Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

This site may earn commission on affiliate links.
I think Waymo and other geo-fenced companies have proved that limited robotaxis are possible.

The question is - is a "generic" robotaxi possible with HW3. The pace of improvement (miles per disengagement) over the last year shows small but meaningful improvement. But we need to see 100x to 100x improvement for robotaxi to happen.

I think I'm comfortable now in saying robotaxi with HW3 will not happen in the next 5 years.
Agreed. But I wouldn’t be surprised to see a level 3 system within two or three years. The amount of potential catastrophic failures has been dropping for us. It still continuously fails on the same three intersections but we have also noticed an improvement in the incorrect signalling. I’ll put 5 bucks on a 2025 level 3 in fair weather conditions.
 
I’ll put 5 bucks on a 2025 level 3 in fair weather conditions.
But will Tesla call it level 3 and take responsibility? I hope not ... Unless they have to for competitive reasons.

Since even a single accident can result in millions in payouts, there is no reason to declare anything other than robotaxi level quality and assurance.

I can also get 70% of my drives to be zero disengagement. But miles / disengagement is still about 1/10. Needs to get to 1/10,000 to be robotaxi level.
 
Amazing. Is there a major incentive for EVs a la Norway?
Taiwan requires annual license tax and gas usage tax and currently EVs are excempted from both. Annual license tax's excempt however will expire by end of 2025.

As for how much they are, it depends on engine displacement and very minuscule. For example
2401-3000cc has license tax NTD$15,210 (USD$495) and gas usage tax of NTD$7,200 ($234)
30001-4200cc has NTD$28,220 (USD$919) and NTD $8,640 (USD$281)

 
Is this a digg like calling CNBC CNBS? I tried googling Tulsa, Mexico but it just offered me flights from Tulsa to Mexico.
As two people have pointed out it is Toluca. Spellcheck corrected me to Oklahoma and I did not notice.
The Toluca plant produces Jeep Compass for USMCA and supplies parts to other locations, while importing parts from the Pernambuco, Brazil factory which was the launch factory for Compass. FWIW, the Compass is built on the DaimlerChryslerMitsubishi GS platform, also used for models from Chrysler, Dodge, Fiat and Peugeot/Citroën.
Apart from Brazil and Mexico Compass is also produced in China and India.
The platform was originally designed by Mitsubishi.
Badge engineering is alive and well. Allegedly this platform will be discontinued when the new BEV platform is launched ~2025, according to the Pernambuco rumors In Recife.

Pernambuco is one of the Brazilian States trying to attract Tesla, as are at least three others. Both Rio de Janeiro and São Paulo are pitching their aerospace and automotive industries.
 
I think Waymo and other geo-fenced companies have proved that limited robotaxis are possible.

Only if you define “possible” as throw so much money at the problem that you can’t hope to financially break even.

The question is - is a "generic" robotaxi possible with HW3. The pace of improvement (miles per disengagement) over the last year shows small but meaningful improvement. But we need to see 100x to 100x improvement for robotaxi to happen.

I think I'm comfortable now in saying robotaxi with HW3 will not happen in the next 5 years.

Eh. Still to be seen. I do keep running into edge cases with FSD Beta. But if the automated learning closed loop is good and getting better all the time, we might see accelerated progress. But I agree that progress in one year has not been super fast.
 
Last edited:
Badge engineering is alive and well.
Honda is interesting, apparently 80% of their global market is the US and China.

US Honda EVs are apparently being made mostly by GM. Chinese Honda EVs made by a Chinese partner.

With the move to EVs, the lawn mower and motor bike markets are under threat.

I can remember when Honda was dominating F1 racing, and that translated into their vehicles having an edge.

Honda might step up their efforts in EV making, but at this stage they are slow starters.
 
Last edited:
Posted in Tesla Semi thread as well. This is the the type of headline/article we need to see more of! Originally posted in Evannex and CleanTechnica this morning;
 
Only if you define “possible” as throw so much money at the problem that you can’t hope to financially break even.
Technically feasible ;)

Eh. Still to be seen. I do keep running into edge cases with FSD Beta. But if the automated learning closed loop is good and getting better all the time, we might see accelerated progress. But I agree that progress in one year has not been super fast.
Unfortunately since Tesla won't publish the disengagement rates, we need to go by what is available through personal logging of many of us. I want to see improvement in this crowd sourced dashboard, if not my personal log.

1670813603900.png


Humans are at about 10,000 miles per accident (similar rate by Waymo). So, we have ways to go .... moreover the rate of improvement is definitely nothing like exponential.

One last thing - we have expected / hoped for accelerated improvement multiple times over the last few years. Even though industry experts have always argued for slow and steady improvement (infact many argue each "9" is more difficult to achieve than the previous one).
 
Technically feasible ;)


Unfortunately since Tesla won't publish the disengagement rates, we need to go by what is available through personal logging of many of us. I want to see improvement in this crowd sourced dashboard, if not my personal log.

View attachment 883947

Humans are at about 10,000 miles per accident (similar rate by Waymo). So, we have ways to go .... moreover the rate of improvement is definitely nothing like exponential.

One last thing - we have expected / hoped for accelerated improvement multiple times over the last few years. Even though industry experts have always argued for slow and steady improvement (infact many argue each "9" is more difficult to achieve than the previous one).
Disengagement rates would be meaningless at present, other than as comparison to the same numbers from a previous time. Many beta-testers are like me and drive in crowded spaces. If the vehicle behind me appears to be in a hurry I am not letting the Tesla take it's time to sort out the intersection. Instead, I take over, get the left turn at the stop sign done, and then reengage. It's not a failure, it's an inconvenience, and there's a real difference between the two.
 
Disengagement rates would be meaningless at present, other than as comparison to the same numbers from a previous time. Many beta-testers are like me and drive in crowded spaces. If the vehicle behind me appears to be in a hurry I am not letting the Tesla take it's time to sort out the intersection. Instead, I take over, get the left turn at the stop sign done, and then reengage. It's not a failure, it's an inconvenience, and there's a real difference between the two.
That is not how a lot of us drive. I've detailed logs on every one of my drives and have recorded all my drives. A lot of people on TMC do that. You should visit the FSD sub-forum more often. (ps : Also try to be not bothered by other vehicles hurrying you up, record your drives and see what disengagement rate you get).

You know how DirtyTesla drives, for eg. Here is his disengagement rate.

1670818061463.png


You can also see CYBRLFT - he keeps detailed logs and posts videos as well.

 
Last edited: