None of this is true.

I don't uncritically accept or defend what Musk says and acknowledge TSLA is a risky stock to buy.

Asking for clarification and sources for someone else's pov is obviously reasonable.

The head of AI and Autopilot (who specializes in deep learning and image recognition) said this at Tesla autonomy day:


Here's the video.

You're saying he is outright lying?

My understanding is Tesla clearly has a data advantage over Waymo and others. What is not yet clear is whether or not that will translate into a superior technology.

saying jaguar isn't profitable because they lost money last year is like saying tesla is profitable for it's 2 green quarters.

but yes, being part of a big indian conglomerate certainly helps.

that gives me a lot of hope.
i need people like you to throw away their money and pay for my porsche.

what other stocks do you like?

Yeah, that's a misrepresentation or wishful thinking. Probably the latter. There is zero chance that the steering wheel movements/braking of real drivers is of any use in training level 4+ autonomous driving. There's no precision and not enough feedback. Compared to the billions of miles of targeted simulation that Waymo and others are doing (they're able to simulate because precise detailed point maps are very different to images), it would be a joke even if they were using it (they're not, except maybe as a brain fart research project; it has no utility). Neural nets are trained with clear measurable definable goals through vast numbers of iterations, and there is no measurable definable goal in a random image feed (which they're not even uploading!!!!) plus steering wheel movements. Besides, driving is too complex to neural net train without precise data. Images are contextual things with vast variation unlike 3D maps of unambiguous point data for path finding that Lidar provides and more importantly can feed extensive rapid simulations to get tens to hundreds of billions of miles of training.

Andrej is an academic with little real-world experience. It's a classic case of when all you have is a hammer, everything looks like a nail. His hammer is neural nets - it's what he knows. But you can't neural-net learn autonomous driving without a near-human intelligence, which is at least a decade away. It just doesn't work. I'll write out why if I can be bothered to explain in detail.

Neural nets as they are today are fairly good for categorizing objects, spoken words, even traits in pictures. They're good for learning precise games that have precise, finite feedback and goals. They're nowhere near the speed and sophistication and low level morphing hardware required to do higher level tasks that require at least a 70 IQ to determine paths and situations and process unreliable vastly varying images into meaningful maps, such as driving.

The only sane approach to level 4+ for the next 10 years or so is precise 4D mapping tied in with visual object recognition and hardcoded edge cases. That's what the guys who've achieved actual level 4, like Waymo and GM/Cruise, have done. There's no other path to get there from pure images until neural nets are capable of near human level intelligence, which requires much better hardware and far greater sophistication on learning algorithms than where we are today.

If you'd watched the Tesla autonomy day and have some coding/design knowledge, it was absolutely cringeworthy. I felt embarassed for Musk, for the team.

Used car prices are way higher than that. But depends which version he got (Performance or not). Even without performance, used prices are close to where his residual value is.

This has to be a joke, right? Jaguar Land Rover is not profitable at the moment and has been struggling for a longer time now.

it's been struggling for about 3 years, despite being profitable every year besides the last, and what happens to the company depends largely on what tata wants to do with it.

i know another car company that's losing money every year but somehow will be a trillion $ company.

Waymo has shipped around 1000 and has orders for 82,000 more. Those 1000 are level 4 who can handle all city and highway traffic conditions with one disengagement less than once per 13,000 miles and dropping rapidly. This statistic is proven in real world traffic in San Francisco, a rather challenging environment for a Western country.

Not sure about GM/Cruise. They have a fleet as well.

So yeah, Tesla are getting the **** kicked out of them by Google and Waymo on deployed level 4. Recap:

Autonomous level 4 cars deployed:

Tesla: 0
Waymo: about 1000, moving to 80,000 soon
GM: 100-300 I would assume.

Level 4 miles tested on public streets in the last two years:

Tesla: 0
Waymo: 5 million with 10 billion simulated
Crusie: 1 million, unsure on simulation

Working, fully autonomous, phone-hailable taxis from anywhere to anywhere in a city:

Tesla: 0
Waymo: around 300 for members of the public who sign up
GM: 0 public, but around 50 for employees.

Last public disengagement data available:

Tesla: Once every 3.2 miles
Waymo: Once every 13,000 miles and rapidly dropping
Cruise. Once every 9000 miles and rapidly dropping

Just to put that in perspective: Tesla has 4000x more frequent disengagements than Waymo in the last objective data we have. And how can you be working on releasing level 4/5 next year and have a million autonomous robotaxis earning money if you haven't even started testing on public roads in the state where your autonomous driving team is??

