I'm not a fan is Tesla. But they have a few hundred thousand cars now on the road? With all sorts of idiots using autopilot or over-relying on it at least.

It's amazing the death-relates numbers are as low as they are...but of course that isn't the total number of accidents.

Show me some data where it says stepping into a Tesla is 5x more likely to result in death. How about the WayMo or Uber data? I'm probably more likely to get into an accident while being ferried around in an Uber than a driver is in their Tesla.

The problem is that you can't really compare an autonomous system with a human unless the human can't intervene. As soon as a Tesla driver hits the brakes or jerks the wheel, Autopilot is considered disabled and therefore "not engaged" during a crash. There are hundreds of videos of this happening just before an accident. To fairly judge a system, it would have to have no human input. Tesla also doesn't test on public roads so we have no true idea what their disengagement rate would be.

View some of the posted videos, you'd be amazed how far they have to go.

There's also the notion that some bring up about hundreds of thousands of vehicles driving millions of miles that gives Tesla a huge data advantage. Those who've studied the data transfer of a given vehicle indicate this isn't the case at all, it's marketeering.

How many vehicles does Tesla have to sell to make a profit?

Nobody knows. There might not be a number. That's why Tesla is doomed.

TennAg said:

---

The problem is that you can't really compare an autonomous system with a human unless the human can't intervene. As soon as a Tesla driver hits the brakes or jerks the wheel, Autopilot is considered disabled and therefore "not engaged" during a crash. There are hundreds of videos of this happening just before an accident. To fairly judge a system, it would have to have no human input. Tesla also doesn't test on public roads so we have no true idea what their disengagement rate would be.

View some of the posted videos, you'd be amazed how far they have to go.

There's also the notion that some bring up about hundreds of thousands of vehicles driving millions of miles that gives Tesla a huge data advantage. Those who've studied the data transfer of a given vehicle indicate this isn't the case at all, it's marketeering.

---

Their 100 watt power consumption constraint seems arbitrary. They've hamstrung themselves in processing power trying to make all of the self driving use less than 100 watts. It's the tail wagging the dog because they've pre-sold self driving, the software is now going to be limited to the hardware decisions made years ago.
Those cars will never be fully self driving in the same way they were marketed.

bmks270 said:

---

TennAg said:

---

The problem is that you can't really compare an autonomous system with a human unless the human can't intervene. As soon as a Tesla driver hits the brakes or jerks the wheel, Autopilot is considered disabled and therefore "not engaged" during a crash. There are hundreds of videos of this happening just before an accident. To fairly judge a system, it would have to have no human input. Tesla also doesn't test on public roads so we have no true idea what their disengagement rate would be.

View some of the posted videos, you'd be amazed how far they have to go.

There's also the notion that some bring up about hundreds of thousands of vehicles driving millions of miles that gives Tesla a huge data advantage. Those who've studied the data transfer of a given vehicle indicate this isn't the case at all, it's marketeering.

---

Those cars will never be fully self driving in the same way they were marketed.

---

why not? what's to say it has the capabilities but the limitations are 'self imposed' because this is relatively new technology and as this evolves they'll be able to deliver more functionality or relax certain restrictions.

It not how many vehicles, it's how many environmental credits do they need sell to other manufacturers.

TennAg said:

---

The problem is that you can't really compare an autonomous system with a human unless the human can't intervene. As soon as a Tesla driver hits the brakes or jerks the wheel, Autopilot is considered disabled and therefore "not engaged" during a crash. There are hundreds of videos of this happening just before an accident. To fairly judge a system, it would have to have no human input. Tesla also doesn't test on public roads so we have no true idea what their disengagement rate would be.

View some of the posted videos, you'd be amazed how far they have to go.

There's also the notion that some bring up about hundreds of thousands of vehicles driving millions of miles that gives Tesla a huge data advantage. Those who've studied the data transfer of a given vehicle indicate this isn't the case at all, it's marketeering.

---

Shouldn't there be data available regarding overall crash rate as well as fatality rate of Tesla's compared to other automakers, regardless of whether self driving function was engaged?

Quote:

---

Shouldn't there be data available regarding overall crash rate as well as fatality rate of Tesla's compared to other automakers, regardless of whether self driving function was engaged?

---

Funny you say you that, the data does exist. Tesla chooses to compare their late model luxury fleet with the existing automobile fleet at large. i.e. motorcycle crashes, old cars without abs, insufficient airbags, etc. When compared to late model luxury brands, it's a whole different story.

