Even if this was true today, we're clearly talking about tomorrow. FSD is a prospective technology. Not only do you not know the state of the art in this field, it is changing every day, and the rate of change is increasing. This dynamic always leads to people underestimating what is possible over year+ timeframes, including people who have some knowledge of the field (which you claim to have).

You've repeatedly stated ITT that the tech is complete, in testing, but not shipped to customers. You've also repeatedly stated that FSD and a fleet of robotaxis is a 100% lock to happen by the end of 2020.

Now you're stating that the tech is unproven and very high risk. Which is it?

if you can make 80% margins on billions of revenue 1.5 years from now you definitely can get that money.

drum up some private equity firms and wallstreet banks and show them that you are basically there today - which, let's be honest, you would have to be to get to 100% 18 months from now. drive them around in your autonomous cars for hours in the snow/rain/night/inner city.

tesla could easily get that kind of money if they could do this.

Do I believe FSD will ship by end of 2020? Yes.

I don't think it's 100% complete yet. There are clearly edge cases that are still being worked out.

As for "a fleet of robotaxis", there is already a 'fleet' of Tesla vehicles in operation and this only increases in size with each new delivery. So the fleet part is a lock unless Tesla sales plummet by a huge multiple, which I see as super unlikely.

EDIT: if Tesla doesn't deliver on FSD there will be a lot of damage to the brand and future sales.

why haven't we seen a tesla navigate san francisco, new york or la rush hour traffic? how about minnesota snow?

are they going to test all that in 18 months?

Edge cases? You mean like snow, parking lots, or recognizing a ****ing truck is across the road?

Others are definitely better qualified than me to address issues relating to debt financing.

There are various reasons why it would be advantageous for Tesla to sell vehicles: (1) revenue; (2)Tesla's approach to FSD is heavily reliant on training data and having a fleet of vehicles in real-world conditions provides that; (3) building a world famous brand with a cult-like following, similar to Apple (I have no idea how to value a brand but it sure has successfully built that already).

I am aware that some ppl ITT (1) disagree with Tesla's approach to FSD, and (2) claim Teslas do not transmit significant data, but (1) is not relevant to my point and (2) I disagree with.

Question for you: why is Waymo seeking to raise outside capital? Its parent company is cash rich and according to some ITT it is much further along the path to FSD than any other copmany?

have you looked that up like i told you to? how much are people on tmc reporting their car is sending back on average?
how many seconds of video of the car's surroundings (every camera) does that equate to?
because they can and - in contrast to tesla - people with big money are lining up to throw money at them.
if google internally values waymo at, let's say $50b, and people are lining up to pay $5b for a 5% share, that's awesome.

have seen stats on tmc, also the data available from Teslas and what is required to train the ANN is addressed somewhat by Karpathy in the autonomy day presentation I linked earlier.

Tesla's approach not 100% reliant on customer data broadcast, but it does contribute to the training sets and is useful.

I think it's fascinating that there are such strongly held opinions about the best (or only) path to FSD, when a bunch of really smart people are pursuing competing technical visions.

Here's a transcript of part of the Q&A session which touches on the volume of data issue...

It's not really fascinating. It's Tesla's 2nd rate engineers vs every single other team that uses Lidar despite the cost.

It's a bit like Musk thinking he could outdo the majors and robotize his entire factory. He nearly destroyed his company doing that, and anyone sane could have (and did) tell him that it was a terrible idea. He didn't want to hear that because he's a grandiose narcissist who's the smartest guy in the room with grand visions. It's the same thing here. Musk is incompetent at complex engineering. He is a big grand thinker who lacks the genius to see detail, so he gets himself into these dumb tangles. His dumb brain thinks "humans can drive with vision, why do we need lidar?", and that's the end of it. It's a bit like thinking "robotizing the whole line is theoretically possible, so why wouldn't we do it?", even though that makes him a dumbass incapable of understanding complexities and running a manufacturing line.

With this philosophy he's drawn the theoretical but impractical type of employees into his autonomous driving team. Which is why this is the state of his technology:

descapito, try reading the last few months of this thread rather than tarding the thread up with your dumb questions and opinions.

And here again is the hardware keeping track of 40% of the field of view of the car, which Musk claims is sufficient for fully self driving:



Tesla are dead last on autonomy and Musk is a shameless liar who's using "2020 autonomous robotaxis" and the eager help of level 5 dumbasses like yourself to pump his stock and keep investors putting money in and customers buying FSD vaporware. It's that simple.

Quote from autonomy day...

despacho - Perhaps if you could cite an "expert" that's not Musk. Or anyone associated with him.

that's just gullible-investor-speach for:
'we are telling you this fancy story about the hundreds of millions of miles you drive to train our model which is this enormous competitive advantage that no other company has.*'

*we don't actually collect or use that data in any meaningful way, nor is it any kind of advantage since we're years behind competition. please buy our secondaries.

despacito, why are quoting a known liar and fraud with a 100% autonomous driving prediction failure, on autonomous driving? Seems weird to choose the least reliable source you possible could.

