Featured Tesla vs Bolt production ramp up

It seems to me that the original article is leaving out some details.

I don't remember Chevy taking hundreds of thousands of deposits on the Bolt. So as far as I can tell they mostly ramped up their production to meet their perceived demand. (They also have to work with dealerships, who may push back, while Tesla doesn't have this issue.) I haven't seen any indications that Chevy could've, say, built twice the number of Bolts and sold them all. And they had to build demand in the face of the Tesla hype: overcommitting at the front end would have led to lost money and front-page photos of acres of unsold Bolt's, titled "Chevy just can't compete!".

Maybe Chevy really was hard-limited in their production. But they certainly didn't drum up hype and then fail to deliver in a timely manner.

Looking more closely at the word "deliver". The 3's that Tesla is delivering will probably not live up to the promise of autonomous driving. It's taken how long for their 2nd generation Autopilot to catch back up to the 1st generation? And they probably don't have enough computing power -- and arguably don't have enough sensors -- for full-automatous driving. Let's see what happens two years from now, if Tesla 3 owners still don't have fully autonomous driving, whether Tesla actually ramped up production of the promised vehicle or if they ramped up something else.

Of course current Autopilot is superior to what current Bolts have, but no one is buying a current Bolt based on believing it'll get an over-the-air update that adds fully-autonomous driving. (I didn't carefully parse the announcement's words and perhaps Elon said something less than "fully-autonomous", so I could be mistaken. But I do remember him hyping that it has all of the sensors and computing power necessary for future self-driving.)

Also, Chevy's got fleet fully-autonomous uses in sight and that could also explain how they're allocating resources.

Last, Elon has overpromised a lot in the past. His reputation is that he eventually delivers, but he's legendary for overpromising. Even simple things like "silky-smooth Autopilot release", but usually deadlines. So it's understandable that his ramp-up -- which still hasn't actually met his production number goals -- would draw harsher criticism than might be justified purely based on numbers. Still, Tesla's been running full-out to bring production up, while Chevy didn't have that fire under them so could basically afford a measured ramp-up which biases the comparison.

 

Yeah, I've found some sites, especially Evanex, to be highly pro-tesla/anti-all-others. FWIW Just like FoxNews, past the headlines, I've stopped reading them and don't give 'em clicks.

Hybrid and Electric Car News and Reviews - Green Car Reports and http://www.hybridcars.com/ are more "fair and balanced" while Inside EVs | Electric Vehicle News, Reviews, and Reports usually is most of the time but they also use evannex and electrek hit pieces/clickbait way too much IMO.

As always, YMMV

 

GM sure they have done that a lot. On the bolt no, but as a company much worse than tesla. Current management is much better than the team that bankrupted the company and got bailed out by taxpayers though. ouch.

It is very doubtfull autonomous driving will be legal in the US in 2 years. I don't know anyone that put down a deposit that thought that was included. If someone pays for it, and the software works, and its legal, tesla will upgrade the hardware for free if its not good enough. That's not much of a risk to tesla, but silly to pay the $3K extra for. No one has autonomous driving. Uber just killed a pedestrian testing their autonomous driving because the safety monitor failed (that is the human in the car that was supposed to stop it from killing).

Tesla's hardware is much much better than autopilot 1. Mobileye's software was better until recently. Software is not easy, but it is getting there.

What are you a tesla short? Yes all the sensors and computing should be there if they can write the software and the laws change. Software and laws are hard. Its a $3K option that very few are going to take a flyer on.

The model 3 is a much nicer car than the bolt in just about every way. They now have production where they could reasonably make a profit selling the short range version for $36,200 after destination before tt&l and tax credits. They won't because they have enough people waiting that they can require $14,000 in options and almost everyone will also opt for the $5000 autopilot. After the federal tax credit is up there may be some base models available. If they could crank out 10K a month they would, but they still are working on manufacturing.

They also don't have automous driving. I don't think you understand what chevy has.

 

From the articles that I've read, Google and GM have orders of magnitude more hours between test drivers overriding their systems than trailing players like Uber. And as far as I know, both GM and Google are testing fully autonomous cars in urban environments, not just an Autopilot equivalent. As I understand it, GM has already made 200+ autonomous Bolts (including LIDAR, the sensor that only Tesla rejects).

Currently, GM is selling SuperCruise (or whatever they call it) that you can get on Cadillacs, but that's not what I'm talking about.

Speaking of LIDAR, that's what I was hinting at. All the players except Tesla believe that LIDAR is required for full autonomy, and of course Tesla doesn't have them and they'd be difficult to retrofit. From what I've read, Nvidia doesn't believe that their current hardware in the Tesla is sufficient for fully autonomous driving. And I've read comments by other players -- which could be sour grapes or trash talking -- that Tesla doesn't have enough redundancy in its sensors to be fully trustworthy.