Quote:

---

why not? what's to say it has the capabilities but the limitations are 'self imposed' because this is relatively new technology and as this evolves they'll be able to deliver more functionality or relax certain restrictions.

---

Haha there's nothing "self-imposed" about it. These cars are hard braking at shadows, overhead highway signs, overpasses, etc. They also sometimes can't handle peaks after a rise in the road, lane splits, stopped fire trucks, and many situations when a vehicle exits a lane directly in front of a Tesla. This is all the supposedly easy stuff on highways, btw.

Flashdiaz said:

---

bmks270 said:

---

TennAg said:

---

The problem is that you can't really compare an autonomous system with a human unless the human can't intervene. As soon as a Tesla driver hits the brakes or jerks the wheel, Autopilot is considered disabled and therefore "not engaged" during a crash. There are hundreds of videos of this happening just before an accident. To fairly judge a system, it would have to have no human input. Tesla also doesn't test on public roads so we have no true idea what their disengagement rate would be.

View some of the posted videos, you'd be amazed how far they have to go.

There's also the notion that some bring up about hundreds of thousands of vehicles driving millions of miles that gives Tesla a huge data advantage. Those who've studied the data transfer of a given vehicle indicate this isn't the case at all, it's marketeering.

---

Those cars will never be fully self driving in the same way they were marketed.

---

why not? what's to say it has the capabilities but the limitations are 'self imposed' because this is relatively new technology and as this evolves they'll be able to deliver more functionality or relax certain restrictions.

---

I thought my post explained it.

The tail is wagging the dog.

Their software and hardware is constrained by a vehicle platform they produced in 2016. They collected customers money selling a full self driving option.

Three years have passed and the self driving is not close to being acceptable for release, but all new self driving developement is bound by the design choices they made in 2016 so that it is compatible with those vehicles.

This means no new sensors can be implemented and that they are operating and developing their software with massive computing hardware limitations. This is why they designed their own processor.

As it is, many experts do not believe acceptable levels of reliability and safety can be reached with Tesla's camera only approach.

Other groups working on self driving vehicles have no limitations. They can implement new sensors and processing power as required to improve safety and reliability, Tesla cannot.

When Tesla reaches full self driving it will be on 2016 hardware and sensor architecture. When others reach full self driving it will be on state of the art hardware and sensor package.

In 2016 Elon essentially commanded his company to "Make it work with what we have!!!"

While everyone else asked a question and not a command: "What works given what we have?"

The "What works?" Groups are testing and iterating design.

Tesla's "make it work" directive has so far failed but is still trying to make it work on 2016 model years. So yes, the limitations are self imposed to work on 2016 Tesla vehicles. Those limitations have proven to be a few too many for safe and reliable self driving.

Actually, their compute rack can be replaced with newer hardware, but you are right: they are limited to the sensors they have now.

People are dumping on them because of technical issues. That's silly: they may not have the most advanced self-driving solution, but their battery tech for cooling and software was first to market and as I said, in 250,000 cars?

I think Tesla will struggle because they will have Saturated a market for $60,000+ buyers and have done so with tax payers footing part of the bill. Churning over Teslas at that volume yearly will come to an end...unless I'm severely underestimating the penchant to churn vehicles from the same buyers.

evan_aggie said:

---

Actually, their compute rack can be replaced with newer hardware, but you are right: they are limited to the sensors they have now.

People are dumping on them because of technical issues. That's silly: they may not have the most advanced self-driving solution, but their battery tech for cooling and software was first to market and as I said, in 250,000 cars?

I think Tesla will struggle because they will have Saturated a market for $60,000+ buyers and have done so with tax payers footing part of the bill. Churning over Teslas at that volume yearly will come to an end...unless I'm severely underestimating the penchant to churn vehicles from the same buyers.

---

50% of Tesla buyers live in California. They've absolutely saturate they've
customer base (California tech bros).

And if their new hardware needs to draw more power then they may need updated harnesses and connectors, they might be limited by ability to cool higher output electronics in the same space, and it will also reduce the range of the vehicle. This is why they had to design their own low wattage chip, the computing hardware constraints are real.

good post and a fair assessment.

That being said, one of Tesla's weaknesses and possibly strengths is they're not afraid to change direction... actually, they change what they say too much which causes concerns over stability with investors.

So far the sensor limitation is a directive from the top to make the current software work. Once competitors start to catch up, they can introduce new, better functionality and more sensors in new models similar to how Apple introduces more hardware\functionality on newer devices while still maintaining the old (i'm sure this is going to agitate the anti-apple crowd).