And yes, like I said, Musk is a grandiose narcissist who's not very bright. The quote is a perfect example of that
It's not expensive any more. It's on a production Audi and more will follow. Costs have dropped dramatically and will continue to. It was expensive some years ago.
It's certainly not unnecessary. Right now it is the only path to reliable level 4, as Google and GM have achieved.
And there's the crux of it...once you solve vision.

I'll agree with Musk that if you plugged vision into a human level AI, it could drive just fine (although not on Tesla's current hardware - see picture above). The trouble is, we are so far from solving vision and the situational awareness needed to make sense of vision that this is a decade away.

This is just like Musk saying if you solve the problems of how to make robots work on the harder parts of a car, humans are worthless and I'll get costs down so low and production so high that Tesla will soar. And the dumb cuck literally tried to do that despite any competent person knowing that robot technology couldn't handle that right now (which is why none of the majors, who are well ahead of Musk on robotizing with billions more in R&D funds, don't use it). There's a lifelong pattern of mindless simplistic thinking from a guy who sucks at detail.

Sure.

https://www.mercurynews.com/2019/06/...it-whos-right/

Cliffs: former head of Waymo now says Lidar is unnecessary.

What I took from that quote was that a small startup is farther ahead in fsd using a Prius and aftermarket cameras Than Tesla is.

Prius w/ aftermarket cameras and computers drives cross country. Tesla can't back out of the garage.

i'm not super knowledgeable about this but here's my high level take.

- the approach tesla is using for autonomous driving to me seems like the same approach that bachelor's level comp sci graduates use for kaggle problems, or someone who just discovered the scikit learn python package. it's like "get lots of data --> machine learning bro! --> there's ur answer." all the uber tesla bulls who post about it seem to have ~this level of knowledge, just enough to sort of understand the very basics of ML, and they talk about it like this is such an obvious answer that's going to work. certainly other companies with phd+ career ML ppl have considered this?

- how much money is tesla actually putting into this research? it can't be that much, right? most of their money certainly goes towards making and selling cars and not going bankrupt (and gf3)

- 10% of their autopilot team (is this also the FSD team?) just quit or got fired. isn't this a substantial setback?

all of this + the fact that it's tesla choosing one path vs. 4+ other huge companies with more money/engineers all agreeing on the same other path. to me it just seems like there's no way tesla is ahead. the only way is if karpathy or some other engineer tesla hired miraculously turns out to be an einstein-level genius.

edit: also it's google FFS. you think a car company is able to create better software than arguably the best software company of all time? don't @ me

Xkf, yeah pretty much. And no amount of human genius will be able to machine autonomous driving from cameras. It's just not a solvable problem with current hardware and the lack of sophistication in the machine learning field. This is one of those things that are iterative 10000000x (make, test, improve tools, make, test, improve tools, etc) rather than an area where a giant breakthrough will happen. Most hard tech is slow-complex-iterative rather than breakthrough.

The guy descapito quoted above appears to be a thief with just about zero credibility. He's the guy who stole a bunch of data from Google and sold it to Uber who then fired him when they "found out".

He made this claim just as he launched a funding round for his startup that doesn't use Lidar...he's about as reliable as Musk on this stuff.

I quoted Thrun in bold/underline. He led Waymo. Convenient how you completely ignored this and focused on Levandowski (who afaik is hated by Thrun). This is a disgusting misrepresentation of (a) my post and (b) the truth. It makes me sick just reading it; you must have felt super scummy writing it.

https://www.newyorker.com/magazine/2...ctual-property

despacho - I can't believe you took the time to edit your name when you quoted me. And before you ask, I took the time to spell it wrong (twice!) for my own amusement to see what you would do. I wasn't let down!

Disgusting is a bit strong. From the article, Thrun gives as his sole quoted example of this technology working a PR stunt by a known thief promoting his startup.
Says it all really. Since it's his sole quoted example, the example in question is all that matters. There are no verified examples of this tech working even on level 3. The sole production car that's at level 3 is the Audi which uses Lidar. Level 4 cars all use Lidar extensively. Tesla fails dismally with just cameras even on highway driving level 2 despite working on it for years.

We're supposed to believe that something 5+ years behind is suddenly going to explode to level 5 next year despite zero examples of this tech working and the cameras themselves being incapable of dealing with things like rapid light changes, very poor low light resolution and sun/night light flares?

The only question that matters is: does the technology work? Thrun says yes to FSD without Lidar. And he started using Lidar at least 15 years ago AND led Waymo.

This Levashmowski whatever is a scumbag, but that's irrelevant. This isn't ****ing Sunday school.

It's hilarious that you continue to rely on the A8 example which is not permitted for use in the US and/or too expensive to scale.

Yes that is the only thing that matters. Since there are no working examples of even level 3 with just cameras, the end.

I do too. It's theoretically possible. It's just not practically possible with current and near future technology. A read of the history of AI predictions by very smart experts is pretty hilarious.

What do regulations have to do with the only production level 3 car using Lidar? Put down the bong son.