I spent a good couple years tinkering with NNs back about 10 years ago. The above might send me down the rabbithole of how they are pulling off Int8 math these days. Back in the 'old days' when you had to multiply two numbers uphill both ways in the snow they tended to be at least 32 bit floats which were always sufficient (and thus probably too much) but 8 bits doesn't give you much to work with. Premature optimization is a major trap for engineering minds. Elon talks about this in general for all engineering in his recent Starbase tour video. If they are adding mitigation efforts just now then clearly the optimization hasn't come for free. Would be delightful if a chunk of the presentation is on this since it relates a lot to the custom hardware on both Dojo/car and represents a 'real-world' optimization that doesn't necessarily exist in academic work.
-
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
I spent a good couple years tinkering with NNs back about 10 years ago. The above might send me down the rabbithole of how they are pulling off Int8 math these days. Back in the 'old days' when you had to multiply two numbers uphill both ways in the snow they tended to be at least 32 bit floats which were always sufficient (and thus probably too much) but 8 bits doesn't give you much to work with. Premature optimization is a major trap for engineering minds. Elon talks about this in general for all engineering in his recent Starbase tour video. If they are adding mitigation efforts just now then clearly the optimization hasn't come for free. Would be delightful if a chunk of the presentation is on this since it relates a lot to the custom hardware on both Dojo/car and represents a 'real-world' optimization that doesn't necessarily exist in academic work.
BobbyKings
S85D > M3P > S Plaid
Clearly I misunderstood Audubon.
The mad respect for makers tweet, I assume was more of a general tweet. He’s made similar comments in the past, I believe. Same sort of reference to making things at a production scale vs prototypes. But I’m not certain
Only in America is setting the thermostat to 78 F is reducing “demand“ of electricity. Way too many people have it set to low 70s I guess.
The vast majority of his tweets are found under the 'Tweets & replies' tab, if this was the point of confusion/object of inquiry.
Artful Dodger
"Neko no me"
The 'Never-Never' is NOT a production car: Rimac plans to build only a total of 150 Nevera (the homologation minimum). Their planned production rate is only 1 car per WEEK!. So by end of 2024 for the last of these cars to be delivered? At that rate, they could source their batteries cells from Radio Shock...
This is not a serious production effort. As Elon often reminds us, "Prototype are easy, Production is hard". When bty cell supply allows for producton, Tesla will produce about 150 Roadsters per WEEK! And that's Tesla's tiny volume car.
Never-never is a PR grab, and does NOT move the needle on carbon emissions in any useful way.
Wasted effort, IMO.
Last edited:
CLK350
Member
For those interested in the details and how to interpret these notes further, see our own TMC's Terminator857 post
"
> Clear-to-go boost through turns on minor-to-major roads (plan to expland to all roads in v9.3)
Accelerates faster during turns.
> Improved peek behavior where we are smarter about when to go around the lead vehicle by reasoning about the causes for lead vehicle being slow.
When is peak? When it is needed most? Right before a decision is needed to be made?
> v1 of the multi-modal predictions for where other vehicles (are) expected to drive. Partial implementation
Multi-model means multiple modes. FSD will have predictions for different possibilities what a car will do. For example: 70% chance will continue, 20% chance will turn, 10% will stop. Multi-model may mean that the algorithm used depends upon the situation. For example if near an intersection.
> New lanes network with 50k more clips (almost double) from the new auto-labeling pipeline.
More lanes mapped? Is this a right turn lane, a bike lane, etc...
> New VRU velocity model with 12% improvement to velocity and better VRU clear-to-go performance.
VRU = vulnerable road user. Pedestrian, bike, escooter, etc... Thanks to Daniel in SD for that.
> Model trained with "Quantization-aware-training" (QAR), an improved technique to mitigate int8 quantization.
int8 quantization is used during machine learning training. Initially floating point was used, but to conserve memory a switch was made to int8. This introduces rounding errors. QAR is a technique used to take into account the rounding error, and provide better overall performance.
...
... "
Peek !== Peak
Im guessing they added more code to the car will steer to the left or right a bit to try and peek ahead to see whats going on further down the road.
I think network refers to new NN. Likely meaning they have a new network to the replace an older one with double the training data. Most significant aspect of this sentence to me is the reference to auto-labeling pipeline which I believe means DOJO
CLK350
Member
Interesting situation - Question is whether their stated Mission Statement
"To accelerate the advent of
SUSTAINABLE HEALTHY LIVING
systems around the world."
is Marketing fluff or has real intentions with real power behind it. Need someone with chops to figure out the ins and outs of this company. In the best of cases, great alternative buy at these prices, as it might be a good acquisition candidate for Tesla especially considering SpaceX will need to figure out Mars settlements, even though this is pretty far into the future. The other way to look into it is as communities that are resilient in the face of upcoming weather disruptions.
I disagree - as a halo vehicle, this is the type of vehicle required to convince people that EVs are the future. You could easily make the same argument ("not moving the needle on carbon emissions") about the Roadster / Model S / X as low volume, expensive luxury vehicles, but I doubt you'd declare those vehicles a waste of time.
