!'ll take 1.5 cars at 400 miles range.
If it's a truck, go ahead and give me the 600 miles.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
If it's a truck, go ahead and give me the 600 miles.
a winter to increase sample size to test cases on new system?
Carl Raymond
Active Member
So let’s assume that the payout is 3x, thrice the usual payout for that tragedy. Tesla, as the insurer, must ensure the tech makes this event at least one third as frequent to remain in front. Since they must first show data to demonstrate that FSD is 10x safer (accidents one tenth the frequency) for regulatory approval, this is already a given.
Notably, of ten pedestrians previously deceased, nine get to live. It’s about more than cash, especially to those nine pedestrians.
Artful Dodger
"Neko no me"
Tesla revealed some interesting specs on Semi reveal night. From F=MA, it takes 1,200 hp (at the wheels) to accelerate 0-60 mph in 20 sec at 80,000 lbs (see link below for calc):
Using the calculator in the link above, that weight and 0-60 mph time yields this hp figure:
- 0-60 in 20 sec @ 80,000 lbs: 1,200 hp
Wikipedia lists 283 hp in its specs for the RWD Model 3. Well, perhaps at 50% SOC that might be true, but dyno-testing at 80% SOC has demonstrated 360 hp at the wheels in a new 2018 LR RWD.
So we can easily rate the Semi at 300 hp per motor to obtain 1,200 hp for the Semi. Let's plug in some more numbers into the calculator:
- 0-60 in 5 sec @ 1,200 hp: 19,400 lbs
Regen ability is primarily limited by the battery. But at 2 C burst charge rate, even the 600 KWhr version of the Semi should be able to accept 1,200 KW of regen, which is equivalent to 1,600 hp and near the limits of the motors and electronics.
TL;dr The Semi is likely set to produce around 1,200 hp, and weighs about 19,400 lbs empty.
Cheers!
Last edited:
No company has reached a trillion market cap without revenue to back it up. They'll still need a few years after getting safe and polished FSD working. However, I believe they'll achieve about the same revenues if they can safely drive empty cars where people need them to be.
Some thoughts on what happened lately.
1. Is it possible that Saudi's sold their 4.9% stake? They had a protective hedge on their investment for below $347 I believe. Couldn't they use their hedge, sell all their stock during Q1 in the upper $200's, then buy them back for an easy ~30% profit?
2. The Q1 model 3 margin maintaining at 20% is astonishing given the multiple price cuts, the introduction of the 35k and 37k SR+ lowering ASP, and the extra expense incurred for shipping to Europe and China. That's a much, much better improvement in manufacturing cost than I expected. I look forward to seeing Q2's amazing margins.
3. Here's an interesting thought. What if Tesla's deal with FCA isn't all cash? It is arguable that Tesla doesn't need cash, but manufacturing capacity. It's Europe, there's no way GF4 can be built anywhere near as fast as China, but we do know Tesla needs a Europe GF4 ASAP. Is it possible that Tesla negotiates some sort of factory takeover deal to turn it into GF4? That would be worth much more than a straight cash deal for Tesla
1. Is it possible that Saudi's sold their 4.9% stake? They had a protective hedge on their investment for below $347 I believe. Couldn't they use their hedge, sell all their stock during Q1 in the upper $200's, then buy them back for an easy ~30% profit?
2. The Q1 model 3 margin maintaining at 20% is astonishing given the multiple price cuts, the introduction of the 35k and 37k SR+ lowering ASP, and the extra expense incurred for shipping to Europe and China. That's a much, much better improvement in manufacturing cost than I expected. I look forward to seeing Q2's amazing margins.
3. Here's an interesting thought. What if Tesla's deal with FCA isn't all cash? It is arguable that Tesla doesn't need cash, but manufacturing capacity. It's Europe, there's no way GF4 can be built anywhere near as fast as China, but we do know Tesla needs a Europe GF4 ASAP. Is it possible that Tesla negotiates some sort of factory takeover deal to turn it into GF4? That would be worth much more than a straight cash deal for Tesla
Full and polished FSD with no competition for five years best anyone can see equals trillion dollar market cap before you can blink. You don’t even sell cars anymore. Just put them on the road and collect revenue. One million cars at 100000 miles per year is conservative 100 billion revenue. With no one else in the game.
Won’t get into this with you as I am not predicting ‘full and polished’ FSD within three years. But if you make that assumption and can’t see the trillion in a heartbeat, that is your fail not mine.
I thought that a driverless car (Uber) already killed a pedestrian (?) 
