rrizzi7210
Member
- First Name
- Robert
- Joined
- Nov 18, 2020
- Threads
- 0
- Messages
- 6
- Reaction score
- 22
- Location
- West Palm Beach, FL, USA
- Vehicles
- Cybertruck
- Occupation
- Tech Business Owner
I found a few takeaways during the earnings call today.
I believe that when the prototype was unveiled in November of 2019, Tesla set several automakers into planning mode for what could be done to capture as much as possible of the e-pickup market demand that nobody knew existed before that event, including Elon.
Elon and all the c-level execs knew more or less how long it was going to take to get GigaTexas online and production-ready. They probably figured, "What do we have to lose by building a game-changer prototype of what Elon wanted in a Tesla Pickup Truck." The problem with that is they showed their cards too early, and most of the big automakers "knew" they had to build prototypes and get them to market as fast as possible. Even if they cannot scale production to match demand, it won't matter because they also probably knew how difficult it would be (even for Tesla) to build GigaTexas and scale the manufacturing process of a new vehicle. The big automakers hedged their bets, and it looks like it might pay off now.
Then, COVID-19 hits, and everything changes. The Supply-Chain nightmare isn't even close to over yet. No one has a solution in place to solve the pent-up demand problem without increasing capacity only to catch up to demand AND then.....deal with the problem of "excess" capacity AFTER supply and demand achieve equilibrium. It is too risky to scale only to have to later abandon these massive fixed-asset investments. I won't even get into how the current political climate plays into this either.
Elon had two choices around this point. "Do we [Tesla] build the original prototype, or modify the prototype to be at least equal to or better than what the other companies have proposed in their prototypes?" It appears he chose the latter. Elon revealed during the earnings call today the question of how to produce "an affordable" CT. I think he has recently learned that the proposed cost of the latest CT prototype is far beyond what he finds "acceptable" for the typical pickup truck owner. When the single-motor CT was introduced at $39,900, the satisfaction from potential buyers was reflected in the number of orders that were placed, even though most of those orders were for dual and tri-motor variants. The price was still more than acceptable for these two more expensive variants.
Sometime later, Tesla removes the configurator from their website, and all pricing information is gone.
The problem is when the design team was tasked by Elon (of course) to improve the prototype based on what was learned when the big automakers revealed their prototypes; I think they did so for too long of a period without getting his feedback. Let's face it, Elon is "busy," and while he knew what was being done to the newer prototypes, he probably did not know what the consequences to the selling price were going to be. This is why he reduced expectations today by not saying anything new and not delivering details on a CT product roadmap. It must have happened in the last few days. He is basically "pissed off" at the design team and has to get involved in fixing the problem. He goes where he is needed most, which is another one of those situations.
Lately, he has been too busy with SpaceX. The same thing happened with the new engine for the latest rocket they are working on. He let almost all or all of the c-level execs in charge of the new Raptor engine production line go and personally went in on the weekends to "fix" the problems.
This is why when I read between the lines in the earnings call, he is saying that Tesla will "try" to produce the CT in Q1 of 2023, but I believe it will not happen. He does not even have enough data to know precisely when it will be, but if he were conservative (which he is not), he would have said Q1 of 2024. If they had done that, Elon knows he would have lost most of the momentum earned in late 2019 and early 2020, plus he would have told the competition that they won big on their bets.
Where does this leave us now? Well, like many others, I, too, am just about ready to order a Model Y Performance instead and will wait for probably a few more years to even see the first CT on the road. The problem is, what will I think when I start to see all the competitors' EV trucks on the road? With low confidence from Tesla regarding the CT, I might test drive one. Even though I don't feel that way today, I could change my mind 12 months from now.
The other issue is if Tesla "reveals" the new prototype and the price, it will be like taking candy from a baby for the competitors. They will do what they did before and repeat. They have the advantage in some respects, but in many other ways, they do not. They probably won't scale reliable battery production in less than three years. They don't have an L3 charging network as Tesla does, and they will learn from Tesla eventually.
In closing.....I was very disappointed to hear two things during the earnings call. The timing of choosing to talk about the humanoid robots was terrible. People, in general, are afraid of losing their jobs to machines, let alone humanoid robots. I think Elon even realized while speaking that he almost slipped and was going to talk about how much more efficient and more productive manufacturing could be if he just had them now instead of people. How often have we heard Tesla (or Elon) say that people in Fremont have choices about where to work. It's hard to get qualified people to build Tesla vehicles in Fremont. Maybe the humanoid robots are the answer to that problem. They don't quit, take breaks, get sick, complain or fight with co-workers.