Again, it may have just been me, but at the 3 unveiling it sure sounded like he was saying that the hardware in the 3 was all you'd need for future software upgrades to full autonomy. Maybe he just meant Autopilot 3.0 or something and I got it wrong. But he's certainly over-promised quite a few times ("silky smooth Autopilot update" which wasn't silky or smooth).

I followed Tesla pretty seriously when I thought we might be able to squeeze the money to buy one. Not realistic, though, and as you say the Bolt doesn't fully compete with the Tesla -- at least the Tesla S.

Tesla placed dead last in self-driving race by Navigant, GM and Waymo top list has a mixed message. In terms of what they can do right now, Tesla is dead last and GM & Waymo at the top. The article says that GM and Waymo only achieve this in areas they've heavily mapped (I assume GPS, LIDAR, the works), but I'm not so sure that their vehicles aren't also learning to drive. They also have cameras and Google's pioneered convolutional neural networks for vision, so the only advantage Tesla has is crowd-sourcing.

Report: GM and Waymo lead driverless car race; Tesla lags far behind | Ars Technica : "Today's Tesla vehicles also lack lidar sensors, which most other companies in the industry consider essential for full autonomy. And "even Nvidia has expressed doubt that the computing hardware it sells to Tesla is capable of supporting full automation reliably," Navigant argues."

"As a result, Cruise [Chevy's autonomous subsidiary] has emerged as one of Waymo's top rivals in developing fully driverless vehicle technology. While Waymo has shifted much of its testing efforts to Phoenix (where state regulatory oversight is lax), Cruise has concentrated on testing in San Francisco. Vogt has arguedthat San Francisco is a better testing environment because cars encounter unusual and difficult situations—like construction zones, pedestrians, and emergency vehicles—much more frequently, allowing Cruise to improve its software more quickly."

 

I believe it says that Autopilot 2.0 hasn't caught up with Autopilot 1.0, which was true a year or more after they dumped Mobileye. I was an avid reader of Tesla forums when I thought we might be able to buy one and moving from Mobileye to their own Nvidia-based system really set them back. Again, I keep going back to Elon's "silky smooth autopilot" announcement that was not close to true -- according to the Tesla forums users.

It's possible that they've matched and exceeded Autopilot 1.0 performance now, I figured there's no way we could afford a Tesla and have basically ignored them since. Until the charging station situation dramatically matures, a hybrid makes the most sense for us.

The only advantage the Tesla has ever had is that they have publicly deployed their system, which means they have a very broad (alpha/beta) testing program -- though it's shallow because of bandwidth constraints to the mothership. Everyone else can do the same things with vision, plus they have LIDAR. Heck, I've played with convolutional neural networks (CNNs) for vision and deployed one for text mining, and these are what Tesla (and everyone else's except maybe Mobileye) is using now.

Tesla's downside has been underpowered hardware/sensors compared to the other players. Google, et al's test cars can have a lot more hardware in them and lots of specialized storage and communications. But of course there are fewer of them.

So it makes sense to me that Google and GM have very nice systems at this point. And it feels like Tesla, with it's production and 3 distractions, may have dropped the ball. The first year or more of Autopilot 2.0 certainly reinforces that impression. Just my opinion, though.

I hear you on paywalls, and I do wonder if Navigent is knowledgeable and neutral. I see a lot of similar "experts" in the AI and computer space.

 

Uber stole waymo (owned by google) lidar ideas from a former employee, a court gave google part ownership of uber because of the theft. I do not believe the waymo or gm system could have prevented the accident, and there's the problem, when a human kills a pedestrian its the human that is in trouble, when a self driving car does it, the whole industry is slowed.

GM, waymo, apple, and tesla are testing their self driving systems, none of them are fully autonomous - level 5. GM bought a lidar start up and is building test vehicles. After the Uber incident apple slowed their plans. Tesla has over 100,000 cars supplying data which it is learning from. This is an infant industry. No one is buying a waymo, apple, or gm car with lidar other than the companies and their testers. No one should be buying a model 3 thinking it will go full autonomous, but somehow these stories are out there that because they aren't .... its a failure.

All car companies have over promised. GM, Honda, and Toyota all claimed hundreds would be driving fuel cell cars by now. Telsa does believe the model 3 has all the sensors and computing power it needs to go level 5 in the model 3. Others disagree. It certainly has better hardware than human drivers ;-) The problem is software and laws. The software until recently for the new hardware wasn't as good as older software built for worse hardware. Until the software works and is tested we won't know, and that is at least a year away. My guess is it will be at least 5 years before fully autonomous driving is legal outside these small test sites.