Self-Driving Uber Car Kills Pedestrian in Arizona, Where Robots Roam
So we know the public's reaction to that already ( = none whatsoever).
Self-Driving Uber Car Kills Pedestrian in Arizona, Where Robots Roam
So we know the public's reaction to that already ( = none whatsoever).
Carl Raymond
Active Member
I thought that a driverless car (Uber) already killed a pedestrian (?)
Self-Driving Uber Car Kills Pedestrian in Arizona, Where Robots Roam
So we know the public's reaction to that already ( = none whatsoever).
There was a driver at the wheel not paying attention. Different situation.
FSD will recognise pedestrians and would have avoided this collision if physics permit. If that is not the case, FSD is not ready for driverless operation.
I wondered about this too, but maybe in a different direction.
I find the quotes from the May 3 FCA Earnings Call to be interesting:
Edited Transcript of FCA.MI earnings conference call or presentation 3-May-19 12:00pm GMT
"The Jeep Renegade and Compass plug-in hybrid vehicles were revealed at the Geneva Auto Show in March, and these will start production early 2020 and represent the initial ramp-up of high-voltage vehicles for our European fleet. And they will be followed by the all-new Fiat 500 BEV and 10 additional launches of heavy -- heavily electrified vehicles over the following 2 years.
...
But when I think then through 2021, for example, obviously, in 2020, we will have the 2 plug-in hybrids and the battery electric vehicle. In 2021, I think that conventional tech will give us about 40% of our compliance. Electrification by that time, we will then have 5 vehicles -- we'll have more than that. We'll have 3 from 2020, plus another 6 coming on stream in Europe. 45% of our compliance will probably come from electrification, about 15% from purchase credits."
The numbers are a little contradictory but it sounds like FCA thinks they are going to be selling six new PHEV/BEV models in 2021 (not including the Renegade, Compass and Fiat 500). And they think they are going to be selling a LOT of them (I would guess 80,000 or so to get the level of compliance they are predicting).
Has anyone heard about what these models.are? How do they think they are going to ramp up that fast? Could Tesla be helping them in some way, maybe by selling them batteries and/or motors?
Edit: Looks like they're Maseratis
Last edited:
myt-e-s-l-a
Member
Buffett knocks Elon Musk's plan for Tesla to sell insurance: 'It's not an easy business'
“The success of the auto companies getting into the insurance business is probably as likely as the success of the insurance companies getting into the auto business,” says Buffett.
“The success of the auto companies getting into the insurance business is probably as likely as the success of the insurance companies getting into the auto business,” says Buffett.
Makes for a great quote, but it’s pretty ridiculous. It would be an order of magnitude harder for an insurance business to become an auto manufacturer.
myt-e-s-l-a
Member
All I know about it is what I learned from the Investor Autonomy day presentation and some inexpert googling, which mostly boils down to smaller batch size meaning lower latency of processing (ideally batch size of 1, which is what Tesla's NN chip is built to handle), if your architecture can natively handle it - sometimes a smaller batch size is actually run as a larger one with wasted performance as there's a minimum "width" for processing on GPUs. Batch size means that for example if you need batch size of 128 (not uncommon in NN-on-GPU approaches for efficiency, it seems) you must collect 128 images to be processed and then process them all at once. Clearly this is less than ideal as that means either breaking up fewer images into overlapping segments and hoping to get a useful result, processing a bunch of null images (throwing away performance just to fill the minimum batch size requirement if there is one), or waiting until you have enough unique frames.
Even with the advantage of a camera-agnostic NN that Tesla seems to have (i.e., the same NN can handle images from all cameras, rather than needing duplicate NNs trained to each specific camera), that would mean 128 images / 8 cameras = 16 frames per camera needed to be captured before you begin processing at full efficiency. Even if you operated your cameras at 120fps you'd still end up processing with up to (for first frames in the group) over 0.1 second lag between input and output which would be less than ideal, especially if you're capturing at a lower frame rate such as 30 (~0.5 second peak lag) or 60 fps (~0.25s peak lag). Alternatively, you only process a few frames at a time, and throw out the vast majority of your computing power. This will improve latency but sacrifice performance by filling in the rest of the batch size with wasted empty frames. Your effective frame rate doesn't improve, but the peak time waiting for a processed frame is decreased, so as long as you can throw more power at it wastefully, you can keep up.
