

Isn’t Musk supposed to keep his hands on the wheel when using Autopilot?
comment image

You mean AutoSteer, not Autopilot, which is on all the time. (But Tesla press releases make the same mistake, so you’re in good company.)

But, to answer your question: No. Watching the road and ready to grab the wheel at any time, yes. Hands on the wheel at all times, or even most of the time, is not necessary, at least according to the official Tesla demo for AutoSteer found in this IEVs article:

In this demo, the driver’s hands remain in his lap for the entire ~10 minute drive. about $56,500 all bells and whistles, and more when Dual motors are available.

So Gayle said something about Musk running a couple of stop signs. I’m curious if Elon ran the stop signs or was it Autopilot that ran the stop signs? If it was Elon, that’s a bad move for a guy who is so focused on safety (especially with a camera rolling while you do it). If it was Autopilot, that’s no bueno!

There is blowing through stop signs, and then there is rolling through stop signs without a complete stop. The first is bad, the second is pretty common, and frankly helps traffic flow if done safely, like motorcycles splitting lanes.

Without knowing what they are talking about, it is hard to know which they are talking about…. (QED)

I like in the third video he says that pack assembly is the “lagging element in production”.

And he is not letting CBS into the Gigafactory….let’s see those robots putting packs together….

or is it Lucy and Ethel on the assy line?

I think everything in Fremont is getting tweaked, and running at a speed limited by Nevada.