This is also what is different with Tesla vehicles vs what's out today. Most other 'smart' vehicles can't be upgraded easily. With Tesla, they're constantly adding updates over wifi.

Flashdiaz said:

---

...Once competitors start to catch up, they can introduce new, better functionality and more sensors...

---

1. You're not paying attention to where Tesla stands. They are the ones catching up. You're being fooled because they recklessly bring untested tech to market.
2. This is nothing like apple or an iphone. This is peoples lives we're talking about. Real companies test this stuff for years before letting the public see it. Tesla isn't testing on public roads at all.

I like some of your points, but you are trying to argue some holistic view of "best autonomous" driving regardless of it exists in a lab or is productized.

Don't overlook that some of these systems that are better and in R&D have no immediate path to commercialization in 200,000 vehicles. I rode in Aptiv's "autonomous" BMW with a safety driver in the front and an engineer watching their data stream and Lidar feedback....all while going 10mph in traffic.

Tesla has the best autonomous package available today to a consumer, and you can't argue that. It may not be the case in 5 years, but tell me another car that can be bought which drives itself on a highway and takes exits based on nav. I'll check back in 2021....maybe.

Mu point is that there's way better tech than Tesla's, those companies choose to test it rather than make it available.

Admittedly, I don't think I'm doing a good job understanding what you're saying.

Because they mega cheaped out on the sensor package. Maybe cameras will one day be able to provide sufficient data to run a self driving car, but all these cars will be in a scrap yard long before then

Today, as well as for the reasonable lifespan of all the cars currently on the road that Tesla marketed as self driving capable, it's LIDAR or lol gtfo. Tesla can't include LIDAR, though, because it's way too expensive for consumer use, so instead they're selling self driving snake oil. That isn't my opinion, but the opinion of just about every expert in the self driving space that doesn't work at Tesla (and looking at their self driving division turnover, most of the ones at Tesla don't seem to think it's possible either).

I see the point you're trying to make, but you're missing the main takeaway, which is that Tesla has no business doing what they're doing. The reason that they have the best "consumer available" self driving tech is because they're irresponsibly selling half finished tech while the other competitors in the space are actually trying to develop a safe, finished product before taking it to market.

And again, it's important to again note that it's literally impossible for them to attain their goal with their current sensor package. None of the big researchers in the self driving space think that cameras are currently up to this task. But, they charged ahead anyway because they thrive on marketing and excitement rather than substance.

This. Tesla is promoting a product as roadable when it should still be in beta phase.

actually looks like some researchers at Cornell have some ideas for cameras tad get it close to LIDAR. I'm still skeptical about its performance during edge cases, though.

Flashdiaz said:

---

good post and a fair assessment.

That being said, one of Tesla's weaknesses and possibly strengths is they're not afraid to change direction... actually, they change what they say too much which causes concerns over stability with investors.

So far the sensor limitation is a directive from the top to make the current software work. Once competitors start to catch up, they can introduce new, better functionality and more sensors in new models similar to how Apple introduces more hardware\functionality on newer devices while still maintaining the old (i'm sure this is going to agitate the anti-apple crowd).

This is also what is different with Tesla vehicles vs what's out today. Most other 'smart' vehicles can't be upgraded easily. With Tesla, they're constantly adding updates over wifi.

---

Part of the problem is they sold fully autonomous driving to tens, if not hundreds, of thousands of customers in 2016 and on. They'll have to update those cars' hardware or get sued for fraud. You can't do that over Wi-Fi. They're looking at potentially huge money to update those cars if they need to go that route. New sensors would require new harnesses and body work to fit them, along with the cost of additional computer/sensor hardware.

As mentioned earlier, they face constraints in harnesses, cooling, etc with hardware. Move outside those constraints and you might as well just give out refunds and then some. Even within them, they'd be burning through a lot of cash to make any hardware changes.

evan_aggie said:

---

I like some of your points, but you are trying to argue some holistic view of "best autonomous" driving regardless of it exists in a lab or is productized.

Don't overlook that some of these systems that are better and in R&D have no immediate path to commercialization in 200,000 vehicles. I rode in Aptiv's "autonomous" BMW with a safety driver in the front and an engineer watching their data stream and Lidar feedback....all while going 10mph in traffic.

Tesla has the best autonomous package available today to a consumer, and you can't argue that. It may not be the case in 5 years, but tell me another car that can be bought which drives itself on a highway and takes exits based on nav. I'll check back in 2021....maybe.

---

If I offer s**t for sale, it may be the best and only s**t available, but at the end of the day it's still s**t.