I'm not sure where the limits actually are for NVidia GPU or Tensorflow applications, almost everything I find with google (as a non-expert) is talking about increasing the batch size as much as possible and the problems that means (needing more working RAM), I guess because when you're doing training you just want to throw as much at it as possible since latency is a non-issue.
With Tesla's design built for batch size of 1, however, all those issues are sidestepped and maximum performance is always had. With batch size of 1, the minimum and maximum latency of processing images is identical, and always as fast as it can be.
I highly doubt the goal for Tesla insurance is to make money. I think they do it for owners to get reasonable rate reflecting the cars safety features.
As if Buffet knows nothing about the man at all. What's the chance of success starting a rocket launvl company and an electric car company.?
This is simply ridiculous from him. He is smart enough to stay away from tech industry because apparently he has no idea how disruptive technology works. He is also smart enough to realize that with tech entering the entire economy he can afford to stay away anymore. And his first choice was IBM!
omgwtfbyobbq
Active Member
All things being equal, maybe... With that said, Tesla may have some breathing room if they can open up some more collision repair shops and exploit inaccurate insurance premiums other companies charge customers.
Last but not least, everyone always says that whatever they do is hard. Even if it isn't actually that hard, all (effectively) encouraging competition does is make it harder while discouraging competition makes it easier.
Last edited:
For Europe, given that the pools are public, we can see that FCA isn't buying credits from anyone else. The pools are the only way to do it.
jhm
Well-Known Member
I do think there is evidence that they are taking this sort of approach. I've seen it with AP1. Whenever AP1 gets confused about where the road is it disengages and cedes to control of the driver. Moreover, as time progressed, the frequency of these disengagements decreased. Now this is in the context of having a driver to hand over control to, but it does demonstrate the principle. Clearly Tesla is thinking about all sorts of redundancy to assure high reliability. So I've got to believe they have some sort of fall back responses when the NN is highly uncertain about how to proceed. With FSD they can't cede control to a driver, but they can slow to a stop.
Also I think the example of the bike on the back of a car is an example of this sort of strategy. They recognized that the NN was being confused about bicycles appearing to cross the road when in fact they were strapped onto a car. What alerted them to this problem? I should hope that it was discovered by diagnostic routines. Hopefully, the system was correctly alert to the potential for a bike or pedestrian to cross the lane of travel. That would be a true positive. But then the NN notices that drivers routinely drive toward "bikes crossing the road" in some cases. These would be the false positives where the bikes are actually strapped on to the car ahead. They'd also have visual confusion about an object that looks like it could be a car or a bike. If they are were doing this right, the cars would be taking a defensive response to these false positives. Then there would be diagnostic routines to review these defensive responses. Once the problem has thus been identified, the NN can be trained to accurately observe that a bike is attached to a vehicle and thus resolve the false positive. So in this instance, I think Tesla was rising on the Slope of Enlightenment.
Now you've pointed out that a bike strapped on to a car can in fact fall off. So yes, there does still exist some really risk in following such a vehicle. So this goes to setting a prudent following distances in such a case. If the case is well identified, i.e., the vehicle can perceive when a bike is attached to a vehicle. Then they are also in a position to tell when it in not attached to a vehicle. So bikes falling off a vehicles are probably already identified events which the NN can learn from. Ideally the NN is recalibrating to allow a suitable distance or to consider changing lanes. But all this is part of the general problems of how much following distance to allow and when to change lanes. The more gritty work was just to train the net to see when bikes are attached and when they are not.
So I am much more optimistic about where Tesla is on FSD. There have already been years where AP was just not making much apparent progress. Remember when a car was supposed to solo across the US? That did not happen. Why? I think that was when research was in the Valley of Despair. They had had enough experience with AP to know that it was really hard to avoid having to cede control to the driver or to have the driver actively disengage. I think that's how they climbed back down Mt. Stupidity. So I think AP was Mt. Stupidity, but that was okay because there was still a driver at the wheel to handle the stupidity. I know I did my part. There was one power pole in particular that AP just wanted to drive right into. So I had to take control every time. I sure hope that somewhere in Tesla this data was being analyzed to understand just what weird thing was going on. Even so, I was doing my part to tell the car, "Don't be stupid. That's not the way to go." I'm also encouraged by shadow mode. Every time a driver takes an evasive action, this should trigger data. The NN should be trained to predict when a driver is about to take an evasive action. This is one way to detect false negatives, the situation where the driver is correctly responding to a hazard, but the NN has heretofore failed to detect that hazard. By simultaneously comparing driver behavior to the action that the NN would have taken in that situation, the system can detect a kind of cognitive dissidence. Specifically there is a substantial disconnect between how the driver and NN interpret a situation. So this should signal for some sort of diagnostic work. Of course, sometimes drivers do stupid things where the NN would not have done the same stupid thing. But where there is substantial disagreement between driver and NN, there is the potential for the NN to learn something important. When the driver takes evasive action, either the NN has a false negative, the driver has a false positive or perhaps both. Eventually, the NN will be able to critique the driver and point out where the driver had false positives or false negatives. (Imaging the NN reporting this analysis of driver behavior back to one's insurance carrier!) At any rate, I think there is huge potential to identify edge cases, especially false negatives by having NN analyze human driver behavior. I think a lot of this problem detection can be done analytically. I suspect that hand work is needed more to resolve the false negatives or false positives where feature extraction is underdeveloped. So long as the system can respond with caution where problems have been detected or where drivers and the NN are likely to diverge, you go along way toward eliminating the possibilities for false negatives. This still can leave you with an abundance of false positives to resolve, but you have more time to work that out.
