Really didn't know anything about Tesla, still don't, but saw the news about the Model 3 today and read that Tesla lost something like $2.5B last year and still it's stock continues to rise.

What other companies get such high amounts of public funding with such poor monetary results? I couldn't think of any.

And I get that it's all a bet on the future with Tesla, but jfc $2.5B every YR... How long can you do that for really?

This is hilarious:
When are the people in this thread going to admit they got mind-****ed and lied to by their cult leader, Musk? You all argued vehemently with me and claimed Tesla was far ahead (I claimed they were far behind), and now not a peep out of you clowns.

Musk is less accurate on understanding autonomous driving and its challenges than is credible for someone intelligent. Which means he's either a moron who doesn't understand the basics of tech and difficulties that can arise, or deliberately lying to you, his investors. A year into losing MobileEye, they still can't do automatic braking!!!!!! Or reliable highway driving at speed - long solved problems. And this despite promising AP1-superior a full six months ago.

When do I get an admission from you guys that you got sucked into the Musk lies, like silly schoolgirls with a rock star crush? Including heltok, who claims to work in the field. How did you all get it so wrong?

Oh and in case anyone who doesn't read the Tesla forums wants an update, this is the state of Autopilot 2, a year after breaking up with MobileEye:

Posted yesterday: dangerous truck lust is still there:
And the next one is hilarious.

Tesla programming is so cuck-level stupid it doesn't recognize how and when to disengage safely. Its line scan is extremely limited and it can't recover from anomalies - in fact it does so in a way that's absurd and potentially fatal.
Some drivers don't have these problems - they have concrete barrier affinity instead, which I'm not sure is better than blind lane swerving or truck lust:
Yet the brainless gentlemen in this thread think Tesla is ahead in autonomous driving. Or maybe they don't think that any more and are too embarassed to admit it after they argued so vehemently for the opposite.

How do you think this will go when this level of technology gets out of the hands of doctors and engineers and wealthy sports car enthusiasts, and into the hands of the lower end general public, which is happening very soon?

You all were fools holding at $380 and you're fools holding now at $350. Thank God for Trump and the market rip (large cap tech +40% over that period), but it's over now. Tesla are not competent, not capable of meeting guidelines, not safety conscious, and horribly inefficient and hence incapable of turning a profit. None of that mattered while they were small fry producing $100K cars for rich beta-friendly enthusiasts and able to go to the market for the (comparatively small) amount of capital they needed to sustain that. Those days are done.

Musk himself is terrified. He's done away with normal testing procedures in order to get this out as soon as possible and keep the stock afloat. This is for a company that royally ****ed up the Model X under Musk's silly design, costing them 6+ months and a lot of money. Now they're doing an unprecendented ramp without proper testing:
This is desperadoville, and you're all on the train. If they majorly mess up or miss something important in the Model 3, the stock is worth zero. If a recession happens, the stock is worth zero. If sales drop thanks to the 200K being hit and hence no more $7500 subsidies while their competitors have them, the stock is worth zero.

If they execute the Model 3 perfectly to plan AND the stock market doesn't correct substantially AND the hype continues, the stock is worth what it is now.

TS,

what you get wrong about us cult members is the fact that we don't believe 100% of what Musk is saying but realize that even if he only achieves 33% of what he claims, it's going to be a milestone and the company will be worth more according to it.

They have problems, they will continue to have problems and there will be mistakes. I think Musk is indeed pushing forward and probably making (possibly large) mistakes in the process. It is all relative though.

You see numbers and that's it. You are in love with Ford, I can't fathom how you could possibly be so bullish on a company that recently fired its CEO because of the state of the company. You can only see the short-term. I don't think it's at all relevant if Tesla delivers 1000 cars more or less and if they are testing new things with regards to the battery production then I think now is absolutely the time to do it. They have time now. If they make large progress now, they can scale that in the future. You seem to think that there is no path dependency and it doesn't matter if you do something now and scale later or scale now and change something later.

That's also the reason why you think the large car manufacturers will be leading in the EV revolution (if it comes). The largest smartphone manufacturers were not the largest cellphone manufacturers. The same will be true for EV cars. It will not matter to you because you trade short-term derivatives but it's silly to constantly attack people because they have a way longer attention span than you have and are a bit more patient.

I really believe this quote is correct:
That's why I don't think it matters all that much if they fail on something that they claimed two, three years ago. If they fail, they'll fail. At least they've tried and that is more than Ford has done in the last 30 years.