I think of the Nevera as advertising for electric vehicles, including Tesla. People will see that and say "Whoa! I want one!". And then they will go buy a Tesla Model 3 because that's what they can afford.
This is the same as the Model S Plaid - look at how many people Brooks @ Dragtimes has convinced to buy a Tesla on the spot by showing off the performance of the Model S Plaid, P100D, etc? These same people then go off and either buy another Model S and then follow those up with a 3, Y, etc...
Any good EV on the market (even a halo vehicle like the Nevera) is good for the entire EV market.
Pezpunk
Active Member
I agree with all of this except the last two sentences. Rimac won’t ever be a Tesla competitor, but efforts like this DO move the needle, in regards to the mission. It provides more proof that ICE is just inherently inferior along several vectors, and provides that proof from a source other than Tesla, and like it or not, there are a ton of people who automatically dismiss messages from or about Tesla / Elon. They are wrong, but they exist, so they matter, if we are talking about changing the world’s opinion and trajectory. It also proves that this kind of tech isn’t so magical that Tesla is the only company capable of making a fast EV — arguably another key to getting worldwide adoption.
granted Rimac’s impact is more about perception and optics than physical vehicles on the road, but it’s a mistake to think that doesn’t matter.
Elon Musk FSD
One step back , Two steps forward
One step back , Two steps forward
Chances a high I get an invitation. Expecting a ton of people wanting to participate. In any case, I will report about it - likely with video and/or article
I think the Tesla already did the job because to convince people it's about the entire ecosystem..from getting products into customers hand, attainability, charging stations, software, reliability, and service stations. The problem is Tesla became the only answer to EVs as no other manufacturers solved the above issues.
Rimac pretty much solved none of the above and only proves it came beat every car in speed...which Plaid pretty much did this already and being attainable via price and production. The problem with these superlatives ..like longer range than Tesla, or faster acceleration than Tesla, or less panel gaps than tesla..they all became distractions as no one is focusing on what is important. They end up using these superlatives to attack Tesla by saying "hey look Tesla is nothing special" vs "see you guys should go electric".
You are so right but there is no defined defination of "Production car". I think it should only include cars at a certain widespread production level say 5,000 per year. Less than 200 is not IMHO, a production car but rather a special edition.
In the world of finance the terms large cap, mid cap and small cap are not defined by anyone either. When I worked for Janus (15 yrs ago) we defined small as anything less than 1.5 billion. Other investment companies e.g. Fidelity at the time defined it as any company less than 2 billion. And don't even get started with the whole micro cap and mega cap. The point is that it is not a defined number accepted by the financial services industry. Nor is the term "production car" other than the fact someone is making more than a couple for sale to the public.
Artful Dodger
"Neko no me"
World auto production over the next 3 years could be up to 300 million cars (or as few as 200 million).
Revera plans to produce ~0.00005% of those cars. Imma call that NOT mass production.
Fun fact: there are 18x more Billionaires in the world right now (2,755) than the planned number of Nevers. Its a toy for the idle rich. It does NOT move the needle.
On the other hand, Tesla is likely to produce ~6 million cars during 2022-24. That WILL move the needle. Even more significantly, this growth in EV market share creates a FORCING FUNCTION that makes the other Auto majors accelerate their own EV plans.
This MATTERS. The climate only reacts to net changes in forcings, not PR efforts.
Revera plans to produce ~0.00005% of those cars. Imma call that NOT mass production.
Fun fact: there are 18x more Billionaires in the world right now (2,755) than the planned number of Nevers. Its a toy for the idle rich. It does NOT move the needle.
On the other hand, Tesla is likely to produce ~6 million cars during 2022-24. That WILL move the needle. Even more significantly, this growth in EV market share creates a FORCING FUNCTION that makes the other Auto majors accelerate their own EV plans.
This MATTERS. The climate only reacts to net changes in forcings, not PR efforts.
Last edited:
Elon bought about a decade ago a McLaren F1 that is now worth +$20 millions. One of the reasons why the McLaren is so expensive is the exclusivity. Only about a hundred cars were built. A Corvette is now faster, but every other simp can buy one
When Bugatti release a special edition car, they usually limit production to about 20. Some crazy billionaires will pay $10 millions for those car only because nobody else have them. It will be the same with the Nevera.
If the Nevera is successful, people will be talking about it for years even if they have never seen one in real life. That is the purpose of halo cars.
I feel like some people here don't understand the supercar market at all. 150 seems fine for a small production supercar.
You guys realize, that Rimac is as much an OEM as they are a Tier-1 supplier for many other car companies?
This car is their advertisement. They'll sell drivetrains to whomever needs one. More customers means more EVs on the road. And at much higher numbers....
This car is their advertisement. They'll sell drivetrains to whomever needs one. More customers means more EVs on the road. And at much higher numbers....
Similar threads
- Locked
- Locked
- Locked