You've got to climb down Mt. Stupidity (detecting false negatives) very fast, but can take more time climbing the Slope of Enlightenment (resolving false positives). I think they've got multiple lines of attack on both fronts. It a big tool box. What is harder to assess is just how effectively they are using these tools. No doubt they are still on their own learning curve around how to best to cultivate the NN. But I do suspect they are well down the path where FSD needs very little supervision from drivers. To wit, if the NN can predict when drivers are about to take evasive action and act faster than human response, there should be no need for the driver to take that action, as the NN can beat them too it. You actually get to a moral problem where in shadow mode FSD passively allows drivers to make mistakes. No doubt some drivers will make lethal mistakes that NN would have calculate to be such even before the driver commits them.
There is an analogy here to NN that have been trained to play chess by estimating the probability of winning with each play made. Such a system would be able to watch a human player and know when they have elected to make a move that sends their probably to winning to zero. The NN could tell the player, "Don't be stupid. That play will cost you the game." It's all well and good for a NN to watch a human player lose a game of chess, and it may even learn something in the process. But what about when a NN knows that a human driver will imperil lives? So as NN descend Mt. Stupid, it gains the ability to critique its own driving and that of human drivers. At some point we look back, and NN will show us how much time even the best human drivers stumbling about on Mt. Stupidity. We'll even have little black boxes that bear witness to the final mistakes that a human driver makes. For me, I think this lends some insight into why Musk is so serious about how people will eventually want to ban driving without computer assistance. Shadow mode probably gives you a very dim view of human drivers even as it illuminates how FSD can avoid making the same mistakes over and over again.
Take a look at Ihor Dusaniwsky's graph with black lines and black-ink data added by me to reflect his text estimates of short-selling (as well as clarify which days of the week are associated with the rapidly-climbing short interest).
Four million shares were shorted in a week with really good Tesla news arriving on Thursday and Friday (successful cap raise, so much interest in cap raise that it is likely to grow in size, Elon raising his purchase amount to $25 million, etc.) at a time when TSLA appears to have bottomed out slightly below the October dip levels and is ready to start recovering.
Now, look at the quantity of short interest increase from Monday morning to Wednesday close. The stock price remained nearly level, but close to 3 million shares were shorted in these three days. Think about a million shares shorted in one day. That's like 20 instances of selling 50,000 shares in a minute's time. It's a crazy powerful amount of selling and yet the stock price hardly budged. Why? Perhaps there is some type of white knight neutralizing the efforts of the shorts until the cap raise is completed. If the white knight is an entity such as Goldman and motivated toward maximizing their commissions, then we could expect them to drop us like a hot penny the moment the cap raise is completed. On the other hand, if the white knight is a big stakeholder, someone like Ballie Gifford protecting their current investment, then we could expect some residual help as well. Can you explain the fortitude of TSLA on Monday-Wednesday despite the extreme short-selling taking place if you don't include a white knight in the scenario? Remember that Tesla didn't announce the cap raise until Thursday. I get the rise on Thursday and Friday. It's the first part of the week that puzzles me in terms of TSLA's resilience. Right now I suspect a white knight but would love to hear your view.
stef
Member
I suspect the old man is starting to feel some heat. This is not the candy shop, this is the core of his story.
Edit: as much as he doesn’t understand tech, he knows very well that insurance is based on data, and that Tesla has more data than anyone regarding their cars. His comment doesn’t make rational sense.
Similar threads
- Locked
- Locked