Yeah I find this kind of report hugely flawed and I used to buy them when I was a product manager.

Let me ask you, why do you think google bought waze when it had great mapped data? waze leveraged its user base to find out in real time changes in road conditions. The head unit in my car is awful on getting me real time traffic, often 10 to 15 minutes late. I believe this is key to developing safer cars and better software. When there are hundreds of thousands of tesla's collecting data about accidents and driver reactions, and these lessons can be uploaded to cars over the air, its a powerful thing. My guess is other companies will copy tesla's model. Will a car need lidar? I doubt it, but if it will, a lidar system that is cheap enough still hasn't been invented. If it ever is, and many speculate it will be, the odds are they will want tesla as a customer ;-)

Don't believe the hype. You don't need autonomous driving to sell millions of bevs. Someone will develop it, but my guess is they will need to leverage information from millions of vehicles to make it work well.

 

There is no doubt that Google and GM have some very nice products with Waymo and Cruise. They are going about things totally differently than Tesla (including doing some extreme 3d mapping of areas where it can be used). Time will tell who wins. No one knows for sure yet how Google plans on monetizing Waymo. Will it be standalone ride sharing or will it be a product they sell to other car manufacturers? Cruise will only be on GM's most expensive vehicles for quite a while. And currently it is around a $20k option. It will be quite a bit more expensive than any Model 3 for some time yet. GM using Lidar will ensure it will be expensive for many years to come. Elon leaning on the neural net has a much bigger upside IMO.

 

An even more powerful thing (as in magnitudes greater) would be for all vehicles to share. But will that ever happen?

 

If it does, which company do you think would be the first to implement it? Who is on the path to a (somewhat) mainstream vehicle?

 

Truly a gazillion dollar question!

It will have to be one system, right? No vhs vs betamax, no Mac vs PC, no Apple vs Android, etc. Or can/will a hodge-podge of different tech work as one?

 

It's not really symmetrical, though. Tesla has a hundred thousand beta testers, while Google/GM don't, but Google/GM have cameras as far as I can tell -- they just think LIDAR is also necessary.

If I were at Google/GM there's no way I'd not be seeing what I can do with just cameras as well as with cameras augmented by LIDAR, RADAR, etc. If I can do well enough with just vision, why not deploy it?

Of course, I'd look at redundancy, though. Can a single camera failure cripple me? How well do cameras work in the dark? (LIDAR would not be affected there.) What if a headlight blows out? What if it's a snowy day with lots of salt spray? (LIDAR and SONAR would be as vulnerable as a camera, RADAR would be less so.)

 

As much as I like Apple, I think their plans for building a car never really got off the ground. Building the software for someone else's car maybe, but not their own car. (They have a lot of software design experience -- even in-car software experience via Apple Play. They also have a lot of augmented-reality experience and vision experience. So making the software or even a hardware/software combo that someone else embeds in their car makes sense.) I think actually building an Apple car would be a big mistake on their part.

That's their claimed advantage, but I wonder if there's much substance to it. For example, the recent crash: the guy had evidently repeatedly reported problems at that spot, but Tesla already had his -- and perhaps thousands of other drivers' -- disengagement data available to them. If they were on top of their data, they'd have known that this section of the road is tricky, and could have even had geo-tagged photos of the spot from multiple cameras and angles to figure out why.

They could have certainly flagged the spot somehow, and perhaps updated their neural network. Of course, they may have added it to their training set and either: a) it still didn't learn, or b) it learned but that improved version won't hit the air for months. In either case, I'm not so sure that there's enough upstream bandwidth that Tesla's literally learning all that much from a hundred thousand beta testers.

They changed from old-school methods to deep neural networks a year and a half ago. Presumably with some amount of testing before that. (They also changed from mainly-camera to mainly-radar, evidently. I hadn't known that.) A hundred thousand beta testers for a year and they still hadn't matched inferior hardware and old-school techniques. I don't think progress is inevitable.

Oooo, a sort spot for me. I don't use Waze precisely because I've seen such terrible performance from it when riding in Uber/Lyft. It is the worst router, by far trailing Apple and Google. Why? It's too quick to change. On multiple occasions I've sat in an Uber/Lyft that got routed in weird routes, eventually coming back to the original route, except it took longer to get to that point than if we'd simply stuck with it. So perhaps Google's better routing software and Waze's user base have synergized, but they certainly weren't a great product in my experience. (Don't get me wrong, I'm not a Google or Android fan at all.)

