How Elon Musk Knocked Tesla's "Full Self-Driving" Off Course

article paraphrased:
Musk zeroed in on car radar sensors. He eliminated it on new vehicles in May 2021. He said the 8 cameras were enough. Some of the engineering staff were aghast. They contacted a former Tesla executive on how to talk Musk out of that decision. There are plenty of stories about employees that are fired for disputing Musk opinions.

Older Tesla vehicle had it disabled. The result, according to interviews with nearly a dozen former employees and test drivers, safety officials and other experts, was an uptick in crashes & near misses.

In recent weeks, Tesla has recalled and suspended the rollout of the technology to eligible vehicles amid concerns that its cars could disobey the speed limit and blow through stop signs, according to federal officials.
Customer complaits have been piling up, including a lawsuit filed in federal court last month claiming Musk has overstated the technology’s capabilities.

IF this story is accurate & true; it would appear Tesla is no where near having “Full Self-Driving” capability and Musk interference may have hampered development of said technology.

5 Likes

I’ve lost faith in FSD. From anyone, actually. It will happen (already has) in some arenas. Farm tractors. Warehouse trolleys. A special-purpose Caterpillar super heavy dump truck. All that can be had today. I can see highway driving for trucks. Likely cars. And I think it just stops there.

Robo taxi? Nope. Won’t believe it until I see it.

3 Likes

Sure, but as is typical the story is pretty much garbage. Lots of fear-mongering and innuendo. But by the time you’ve read it you’ll be more ignorant than when you started.

Tesla’s elimination of the old radar has resulted vehicles that have scored record highs on Europe’s safety tests. So the worries have all proven to be baseless. And it appears that Tesla has a new higher definition radar going out now (as part of what is called Hardware 4), that will work even better. The issue has always been whether the cost of sensor fusion is worth the expense, and it seems as though maybe they’ve found a way to make it so.

There’s a similar issue around Tesla’s elimination of the ultrasonic sensors it had on all vehicles.

Tesla actually screwed up one thing on that: Tesla Vision was not up to providing park assist without ultrasonic sensors at the time the sensors were scheduled to be eliminated, but they got rid of them anyway (I’m guessing because hardware supply issues are hard to be flexible about). And then it took way too long to get to the point where the software was ready. So there are a bunch of cars that have been delivered without the sensors and with an excuse that the feature would reappear “soon”.

As it turns out, “soon” is now, a hiatus of months. Or in a couple of weeks for the cars that don’t have FSD Beta. The reports from people who have tried it is that Park Assist with no sensors works even better than it used to work with sensors. So perhaps all the hand-wringing and pearl clutching was misplaced? We’ll know better as the latest software release spreads to more vehicles.

Meanwhile, these scary articles are just fear-mongering. Hardly worth looking at.

-IGU-

3 Likes

Nice find

For someone who has a full engineering background I have a question. How can a radar system be used by every car on the road? Are we talking cancer causing if there was a switch over?

There are several companies that are considered to be ahead of Tesla in autonomous driving, including:

  1. Waymo: Waymo, owned by Alphabet Inc. (Google’s parent company), is widely considered to be the leader in autonomous driving technology. The company has been testing its self-driving cars on public roads since 2009, and its vehicles have driven more than 20 million miles on public roads and over 10 billion miles in simulation. (1 - see footnote below)
  2. Cruise: Cruise, a subsidiary of General Motors, is another major player in the autonomous driving space. The company has been testing its self-driving cars in San Francisco since 2015 and has already launched Robotaxis in San Francisco and plans to launch a commercial service in the near future in other cities.
  3. Mobileye: Mobileye, owned by Intel, is a leading provider of computer vision technology for autonomous vehicles. The company’s advanced driver assistance systems are used by several automakers, including BMW and Audi, and its self-driving technology is being tested in several cities around the world. (2 - see footnote below.)
  4. Baidu: Baidu, the Chinese search engine giant, has been investing heavily in autonomous driving technology and has launched several pilot programs in China. The company’s self-driving cars have already logged millions of miles on public roads in China.
  5. Zoox: Zoox, a self-driving car startup that was acquired by Amazon in 2020, is developing a fully autonomous vehicle that is designed for ride-hailing services. The company has been testing its vehicles on public roads in San Francisco and plans to launch a commercial service in the future. (3)

It’s worth noting that Tesla is still considered a major player in the autonomous driving space and has made significant progress in developing its Autopilot system. Auto-pilot is now known as a Level 2 Drive Assistance by the State of California’s DMV. However, these 5 companies above are often cited as being ahead of Tesla in terms of the sophistication and reliability of their autonomous driving technology.

Also, as far as Level 2, there are probably a half-dozen driver assistance programs in Ford, Cadillac, Mercedes, Audi, Toyota, Hyundai, VW, etc. whose Driver Assistance is beyond Tesla’s. Mercedes is now licensed for Level 3 Driver Assistance on German roads. The first auto manufacture to offer such software for real-life use. Audi and BMW. are closing in on Mercedes with their Level 3 software.