Spurious,

Actually that isn't true wrt phones. Large outsourced contractors made a bunch of cellphones and still make a bunch of smartphones.

Good try tho!

Ok, we can now argue about the IP and the impact of that on manufacturing. They might be assembled in the same place but the majority of profits go to someone else.

It doesn't matter though because my point still stands and it's perfectly clear what I meant. Nokia is not Apple, Blackberry is not Samsung, Ford will not be the market leader in EV.

comparing the advance in smartphones to the use of EVs over gasoline cars has to be one of the stupidest things written in this thread.

phones took a giant leap in core functions in the last 15 years, allowing people to take photos, listen to music, use the internet, watch videos. the usage is totally different to 15 years before.

a cars core function doesn't change. EVs get you from A to B. gasoline cars get you from A to B.

Spurious,
Tesla are far behind in the EV "revolution", despite a large head start. They're not leading anything. This is Europe:



This is China:



The only place in the world where Tesla has done ok is the US, where they captured the high end electric sports car market (in which they have a monopoly thanks to the vast sums of money you have to lose to produce a high end electric sports car), and have been able to take advantage of $7K tax credits.

Tesla are getting crushed by the pack in the electric vehicle revolution. And they're getting crushed while getting large subsidies (soon to run out while their competitors will have it for a while), and losing vast amounts of money (something that's not viable at a higher volume).

Just wait until ICE cost parity is reached and it becomes viable to the majors to manufacture EVs at a large scale.

Here's what you may not understand. A car is VERY different from a phone. In cars:

- Cost matters to the consumer a great deal, as a car is a year's pay for most people, as opposed to a week's pay for a phone, and the low end with tight margins is where the volume is.
- Up and comers are at a huge disadvantage in terms of reliability and cost and supplier networks
- Car manufacturing is highly capital intensive. Ford needs $150 billion in capital to produce 5 million cars. Phones are far less capital intensive
- EVs will look identical to current cars, just with a hidden battery powering them instead of a hidden engine. Unlike the switch from regular -> touch smartphones, everything up top will be identical. How do you think Apple would have gone if they'd merely provided a new power system, but their phones weren't otherwise differentiated from the major smartphone manufacturers? Nowhere is the answer.

The comparison with smartphones has no foundation.

TS,

Even the European numbers look quite good to me. In China they need to catch up, what they will. The 25% (or however high that number is) import tax obviously hurts them.

Spurious,
The point is they're a small fraction of all EVs sold. They're not ahead of anything, not even close. They've occupied an expensive electric sports car niche in which no one thought it viable to compete due to the large amounts of money you had to lose do so, for no real gain. They've basically used their investors' billions, and public money in the billions, to subsidize sports cars for rich people.

Tesla has no chance of getting a foothold in China. Absolute zero.

All of this revolution talk is crazy. Musk has sold himself to fools as the leader in the "electric car revolution" to save mankind.

He also sold himself as the leader in the autonomous car revolution.

The second has been exposed as pure bull****. Why hasn't the penny dropped that the first is pure bull**** too?

Musk's actions will make no difference to the speed of adoption of electric cars. It will make no difference to the environment. Electric cars will happen in meaningful volume once cost parity is reached with ICE, which depends on battery tech, which has nothing to do with Musk (battery are a mutli-tens-of-billion industry that's growing exponentially even without EVs, of which Musk only takes a fraction of anyway).

Musk is a fraud, out to make himself famous and rich, and you're swallowing his pure bull****. It's kind of sad. Tesla aren't ahead of anything and they're not producing anything of worth except toys for some rich people. THeir odds of ever reaching GMs profits/production numbers are extremely low. They're selling you nonsense, and you're swallowing it like the Trumpsters swallowed Trump's nonsense.

They start producing a mass market car that has already reservations in the hundreds of thousands. We will have to wait a year to see how this goes but that will show where they are.

If they execute well they will have hundreds of thousands of cars on the street that collect data for their autonomous/assisted driving program which will in turn dominate everyone else who even tries to start something.

Agreed.
This is nonsense too. I can't believe so many otherwise intelligent people have swallowed it. He truly has Kim Jong Il level of reality warping.

