Whitney Tilson’s email to investors discussing that Audi eTron is cheaper and designed to compete with the Tesla Model X and Jaguar i-Pace and a debate on Tesla Inc (NASDAQ:TSLA)’s autopilot.
Debate on Tesla’s autopilot
An interesting debate among my readers (plus me) about Tesla autopilot:
One started it by emailing me:
40-70 years?? Do these people not understand the power of the neural networks at use here? After hours of training, AlphaZero destroyed the best chess engine in the world. The improvement from v8 to v9 of the autopilot has been incredible in just a few months. Sure it has a ways to go, and there’s a big difference between Tesla achieving level 4 autonomy vs. level 5; but, when has saying “Elon can’t actually do this bit of engineering” ever worked out?
I agree that it’s more likely to be 4-7 years, not 40-70.
But Musk has been making promises he hasn’t kept nonstop for years… A sample:
- No more capital raises, ever (at least since 2012).
- "Three months maybe, six months definitely" -- pick almost any software update ever.
- "Full self driving" (always 9 months out).
- 5,000 per week (the imminent goal every quarter for 8 quarters straight).
- 2% interest rate convert (2013 and 2019).
- Massive service center expansion (always, usually a post-Midnight tweet).
- Superchargers in Ukraine, Mongolia, Montana, Zaire -- always the imminent capex expansion plan.
- Service should take an hour or two instead of a quarter or two -- also the imminent plan to fix.
- The next car will finally be the profitable one (Model S, X, 3 and soon the Y).
- Every quarter from here on, will be Free Cash Flow positive.
- Every quarter from here on, will be profitable.
- Orders are up (on every earnings call, ever).
- Manufacturing quality is up (on every earnings call, ever).
Of course Elon lives on EST. He has not hit, like, any deadlines. But he does achieve the engineering, eventually.
The criticisms here aren’t “the stock is too expensive”; they are “Elon can’t do self driving”. That’s dangerous in my opinion. He gets the business wrong, but he gets the engineering right.
It depends on how you define “doing” self driving.
The danger is the he’ll roll it out when it’s 99.9% effective – when it needs to be 99.999999% effective.
There’s a HUGE difference.
And I have ZERO confidence that Musk will be appropriately conservative in rolling this out – meaning, a lot of people could literally DIE due to this guy’s delusion/narcissism/desperation.
Yeah, but how many 9s effective are humans? In theory it should be rolled out the second it’s better than a human driver, but in practice that would cause “too many” accidents for the public to tolerate and won’t happen. I can’t imagine it’s rolled out before it’s better than a human driver. I imagine Elon would roll it out a couple 9s before GM would, but they’re both rolling it out too late thanks to the public not actually understanding risk, like, say, nuclear vs. coal.
Another friend (2) weighed in:
If you're at 99.999% accuracy -- however that's measured -- you're practically nowhere when it comes to being socially acceptable. Would you put your 4-year-old only child in the back seat of one of these, knowing that Elon just pushed out an A/B beta test for a new software implementation what makes the car a little more aggressive in oncoming traffic on one-lane narrow streets?
And besides, what happens at the other end if said 4-year-old forgets to close the door? Does the Model 3 just sit there, until help arrives? How many bums in San Francisco will enter the car in the meantime to use as a drug lab, followed by going to the toilet in there?
Even if Tesla had some advantage in this field at this point, it's about as relevant as having a backyard trampoline and claiming that you're now closer than your neighbor in terms of catapulting yourself to the moon.
Neither Tesla nor any other company will have a truly self-driving car, in the relevant sense of the term (being as good as a human driver) in several decades at best, perhaps ever.
And if they would ever get there, the contraption would swiftly be banned as matter of national security. We cannot have 5,000-lb. remote-controlled bombs floating around American society, potentially under the influence of bad people/regimes. What happened to drones will of course happen to automobiles: They'll be severely restricted, if not outright banned.