Footnotes:

(1) Waymo is currently operating its self-driving cars in several locations across the United States for various purposes. Some of the locations where Waymo is currently being used include real-life ride hailing services with no driver in the car. These cars have no steering wheels, charge for rides, and the revenue is already flowing back to $GOOGL.

Some of these Waymo test areas are still in “beta” mode and are not yet charging.

:pushpin: Phoenix, Arizona: Waymo is currently offering a commercial ride-hailing service in the Phoenix metropolitan area called Waymo One. The service uses a fleet of self-driving Chrysler Pacifica minivans and allows riders to book trips using a mobile app.

:pushpin: San Francisco Bay Area, California: Waymo has been testing its self-driving cars in the San Francisco Bay Area since 2009. The company has also been testing its autonomous trucks in the area.

:pushpin: Detroit, Michigan: Waymo has been testing its self-driving cars in the Detroit area since 2017. The company has been working with Fiat Chrysler Automobiles (FCA) to develop autonomous versions of the Chrysler Pacifica minivan.

:pushpin: Dallas-Fort Worth, Texas: Waymo has been testing its autonomous trucks in the Dallas-Fort Worth area since 2019. The company has been using the trucks to transport freight for several customers.

:pushpin: Kirkland, Washington: Waymo has been testing its self-driving cars in the Seattle suburb of Kirkland since 2017. The company has been using the area to test its cars in rainy conditions.

It’s worth noting that Waymo is constantly expanding its operations and may be testing its self-driving cars in additional locations in the future.

(2) https://www.cnbc.com/2022/10/26/mobileye-pops-more-than-30percent-in-ipo-after-spinning-out-of-intel.html

Intel spun off $MBLY last October:

Mobileye shares closed up more than 37% in their stock market debut on Wednesday after the maker of technology for self-driving cars was spun out of Intel.

In a year that’s seen no significant tech IPOs in the U.S., Mobileye offers investors an opportunity to get in on area of growth. But it’s not a new name for the market.

Mobileye was publicly traded before Intel bought the Israeli company in 2017 for $15.3 billion. At its IPO price of $21, Mobileye was valued at just $17 billion, resulting in minimal gains for Intel thus far. The stock, trading under the ticker MBLY, rose to $27.85 on Wednesday.

(3) Zoox website is pretty cool. “Built for Riders, Not Drivers.”

https://zoox.com/vehicle/

Bonus: Tesla Robotaxi bot on Twitter:

5 Likes

Not only will all of those companies compete head on with Tesla over the coming months but Tesla read Musk does not know what he is doing. Buying Twitter proves that hands down.

Is Musk more brilliant than the rest of us? Of course he is but that does not mean this year he knows what he is doing. Pay attention Elon.

Ah, the anti-Tesla echo chamber is back in session.

3 Likes

Exactly!

The Captain

1 Like

It is not anti Tesla to be realistic about the competitive landscape.

1 Like

Rob Maurer at Tesla Daily reveals the author of the article, a Tesla ex-employee fired for making unauthorized videos about FSD

In MacroNews…

The Captain

3 Likes

No the author is Faiz Siddiqui:

Since joining the tech team, he has focused on Tesla’s rollout of driver-assistance technology, labor and workplace issues inside the company and the decisions of its chief executive, Elon Musk. Prior to joining the tech team, he covered the D.C. Metro and local transportation scene, including, the system’s chronic safety and reliability issues.

Bernal was the fired employee.
Maurer a Tesla proponent poo-poo the accident data but states he & we don’t know the data only Tesla has the data. We don’t know if any of the multiple individuals [about a dozen] interviewed have a summary of the accident data. I would think that if accidents were less Tesla would proclaim that information. They haven’t so we, the public, are still in the dark.

Tesla & other EV makers have some mountains to climb.
Some buyers of the FSD system are cheesed off & have filed a class action suit against Tesla.
Tesla has a parts supply problem for Tesla vehicles involved in crashed. There are long wait times for those vehicles. And there is the EV insurance problem with an EV involved in an accident spelled out here:

An EV would be a valuable second vehicle used in commuting & shopping if one has a 240 volt charging system at home. But as a sole vehicle in the hinterlands [where I reside]…nah more time needs to pass. Especially if one is a late adopter of technology, as I am. I did not own a computer until 1995 or a cellphone until 2014. And I was late to online bill payment [2014]. I still get the paper statement for my credit cards. And I still like reading physical books, no kindle for me. I like to see a vast majority of bugs worked out on any purchased technology. Yep I is a fuddy duddy Luddite! LOL

1 Like

Owning a gas powered vehicle will likely get increasingly inconvenient. The number of gas stations has been in decline for a number of years now. There were over 200,000 stations in the US in 1994. That has declined to a little of 100,000 today. At least one analyst projects that 80% of gas stations will be out of business by 2035.