I was disappointed to hear that Elon thinks that FSD will be ready by the end of the year. I want to go back and listen to precisely what he said again, but I think he said: "would be done by the end of the year." There is no way in the world that FSD Beta will be ready by December of 2022. Yes, the features in Autopilot might be 99.99999% or at least better than any human, but FSD Beta will not be there for at least another three to five years.
I am a software developer (and hardware designer) and have worked with embedded systems, neural network processing modeling, and many other areas within computer science for over 30 years. It has been by "day job" for over 30 years. While I am incredibly impressed with the progress so far, the next 1% of improvement will take another 90% of the time and effort spent thus far. I believe the problem is complex enough to be classified under what I would call the "law of diminishing returns." The progress vs. time graph is asymptotic, meaning the complexity and cost of increasing the performance of a system increases geometrically to the feature set or the inverse of the error rate.
I believe Elon already knows this, and he was hoping that Dojo would allow them to break through this barrier when they begin the development several years ago, through machine learning to improve the labeling system by order of magnitude or two. He could be correct, but during the earnings call, he even said he wasn't sure they would be successful with Dojo in the future. Dojo would only be used if the FSD development team felt the gains achieved with Dojo compared to GPU clusters increased. It might be too early to determine that, and I think the gains would be higher with Dojo if neural network processing can be shown to improve and solve the labeling problem as it stands right now. I think it will take several years to get there, regardless of how many existing Tesla owners are helping to train the models. FSD is the type of problem such that improvements occur when orders of magnitude in capacity or performance are achieved, meaning to double the benefits, you have to increase the performance or capacity by a factor of 10 and to gain four times the improvement; you have to increase performance or capacity by a factor of 100 and so on.
The only other way Tesla can deliver a working, useful replacement for human drivers would be for the DOT to work with Tesla to change the design of roads and highways, or for Tesla engineers to have a breakthrough in terms of how neural network processing is done and I am not talking about speed or depth here.
I'm not going to cancel my order for my CT, and I am around 113K in the queue, and I expect to see my CT (quad motor) in Q2 or Q3 of 2024. In the meantime, I will be a part of those buyers who continue to push Tesla vehicles sales upward and onward because I need to replace one of my existing vehicles soon.
I believe that when the prototype was unveiled in November of 2019, Tesla set several automakers into planning mode for what could be done to capture as much as possible of the e-pickup market demand that nobody knew existed before that event, including Elon.
Elon and all the c-level execs knew more or less how long it was going to take to get GigaTexas online and production-ready. They probably figured, "What do we have to lose by building a game-changer prototype of what Elon wanted in a Tesla Pickup Truck." The problem with that is they showed their cards too early, and most of the big automakers "knew" they had to build prototypes and get them to market as fast as possible. Even if they cannot scale production to match demand, it won't matter because they also probably knew how difficult it would be (even for Tesla) to build GigaTexas and scale the manufacturing process of a new vehicle. The big automakers hedged their bets, and it looks like it might pay off now.
Then, COVID-19 hits, and everything changes. The Supply-Chain nightmare isn't even close to over yet. No one has a solution in place to solve the pent-up demand problem without increasing capacity only to catch up to demand AND then.....deal with the problem of "excess" capacity AFTER supply and demand achieve equilibrium. It is too risky to scale only to have to later abandon these massive fixed-asset investments. I won't even get into how the current political climate plays into this either.
Elon had two choices around this point. "Do we [Tesla] build the original prototype, or modify the prototype to be at least equal to or better than what the other companies have proposed in their prototypes?" It appears he chose the latter. Elon revealed during the earnings call today the question of how to produce "an affordable" CT. I think he has recently learned that the proposed cost of the latest CT prototype is far beyond what he finds "acceptable" for the typical pickup truck owner. When the single-motor CT was introduced at $39,900, the satisfaction from potential buyers was reflected in the number of orders that were placed, even though most of those orders were for dual and tri-motor variants. The price was still more than acceptable for these two more expensive variants.
Sometime later, Tesla removes the configurator from their website, and all pricing information is gone.