There is no way a country of nervous nellies who ban peanuts on airplanes are going to accept having robotaxis running around neighborhoods killing people, including their kids.
Another (3) wrote:
Musk’s strategy running the company is “what do I need to survive”? Then he announces it, fires people if they tell him it can’t be done, slave works everyone else. Reality doesn’t enter the equation. At this point he seems to believe that the act of announcing robotaxis is equivalent to having them in a year. Very bad outcome is expected.
To which (2) added:
Indeed, and he had the same announcement already on October 16, 2016:
“Basic news is that all cars exiting the factory have hardware necessary for Level 5 Autonomy so that’s in terms of cameras, compute power, it’s in every car we make on the order 2,000 cars a week are shipping now with Level 5 literally meaning hardware capable of full self-driving for driver-less capability.” (transcript)
People forget, that when it comes to Tesla, everything that's old is new again. The company keeps announcing and promising the same thing over and over again, with a very low hit rate.
Of course, some things are real. The Supercharger network has been the best of its kind -- even if, finally, Electrify America is showing proof of being on a strong path to catch up -- and of course the cars are real and look good, while being able to accelerate 0-60 MPH so quickly that it makes the fatal moment at least quick and perhaps therefore on the less painful side.
Tesla's autopilot video
From a friend:
Tesla tweeted out an autopilot video that’s on their website: Tesla.com/autopilot
This is their show-off video. Pls go to 1.23-1.24 into this video. There is a guy walking a little white dog. The car absolutely does not “see” the dog. You can tell by the fact that the middle camera view doesn’t put a box around it. One foot to the left and squished dog.
Another friend commented:
Round numbers, humans collectively are involved in one traffic fatality for every 100 million miles driven. A Tesla could not go, on average, more than 10 miles without a so-called "disengagement" -- a human having to take over, in order to prevent a probabilistic accident (perhaps not fatal). We know this from the people who were privileged to try out that pre-selected route on Tesla's "autonomy investor day" on April 22. Maybe if the route had not been pre-selected, the Tesla would not have lasted, on average, even one mile without help.
In any case, this indicates that Tesla is somewhere in the neighborhood of being close to 7 decimal points away from replicating the safety of the average human driver.
I'm thinking that this is at least 40-70 years away from fruition, if even then. Given the sheer orders of magnitude involved, that's probably way optimistic.
I watched the 24-min video (link below) of the review of the Audi eTron. A very cool car!
From a friend:
1) An incremental positive for Tesla today is that the deliveries of the Volkswagen ID.3 seem to have been delayed from around January 1, 2020, to the middle of 2020. If true, that’s disappointing. Manufacturing is still supposedly starting this November, 2019 -- so why the delay in deliveries until closer to mid-2020 as opposed to January 1, 2020? I don’t know -- yet. Maybe this is false alarm. But if true, an incremental positive for Tesla, in terms of the Model 3’s competitive climate in Europe, which is where the ID.3 is scheduled to be sold in 2020.
2) Doug DeMuro reviews the Audi eTron, and it gets almost as high a score as the Jaguar i-Pace:
Doug’s review may not be my #1 favorite one, but he’s important because his YouTube channel is huge. This video already has almost 1 million views in its first 24 hours, and over 4,000 comments.
You may also recall that Doug had what may be the first independent review of the Tesla Model 3, and that his reviews of Tesla cars -- at least until recently -- have been very positive. That may have changed now that the Jaguar i-Pace and Audi eTron have shown him and others that they are extremely competitive with Tesla Model X and S -- but at least he used to be called a little bit of a (mild to moderate) Tesla fanboy until the last few months.
3) One more thing: First drive reviews of the Mercedes EQC are getting very close. Could be within a couple of weeks. It’s also very similar to the Audi eTron and Jaguar i-Pace. It will bring further competitive pressures to Tesla’s already declining Model X and S business -- first in Europe 2019, then in China by early 2020, and finally in North America later in 2020.