What Tesla has done could be easily commercialized and rolled out in upcoming model years into hundreds of thousands of vehicles by other manufacturers. They're not doing that because their systems are not 100% capable and those other manufactures aren't taking any chances. They tend to do things right or not at all, and fit good reason. a) You're taking about people's lives and b) going the wrong route could prove incredibly costly.


A perfect example of the difference between all other vehicle manufacturers and Tesla is in their touch screens. Musk wanted a huge screen and couldn't understand why other cars didn't have one. After all, laptops had big screens, and they were really durable. Then Tesla went shopping for giant car screens and found out no one made them. Other manufacturers use smaller screens because they are automotive grade and built to withstand some insane conditions that laptop and computer screens simply can't. Going bigger with automotive grade is just not economically feasible.

Musk said **** it and went with the best thing they could get. It was like 17" and had really good durability specs. It was a screen designed for industrial, but not automotive, conditions. It has some impressive operating specs until you look at exactly what automotive grade is. It's just barely sorry of the lowest automotive grade. Tesla tested, thought it worked fine, and ran with it.

Turns out they didn't work really well and ran into all kinds of thermal issues. Tesla was replacing them left and right. That's where their cabin overheat protection came from, which mitigated some of the worst problems. They still had problems with banding though, and they started telling customers late last year to pound sand on warrantied screen replacements, calling the remaining issues,"cosmetic." Now they're trying to develop an in-house screen. They're supposedly rolling those out, but they may just be remanufactured screens.

Long story short, other automakers tested having and used proven products with incredibly high standards for their tech. Tesla went the cheapest, coolest route. Only of them has spent a ton of money mitigating and warranting an issue they have yet to solve, all the while pissing of their customer base.

Tesla may be selling Autopilot right now, but they may also end up eating a ton of money later on with hardware or sensor upgrades the same way that ate money on screen replacements. It may be, "productized," but it could also end up as an albatross, whereas those systems still in the lab actually pan out profitably.

That's a lot of words to say, "Their consumer product is garbage in my opinion and they are cutting corners."

I'm not rushing out to buy one of their cars for a number of reasons, nor buying their stock. You don't have to convince me. But you are now waging (and others on this thread) that their self-driving tech is crap and risky and they are going to be blown away by a *real* product is...some sort of "I know best" holy war argument.

I'm sure someone's cool R&D project at Nissan, Uber, Mercedes or even Audi (one could argue that Audi is closest to commercial) has better tech and a safer solution. Okay, they are more ethical. Okay, theirs is better. Where can i buy it? When can I buy it?

Tesla was first to market with some sort of mode that allows you to hit a button and drive on a highway. I've, reluctantly, been a passenger in this semi autonomous mode and it scares the **** out of me. Folks seem perturbed to admit that or acknowledge it. The first jet airliner came out and 3 crashed killing hundreds within a few years. The engineers went back and redesigned square windows to become round. There are going to be a lot of risks taken before something is deemed worthy for consumer readiness. What point that is? I don't know, I'm not going to argue.

Tesla won the first battle in consumer tech. But pretend otherwise if it helps you sleep.

They didn't win a battle, they're just forcing customers to buy the self driving package because it's probably one of the few things on that car where they actually make money (I bet they have a nice margin on all the sensors). the other companies in the space haven't released a product because they don't have to (and there's no real first mover advantage here because they aren't creating any sort of ecosystem lock in like Apple or Google with their smartphones).

This quote is from a thorough review of the lane changing feature that you're has so impressed you:


https://bgr.com/2019/05/22/tesla-autopilot-consumer-reports-navigate-lane-changes/

"In practice, we found that Navigate on Autopilot lagged far behind a human driver's skill set: The feature cut off cars without leaving enough space and even passed other cars in ways that violate state laws, according to several law enforcement representatives CR interviewed for this report. As a result, the driver often had to prevent the system from making poor decisions.

"The system's role should be to help the driver, but the way this technology is deployed, it's the other way around," says Jake Fisher, Consumer Reports' senior director of auto testing. "It's incredibly nearsighted. It doesn't appear to react to brake lights or turn signals, it can't anticipate what other drivers will do, and as a result, you constantly have to be one step ahead of it." "



You seem to think Tesla won some sort of battle by rushing what is clearly a half baked product out the door. I don't understand that. We're in mile one of a marathon, Tesla is behind the other competitors, and you're assigning them some sort of victory just because they're showboating louder than the companies that are beating them.

evan_aggie said:

---

That's a lot of words to say, "Their consumer product is garbage in my opinion and they are cutting corners."