There will be some threshold of BEV adoption that when reached will result in a sharp decline in ICE related industries that will include gas stations, mechanics, auto parts etc.

1 Like

Thanks for the correction!

The Captain

1 Like

Actually, I have seen reports on accident data from Tesla multiple times over the years. The one problem with them is the question of whether Tesla with FSD can really be compared to the average car because of the likely difference in the driver.

2 Likes

You do seem to love to parrot BS you read somewhere. Tesla publishes its accident statistics, and has for years. It’s hugely better than average, and much better still when using its Autopilot. The link has been posted here many times. Rob Maurer’s comment about only Tesla having the data is in regards to having more detailed insight into how exact routes compare.

Some buyers of [anything you care to name] are cheesed off and …

No worse than anybody else. Probably better. Provide numbers for your bogus claims.

There is no Tesla insurance problem. The article is just more BS. Tesla is even available to provide insurance directly to about half its US customers at this point, and the number is growing as they expand the program.

The silliness about battery packs is just ignorant speculation. Batteries are easily recycled. Essentially old battery packs can simply be treated as high grade ore. Check out Redwood Materials for how it’s currently going.

-IGU-

4 Likes

IGU,

Thank you for the facts and analysis. The jury on those items had been out in my mind before your post.

That does not mean the competition wont show up very strong in the next three years.

Which doesn’t tell you whether Autopilot makes the cars safer.

Tesla’s cars should have fewer accidents than average even if they didn’t have Autopilot, because they’re not representative of the average car. They’re vastly newer than average - meaning less time for them to start suffering the safety declines that come with vehicle aging, like worn brakes and tires. Some of the biggest contributors to accidents are weather and teenaged drivers - with Teslas being geographically concentrated in fair-weather California and economically concentrated among people who can afford $45K-120K new cars, which will disproportionately exclude teenagers. To say nothing of the fact that expensive cars are spec’d out better than the average car - so being better than average doesn’t tell you whether Tesla’s any better than the average $45K-120K new car.

And as for Autopilot…well, it’s entirely up to the driver to choose whether to engage Autopilot or not. Which means that the results aren’t representative of average driving. The driver is going to disproportionately choose to engage Autopilot in circumstances that Autopilot will do well in (such as fair-weather driving on normal highway conditions), and less so in the circumstances where accidents are more prevalent (bad weather conditions, unusual traffic conditions or disrupted traffic patterns, complicated environments to navigate).

IOW, Tesla’s comparison of it’s average accident rate to the national average accident rate tells us virtually nothing about whether Teslas in general, or Autopilot-driven Tesla’s in particular, are an improvement over baseline. Which is one reason why Tesla is so, so, so far behind in getting autonomy to market. Other AV entrants are racking up the miles in official Level 4 testing (when the car is self-driving all the time, not when a driver picks and chooses to engage it), so regulators can actually assess whether they’re safer or not. Tesla’s losing years on getting regulatory approval by refusing to start doing that.

10 Likes

I agree with all your points on how we have no statistics that allow for a true apples to apples comparison. Maybe the problem, though, is that the national statistics don’t include all the required details, such as the car age, cost, driver age, etc.

I don’t see how your first sentence leads to Tesla being so so so far behind, can you explain?

This is factually false.
I took two different autonomous taxi rides in Las Vegas. The safety driver said he had to drive the car while in the hotel property (the harder parts of the trip, IMO) because their city approval only was for city streets and not private property. I don’t know about other cities, but the first and last hundred feet can require fairly difficult driving, while usually slow and not dangerous.

Mike

2 Likes

IGU,

Al’s analysis is even better, sorry.

Sure. Because they keep relying on the “apples to oranges” comparison, Tesla is years behind in compiling the data on product safety that will be necessary to obtain regulatory approval for a Level 4 or 5 autonomy assist program.

Cruise and Waymo have been operating fully-driverless Level 4 programs in certain California cities for several years now (Cruise just got permission to expand statewide). So they have years of “apples to apples” data - how the cars operate when a driver isn’t picking and choosing when to activate self-driving.

Tesla doesn’t have that kind of data. And I think it’s very unlikely that once the FSD system is “feature-complete” to Tesla’s internal satisfaction they will be allowed to just turn it on in a million cars. They’ll have to “show their work” to the government - which will mean doing the types of iterative programs to demonstrate safety that Cruise and Waymo have been doing for the last few years in California (not Nevada).

Given Musk’s disdain for regulators in other contexts, it may be that he genuinely is planning to just turn the system on once he thinks it’s done - counting on a million Tesla owners to pressure the government into letting them use the feature-complete system that cost them so much money. But I don’t think that’s going to be successful.

5 Likes