Again, I think that Tesla's Autopilot falls into an awkward gap. (I guess it's Level 4?) The bottom line is a car that assists me but I'm driving (Level 3 and below?) is helpful and cars have been doing that since the invention of hydraulic brakes and power steering -- certainly things like ABS and pre-collision systems. But having a car that proactively "drives" -- except you have to maintain strict attention because it may at any moment do something unexpected -- is just a fun toy for us geeks, not a finished product. I read enough in the Tesla forums that I would not trust my mother, father, or wife in a Tesla with Autopilot.

As you say, full level 5 is a long ways off, and might take a LOT of beta testers to perfect. But I think there's a gap between "assist" and "autonomous" that can't really be bridged incrementally.

 

@wfolta I'm sorry, but you've been listening to the Tesla FUDsters too long. Much of your assumptions and arguments have already been proven wrong. I don't have the time or desire to sit here and refute every single one of them.

 

Like I said, you really seem to not be thinking this through. Apple, Waymo (google), Tesla, Uber, and all the car companies are trying to get a suite of hardware and software to work. No car manufacturer is as close as tesla. Tesla wants to sell the tech if they get it right as does waymo and apple. Uber is the only one that I know of that wants to keep it for itself, and they have been shady.

Tesla had thousands of cars pass this point sucessfully. Satelite data shows that the barrier was damaged recently before the crash. The driver also was not paying attention. We won't know everything until the ntsb is through with their investigation. There is too much data to actually process.

Telsa collects shadow mode data. That is it collects data from sensors and how drivers react while not self driving, and when they take over from autopilot. This is very valuable data to improve future systems.

Stop with the FUD. There are not a hundred thousand beta testers for the software. The software works, and it works well. Product impovements improve as new situations are encountered.

They changed to better hardware. They did not throw away old methods. Stop with the beta tester FUD. Their previous vendor mobileye had some great software. Software is not easy. Tesla didn't pay to port the software to the nvidia hardware but decided to develop it themselves. That takes time. Intel bought mobileye not for their hardware but for the software.

Two pieces in the gm suite are better than teslas. 1) driver facing camera with software to detect if the driver is paying attention, 2) better maps. Tesla has added a driver facing camera in the model 3 but doesn't have the software yet developed. My guess is someone like google will make the best maps and license it. Tesla does a large number of things better than gm or waymo. Their system can do better in fog, heavy rain, or snow. Their system can handle unpredictable events. This makes the cars safer but not self driving fully autonomous. No car has that capability.

+1

 

Sorry, conversation ends here. I have developed a lot of software and I am not engaging in FUD and will not respond to someone who resorts to personal attacks.

 

I am sorry that you consider that a personal attack, instead of an instruction to think about the information you are repeating. If you develop difficult software that you should definitely understand that people buying a software product in an early generation are not beta testers. Many will buy for current functionality and possible future features, that may never exist.

Despite the name - enhanced autopilot - self driving software that can prevent all accidents is not the product. It requires driver attention. The system does auto steer, change lanes, maintain distance quite well, but it is not self driving. IMHO there probably are plenty of sensors in hw 2.5 (which really just adds a little more computing power and a driver facing camera to 2.0). The software is extremely difficult though, and computing hardware has been built so it can be upgraded just in case the software needed to do self driving requires faster hardware. Hardware is approximately 40x faster than it was in autopilot 1. There are now 8 better outfacing cameras but they are not all implemented in software, along with better radar, and ultrasonic sensors (for close range). You can prepay $3000 for autopilot software that does not exist yet, but I don't think anyone is buying the model 3 based on being able to prepay for future software. If you prepay for the software and it requires that extra hardware tesla will install it for no extra cost, but IMHO the chances that they develop it and it is legal before you are ready to buy a new car is fairly low.

 

Seems like the OP's "ramp up" went from the meaningfulness of #'s of cars 1st manufactured versus timeline, to quality of driver assist. REAL quick And here i thought it was MY task to derail

As for the best thing going on ANY EV, it doesn't mean a thing if it aint got no quick charge infrastructure. Fit & finish is important ... a lack of nickel & dime aggravation issues is too. Highest safety rating, driver assist - all important. But if anything proves QC infrastructure is the make or break. Look at the deposit holders & look at the most advanced QC infrastructure - bar none. A lack of long range QC infrastructure makes all other EV models insignificant - according to the hand wavers.
Then there's the charge speed thing. The SAE QC is seldom putting out 50kW's despite a higher theoretical number. The Model 3 has been able to suck down at 116kW's for a 30 minute 200 mile quick pick me up.
Supercharger speed: 116kW | Tesla Motors Club

The other stuff, although important, is superfluous in comparison. When is GM going to start breaking ground on 1,000's of QC stalls might be the better question.
.