The problem is when the design team was tasked by Elon (of course) to improve the prototype based on what was learned when the big automakers revealed their prototypes; I think they did so for too long of a period without getting his feedback. Let's face it, Elon is "busy," and while he knew what was being done to the newer prototypes, he probably did not know what the consequences to the selling price were going to be. This is why he reduced expectations today by not saying anything new and not delivering details on a CT product roadmap. It must have happened in the last few days. He is basically "pissed off" at the design team and has to get involved in fixing the problem. He goes where he is needed most, which is another one of those situations.
Lately, he has been too busy with SpaceX. The same thing happened with the new engine for the latest rocket they are working on. He let almost all or all of the c-level execs in charge of the new Raptor engine production line go and personally went in on the weekends to "fix" the problems.
This is why when I read between the lines in the earnings call, he is saying that Tesla will "try" to produce the CT in Q1 of 2023, but I believe it will not happen. He does not even have enough data to know precisely when it will be, but if he were conservative (which he is not), he would have said Q1 of 2024. If they had done that, Elon knows he would have lost most of the momentum earned in late 2019 and early 2020, plus he would have told the competition that they won big on their bets.
Where does this leave us now? Well, like many others, I, too, am just about ready to order a Model Y Performance instead and will wait for probably a few more years to even see the first CT on the road. The problem is, what will I think when I start to see all the competitors' EV trucks on the road? With low confidence from Tesla regarding the CT, I might test drive one. Even though I don't feel that way today, I could change my mind 12 months from now.
The other issue is if Tesla "reveals" the new prototype and the price, it will be like taking candy from a baby for the competitors. They will do what they did before and repeat. They have the advantage in some respects, but in many other ways, they do not. They probably won't scale reliable battery production in less than three years. They don't have an L3 charging network as Tesla does, and they will learn from Tesla eventually.
In closing.....I was very disappointed to hear two things during the earnings call. The timing of choosing to talk about the humanoid robots was terrible. People, in general, are afraid of losing their jobs to machines, let alone humanoid robots. I think Elon even realized while speaking that he almost slipped and was going to talk about how much more efficient and more productive manufacturing could be if he just had them now instead of people. How often have we heard Tesla (or Elon) say that people in Fremont have choices about where to work. It's hard to get qualified people to build Tesla vehicles in Fremont. Maybe the humanoid robots are the answer to that problem. They don't quit, take breaks, get sick, complain or fight with co-workers.
I was disappointed to hear that Elon thinks that FSD will be ready by the end of the year. I want to go back and listen to precisely what he said again, but I think he said: "would be done by the end of the year." There is no way in the world that FSD Beta will be ready by December of 2022. Yes, the features in Autopilot might be 99.99999% or at least better than any human, but FSD Beta will not be there for at least another three to five years.
I am a software developer (and hardware designer) and have worked with embedded systems, neural network processing modeling, and many other areas within computer science for over 30 years. It has been by "day job" for over 30 years. While I am incredibly impressed with the progress so far, the next 1% of improvement will take another 90% of the time and effort spent thus far. I believe the problem is complex enough to be classified under what I would call the "law of diminishing returns." The progress vs. time graph is asymptotic, meaning the complexity and cost of increasing the performance of a system increases geometrically to the feature set or the inverse of the error rate.
I believe Elon already knows this, and he was hoping that Dojo would allow them to break through this barrier when they begin the development several years ago, through machine learning to improve the labeling system by order of magnitude or two. He could be correct, but during the earnings call, he even said he wasn't sure they would be successful with Dojo in the future. Dojo would only be used if the FSD development team felt the gains achieved with Dojo compared to GPU clusters increased. It might be too early to determine that, and I think the gains would be higher with Dojo if neural network processing can be shown to improve and solve the labeling problem as it stands right now. I think it will take several years to get there, regardless of how many existing Tesla owners are helping to train the models. FSD is the type of problem such that improvements occur when orders of magnitude in capacity or performance are achieved, meaning to double the benefits, you have to increase the performance or capacity by a factor of 10 and to gain four times the improvement; you have to increase performance or capacity by a factor of 100 and so on.
The only other way Tesla can deliver a working, useful replacement for human drivers would be for the DOT to work with Tesla to change the design of roads and highways, or for Tesla engineers to have a breakthrough in terms of how neural network processing is done and I am not talking about speed or depth here.
I'm not going to cancel my order for my CT, and I am around 113K in the queue, and I expect to see my CT (quad motor) in Q2 or Q3 of 2024. In the meantime, I will be a part of those buyers who continue to push Tesla vehicles sales upward and onward because I need to replace one of my existing vehicles soon.
Sponsored