The data collected is worthless. Why? Because, as I've said long before it was realized by the people here, autonomous driving is about solving high bandwidth image processing and object recognition and surface mapping in real time. You need multiple high resolution cameras pushing through huge bandwidth video feeds to a processing unit like the nVidia Drive PX. There simply isn't the bandwidth to upload even a tiny fraction of this data. What Musk is supposedly collecting is worthless. It's likely just mapping data to notice issues at certain map points and avoid them, which is a bandaid on a broken system and has nothing to do with level 4 or 5. Read the Tesla Motors forum quotes a few posts above to see how well it's all going.

TS on robotic vision is what is truly worthless.

How exactly do you acquire the dataset to drive feature extraction (which is an extremely low number of features compared to theoretical possible which is <pixels>!)? Oyah, collect a **** ton of data.

That being said, their algos are way way way behind. Some of the worst actually being tested in Cali on public roads. Unless you believe they are intentionally going on public roads and intervening as some part of weird disinformation campaign. They have a solid structural advantage, but their actual technology is much much farther behind than anyone thought.

I mean, lets take this jackassery about "high bandwidth image processing and object recognition and surface mapping in real time". The Drive PX2 is about as fast as the new Titan card. Not dual, not quad, a literal single card. And its expected out late this year. So it'll probably be slower than multiple cards.

If the problem is really requiring higher end cameras and hardware, why are they using cheap lowend hardware and devices? Surely google must have someone who knows how to rig a couple graphics cards together? Or maybe someone can afford an 8k video camera? (Or atleast a few HD ones)

The narrative you put forth isn't even substantiated by the facts, which is that the cameras and hardware are low end. The real issue is simply rejoining the datasets into simpler algorithms that can be processed on these devices without creating massive power draw so they can actually work in a real car (EV or not). Optimizing this is the real issue which is powered by datasets, the kind that Tesla will be able to acquire for free.

Yes I've been right and you've been wrong - both on theory, practicality, and where Tesla is at. How does that work?
You're assuming that this pared-down machine learning data can be combined. I believe that it likely can't. There's no reason to think machine learning can be parallelized across different units (it probably can't, humans aren't that smart).

What you'll eventually need to do is capture raw image data of every possible variant and feed it into a continuous training stream. There's no reason why 100K cars is gong to do that better than 100 if you can't upload the ultra high bandwidth image streams.

Deep learning, by its nature, very likely isn't parallelizable. Computer vision is:

- Object recognition - this can be deep learned from image streams within one unit
- Surface and object position mapping
- Creation of a 3D world model through which paths are found. This can't be deep learned at present
- A decision engine about which branch to take from available paths (this can't be deep learned at the boundaries, imo, at least for a decade or so).

Tesla's data isn't going to do a damn thing for any of these. How is uploading data from the Drive PX unit going to help them with these? It's not.

The narrative that Tesla will "deep learn" how customers drive is pure bull**** to fool morons. It doesn't work like that.
FYP. I mean, I actually told you and others in this thread that it was OBVIOUS they were very far behind. You and others thought that was laughable.

It was YOU who didn't realize they were far behind, not "everyone". Own it, dude.

Because the optics still aren't there.

Yes, you can get by just fine with a 720p camera on a highway in good conditions with nicely marked lanes and signs. A piece of cake. In fact, you can probably make it work with just one camera and a low end graphics card.

Try that same camera in poor light, with rain or snow, or with light conditions changing (exiting a tunnel, etc), or requiring subtle color difference or hard-to-see boundary detection (a light color truck vs the sky, for example, which is what caused the Tesla Decapitation in Florida) and the whole thing goes belly up.

It's the boundary conditions that matter, and increasing resolution by a large amount is the only way to solve it. There's a reason cameras - even high resolution cameras - increase ISO in low light. It solves a problem. Guess what's also the same as higher ISO? Higher resolution. Current resolutions, even 8K, aren't up to the task.

I mean, let's take our cues from nature:
Get it? Higher resolution eyes with higher density low resolution sensing equipment in them solve this problem for cats. The same will happen with cameras - multiple sensor units in the same chip at much higher resolution than today, which can handle low light, flares, etc. What we have now in terms of cameras is inadequate for the task in boundary conditions.
They're low end because the difference between low and high isn't large, and no one is trying to solve the boundary cases right now (snow, terrible light). Google relies on LIDAR, hence high resolution isn't needed.
I do not think combining deep learning data sets is going to be a solution to autonomous driving.

I almost spit my coffee out when you said that machine learning can't be parallelized. It obviously can. Try actually doing something with tech before talking about tech? The Drive P2 has multiple GPU and SoCs.