I'm not rushing out to buy one of their cars for a number of reasons, nor buying their stock. You don't have to convince me. But you are now waging (and others on this thread) that their self-driving tech is crap and risky and they are going to be blown away by a *real* product is...some sort of "I know best" holy war argument.

I'm sure someone's cool R&D project at Nissan, Uber, Mercedes or even Audi (one could argue that Audi is closest to commercial) has better tech and a safer solution. Okay, they are more ethical. Okay, theirs is better. Where can i buy it? When can I buy it?

Tesla was first to market with some sort of mode that allows you to hit a button and drive on a highway. I've, reluctantly, been a passenger in this semi autonomous mode and it scares the **** out of me. Folks seem perturbed to admit that or acknowledge it. The first jet airliner came out and 3 crashed killing hundreds within a few years. The engineers went back and redesigned square windows to become round. There are going to be a lot of risks taken before something is deemed worthy for consumer readiness. What point that is? I don't know, I'm not going to argue.

Tesla won the first battle in consumer tech. But pretend otherwise if it helps you sleep.

---

GM and Nissan have pretty highway assist tech being rolled out.

GM is actually using geofencing where the driving assist can only be enabled on specified highways via GPS, and when you exit the highway it gets disabled.

Honda has auto distance keeping that will follow cars in traffic, you can get it in a Civic.

Mazda has some simple lane keeping assist and emergency brake assist.

Tesla is ahead, but their implementation is less safe and backwards. They have humans monitoring the machine and intervening when the machine is unsafe. Everyone else has the human more engaged and the machine intervene when humans are unsafe.

An analyst on Squawk Box yesterday put a target price of $4000 in 5 years. That would put TSLA at a market cap in the $600B range. Hmmm....

What's that analyst smoking? Because I want some too.

IrishTxAggie said:

---

What's that analyst smoking? Because I want some too.

---

He is just saying that to act as click bate for the headline to get more eyeballs and advertising dollars for his employer. The number is not based in any rational analysis.

It was a buy side analyst talking her book.

95k deliveries last quarter, new record. They also announced new orders exceeded deliveries and their backlog has actually increased.

Fizban said:

---

95k deliveries last quarter, new record. They also announced new orders exceeded deliveries and their backlog has actually increased.

---

The problem I see with the deliveries number is that they claimed q1 was super low due to having roughly 10k cars in transit. Seems like q2 was an average quarter inflated by late q1 deliveries.

In q4 they delivered 85k. In q1 they delivered 63k. In q2 they delivered 95k. Say q2 was average. Q1 was still 12k vehicles low. The question becomes, "Exactly how many q1 in transit vehicles are counted as q2 deliveries and what does q2 really look like?" If you subtract the 10k cars reported in transit in q1, q2 was just an average quarter. For Tesla that's not good.

Fizban said:

---

95k deliveries last quarter, new record. They also announced new orders exceeded deliveries and their backlog has actually increased.

---

Any quarterly profit guesses?

they also switched who got cars first so most of the cars produced in April were sent overseas (since that takes longer) and US customers started getting theirs in May.

ABATTBQ11 said:

---

Fizban said:

---

95k deliveries last quarter, new record. They also announced new orders exceeded deliveries and their backlog has actually increased.

---

The problem I see with the deliveries number is that they claimed q1 was super low due to having roughly 10k cars in transit. Seems like q2 was an average quarter inflated by late q1 deliveries.

In q4 they delivered 85k. In q1 they delivered 63k. In q2 they delivered 95k. Say q2 was average. Q1 was still 12k vehicles low. The question becomes, "Exactly how many q1 in transit vehicles are counted as q2 deliveries and what does q2 really look like?" If you subtract the 10k cars reported in transit in q1, q2 was just an average quarter. For Tesla that's not good.

---

Given Tesla's rate of growth there really isn't such a thing as an "average" quarter. Q4 2018 was somewhat boosted by the expiration of the tax credit. Q1 2019 was somewhat depressed by the additional shipping time and logistical complications associated with introducing the Model 3 to Europe and Asia. Q2 2019 picked up some of the Europe/Asia deliveries that missed Q1, but it is still an all-time record. This combined with orders exceeding deliveries and continued improvement in the Model 3 production rate (highest ever in Q2) make it hard to see Q2 as average.

TennAg said:

---

Fizban said:

---

95k deliveries last quarter, new record. They also announced new orders exceeded deliveries and their backlog has actually increased.

---

Any quarterly profit guesses?

---

Nah, certainly it will be a far stronger quarter than Q1 but I would basically just be throwing darts. Neither a small loss nor a small profit would surprise me.