You didn't answer the question. Ship = deliver, not order. Waymo did not deliver those vehicles, it ordered them for its hypothetical future fleet.

According to the article you linked, Waymo is considering selling to individual customers in the future, but has not yet done so.

So the answer appears to be zero.

Also not clear where you're pulling these stats from...

decapito is being deliberately obtuse

How can you possibly equate purchasing vehicles from Chrysler to building and delivering vehicles to paying customers? It is literally the opposite thing.

doesn't that make them smart?

why is elon selling these cars for $50k now when he thinks they are worth $150-200k 1.5 years from now?

shareholders should sue him, that has to be one of the biggest value destructions in business history.

And just to give a little color on why the hardware in current Teslas is lolwtfareyoucrazy nowhere near suitable for autonomous driving yet, here's a feed from their cameras



Look at the camera on the right getting completely ****ed up by the sun on it. How are you going to do left turns into potential pedestrians with ONE SIDE CAMERA that gets its images ****ed up by the sun? You gonna rely on that to see pedestrians? ONE CAMERA that suffers bad lens flare?

In contrast, here is what GM/Cruise can do in busy San Francisco traffic that even human drivers don't enjoy/have trouble with, thanks to their precise 360 degree point map:



Look at the 18 second mark. How is Tesla going to do left or right turns with pedestrians and bikes coming alongside? It can't even see and identify them properly with that single camera and low image quality, let alone when light flares on it. Its ultrasonic sensors are worthless for complex situations with multiple pedestrians and bikes and cars going in different directions or at a distance, for example as it speeds up taking the left turn and has to decide whether to go faster or slow down.

It's obvious that Musk is a liar and a fraud. It's sad descapito that you are so butt-stupid. Try to work on that.
Yeah, it's basically an admission that I'm right, he has nothing to say but he has to come back with something because his ego is bigger than his intellect.

The company has to either (a) achieve breakeven/profitability or (b) rely on debt/capital markets. Achieving (a) has a lot less existential risk, so it's immensely valuable. Selling vehicles now helps to achieve (a).

Also, the time value of money means [x] now is worth more than [x+y] later.

From the video above: the view from their single side camera on which life and death death driving decisions will be made when turning, changing lanes, etc:



Oh and they're going to use machine learning on this high quality image to decide what to do.

How much of an assclown/sheep is despacito to believe Musk when he tells pure bull**** such as "all cars produced now have the hardware for Fully Self Driving"? The above is open and shut proof that they don't. He's saying that to sell his FSD vaporware. He is obviously and transparently lying.

Comparing low-resolution videos from Youtube is not a credible way to prove the current and future state of either company's technology. You're reaching.

Right, which is why it's great we have to compare:

- Actual fleets operating on public roads
- Public disengagement data

Which back up the video completely. You've been so transparently owned by your own stupidity at this point that you're reduced to merely weakly trolling, because your ego can't handle the fact that you're a total idiot (which, ironically, is the probably the reason you are).

the time value of money is negligible in this case. we are talking about the end of 2020 when he has promised robotaxis.

why do you think he is not bridging those 1.5 years with financing from the market if he can sell all his product at ~80% margins after that?

I assume you mean debt finance.

How would they service the loan with no revenue?

Where is your public disengagement data from?

DYOR is what you should say to someone who is asking for other people to feed him conclusions.

In this case I'm asking for sources so that I can form my own opinion about the validity of the data and OP's interpretation of it.

Can you understand the difference?

you just structure a loan without payment for 1.5 years?

or sell stock for whatever amount gets you there.

this is a no-brainer if you have 80% margins on billions of revenue after that.

idk enough about debt markets to know if that is true. Intuitively it seems like this should be really hard to do for an unproven technology which is obv very high risk. Do you think Tesla could raise enough debt to get it through 24 months with zero revenue?

of course they won't get that money because their claims are ludicrous.
that's kind of the point i was getting at.

He interned at Google and worked at Open A.I. Sure, Open A.I. is a non-profit, but they're doing cutting edge research including practical applications such as winning vs humans at Go and DOTA. Hardly dry academic stuff.

It's also ridiculous to claim that having a strong academic background is a disadvantage in this field. The founders of Google were both doing PhDs at Stanford.

So in other words you think selling vehicles now is perfectly reasonable given the difficulty of raising sufficient debt to ensure the company survives long enough to achieve FSD.

I guess we agree then.

btw I don't know if the income projections on robotaxis check out, there are a lot of variables on that one, even if FSD is achieved on schedule (competitive risk, for one).