That was quite the riot. This is exactly what I mean by being so deep into Dunning-Kruger. Do you even understand how modern computers work? The one I'm literally typing on now has 12 parallel threads. Every Khronos technology is parallel out of the box. I can't even. Like you could literally type anything else and be closer to correct.

On the last part: I would easily say that until the point California released their data many people thought that the AP technology would transfer much better than it did to SDC. Considering that was the consensus at the time. As is, they are one of the worst in the world, despite having one of the better (if not the best) automated cruise control system.

LIDAR works even worse in rain/snow than cameras. Again, a fundamental lack of knowledge about the most basic aspects of the technologies you purportedly are an expert on. This is literally wiki level knowledge, which you have not mastered in either finance or technology.

I was talking about low light.

Given that we have previously discussed this very fact (Lidar fails with particles in the air), at this point your fragile ego is just making stuff up to feel about the fact that I'm wiping the floor with you.
Good lord you're dense. What you said has no relationship to what I said. Parallel processing of streams vs combining of deep learned algorithms are two completely different things. How is this not obvious to you? Dunning-Kruger indeed.

Probably because you don't have the faintest clue what you're saying and think that bulk posting makes you somewhat correct or less wrong or who knows what you think (Glad you jumped ship on that risk/asset class discussion, that must've been pretty brutal). There would be a single algo that is run in parallel. This is obvious to anyone. Who once mentioned multiple algorithms?!?

Also, what is a deep learning data set? That makes literally no sense. Sensors produce data. This is aggregated into a data set. Whether it is used in machine learning or deep learning makes no difference. So the appending "deep learning" to the front of data set is just more buzzword bingo to make you appear intelligent, but in reality reveals your deeper lack of understanding to people who have even passing familiarity with the topic.

why are you even talking about their data? it's totally worthless, they are years behind competition (look at breaking). by the time they got enough m3 on the road and collected whatever ****ty data they want mobileye(ibm)/google/ford/zegermans will have them beaten anyway.

This is just comical dude. The conversation has gone right over your head, I see. You don't understand the difference between data and algorithms, which is hilarious. There's a reason you can't parallelize the training of something that gets deep learned. I'm sure some day as the field gets more advanced, someone will teach this to you, and the light bulb will go on.
It's amazing how much goes over people's head. What I've said above is straightforward and correct. But people can't think, even those who work in the field. When I patiently explained to everyone a year ago why Tesla was far behind on autonomous driving, I was correct while someone who actually works in the field - heltok - was dead wrong. You made the same spiel then about me "lacking an understanding". It seems you don't learn from mistakes.
Yeah, it's obviously worthless if you know anything about how this will work (I laid out a little above), and Tesla are obviously far behind. Not obvious to idiots, apparently. Hence why Musk can spin his pure bull**** over and over and have people believe it. That's literally proven above with his prior statements and how all his fans swallowed it.

Training Deep Learning models wouldn't even be realistically possible if it couldn't be parallelized. The whole idea of training it on Nvidia hardware is that it benefits so much from parallelization.

TS (a supposed trader) burying his incredibly poor stock predictions with technical stuff he most certainly is not an expert at.

Color me surprised.

Sent from my SM-N910V using Tapatalk

Again, this isn't hard. Or maybe it for people; even the experts who work in the field like Heltok were dead wrong about where Tesla was, when it was absolutely obvious they were far behind. Even when told, with reasoning, they couldn't see it.

You can parallelize inputs into deep learning
You can parallelize trivial discrete stuff like object recognition.
You can't parallelize what the algorithms come up with when it gets complex. Which is why Musk's 50K cars returning data are worthless, since that data isn't video streams.

This is just another reason (among many), why Musk's supposed advantage in data capture is non-existent. Autonomous driving models don't benefit meaningfully from parallelized low bandwidth processed data. Not in a way that takes you from 3 to 4+. They may be of limited benefit in mapping local highway obstacles and anomalies, but that's not getting you anywhere on the autonomous driving scale.
Yet I get the right answers, better than the "experts" here, who can't even think in simple novel ways to realize what the data says.

If you listened to anyone but me (including Elon Musk) for most of this thread on autonomous driving, you got entirely the wrong picture.

And covering what exactly?? My advice/predictions in this thread has been excellent, apart from the latest runup from the high 200s, which I didn't see coming.