Elon: FSD Beta tweets

[spacecoin]

Where does that figure come from? Most disengagements aren't safety-critical anyways, in my experience. Mostly comfort issues, not getting honked at, or missing a turn which would then add travel time on re-route. None of those things really have to be perfect to get the job done in a basic sense.

If you feels it’s close to autonomy then I would challenge you to upload a video where you sit in the back seat or blind folded while the car drives for a few hours.

The number has been used by experts, and typically humans have an accident about 1-2 times per life time on average.

 

[AlanSubie4Life]

So many maybe-not-Beta Tweets today.

FSD (Supervised, not Beta) is going wide!!!

Whole Mars:

FSD licensing
1. Allow OEM to integrate it into any car for free or at cost. Car owner pays for FSD license
2. Active safety & basics comes free
3. Leverage Supercharger deals. Offer same pricing as Tesla cars for EVs that integrate FSD hardware & share data

Elon:

Tesla would be happy to do such deals

https://twitter.com/x/status/1775731044836381042

I wonder if these cars will also be capable of FSD Beta, not just FSD Supervised?

What a time to be alive.

 

[JulienW]

.......I wonder if these cars will also be capable of FSD Beta, not just FSD Supervised?

What a time to be alive.

Cars would have to buy and use Tesla's FSD computer and use the same camera/locations (be complete ground up redesigns). Then each car would have to be tuned in FSD which is obviously a HUGE bar. Even the Tesla Cybertruck having all the hardware and even BEING a Tesla doesn't have FSD (or even AP) yet. So easy to say and offer but likely a manufacturing and implementation nightmare that no OEM wants to endure along with capitulation to NEEDING Tesla.

 

[DanCar]

FSD can go 4 wheeling.

https://twitter.com/x/status/1775414062136078398

 

[jabloomf1230]

Where does that figure come from? Most disengagements aren't safety-critical anyways, in my experience. Mostly comfort issues, not getting honked at, or missing a turn which would then add travel time on re-route. None of those things really have to be perfect to get the job done in a basic sense.

You left out the main reason that I personally disengage. The navigation isn't always IMO, the best route. Tesla really needs to allow both dragging route nodes and saving a route as a favorite. The kluge of adding intermediate stops to a route too cumbersome to use.

 

[diplomat33]

Are you aware that you need at a minimum 10000 miles per disengagement to leave L2? Probably a lot more.

SAE levels are not based on specific disengagement rates. You do not graduate from one level to another when you reach a certain disengagement rate. You can have a system with 10k miles per disengagement and still be L2 because it is just does lane keeping on highways. And you can have L4 with only 100 miles per disengagement as we see from several of the companies that report L4 testing to the CA DMV each year.

What I think you meant to say is that you need ~10k miles per safety critical intervention in order to go from L4 supervised to L4 unsupervised in a certain limited ODD.

 

[spacecoin]

SAE levels are not based on specific disengagement rates. You do not graduate from one level to another when you reach a certain disengagement rate. You can have a system with 10k miles per disengagement and still be L2 because it is just does lane keeping on highways. And you can have L4 with only 100 miles per disengagement as we see from several of the companies that report L4 testing to the CA DMV each year.

What I think you meant to say is that you need ~10k miles per safety critical intervention in order to go from L4 supervised to L4 unsupervised in a certain limited ODD.

I thought that was self-evident... And same rules apply to L3 as L4 in terms of required reliability in the ODD.

In simpler terms: "In order to take on the full driving task the system need not to break rules, hit things or drive off he road too often". If deploying your software to consumer vehicles in particular. Tesla would go bankwupt in two days if they shipped an "L4 with 100 miles MTBF" since there is no safety driver in a consumer L4, only passengers.

I think it's quite silly to say that you have deployed an L4 under development if it has a crappy MTBF and only a testing permit. It's still under development then, and not a deployed L4, in my view. Otherwise we're blurring the lines and start calling FSD (Supervised) for L4 soon. The consumer can never be an L4 safety driver. You're either driving (L2) or you're not. OEM liability is the difference in practice.

And to be able to take on liability drive the system, you need a 10k+ miles MTBF Q.E.D.

 

[EVNow]

FSD licensing
1. Allow OEM to integrate it into any car for free or at cost. Car owner pays for FSD license

I vote for getting FSD in CTs first

 

[JulienW]

I vote for getting FSD in CTs first

.....and I bet those several 100k Tesla's on 24.8.x or hoping the "next" week for FSDS will happen sometime in the next........whenever too.

 

[ItsNotAboutTheMoney]

I thought that was self-evident... And same rules apply to L3 as L4 in terms of required reliability in the ODD.

In simpler terms: "In order to take on the full driving task the system need not to break rules, hit things or drive off he road too often". If deploying your software to consumer vehicles in particular. Tesla would go bankwupt in two days if they shipped an "L4 with 100 miles MTBF" since there is no safety driver in a consumer L4, only passengers.

I think it's quite silly to say that you have deployed an L4 under development if it has a crappy MTBF and only a testing permit. It's still under development then, and not a deployed L4, in my view. Otherwise we're blurring the lines and start calling FSD (Supervised) for L4 soon. The consumer can never be an L4 safety driver. You're either driving (L2) or you're not. OEM liability is the difference in practice.

And to be able to take on liability drive the system, you need a 10k+ miles MTBF Q.E.D.

A simpler definition would be that if it's profitable, it's deployed.

 

[spacecoin]

A simpler definition would be that if it's profitable, it's deployed.

I think that’s more complicated. Both Tesla and Waymo fail that test.

Waymo is great, not yet profitable.
Tesla sells L2 as L4, profitable, not autonomous.

 

[DanCar]

FSD transfers back on. No source given.

https://twitter.com/x/status/1776074675115766060

 

[tivoboy]

FSD transfers back on. No source given.

https://twitter.com/x/status/1776074675115766060

Technically, it’s been “on” since Sept 2023…just keeps getting renewed. MAYBE a couple weeks off, but it’s a current certainty.

 

[AlanSubie4Life]

Technically, it’s been “on” since Sept 2023…just keeps getting renewed. MAYBE a couple weeks off, but it’s a current certainty.

Yep. One-time deal only.

 

[tivoboy]

Yep. One-time deal only.

Just for two weeks

 

[Knightshade]

Likely they'll keep running it until they've filtered as many of the pre-3/2019 buyers they still owe L4 to out of the mix as possible, as each time they replace that obligation with the newer L2-only promise they've already delivered (once summon works on no USS cars anyway)

 

[spacecoin]

Likely they'll keep running it until they've filtered as many of the pre-3/2019 buyers they still owe L4 to out of the mix as possible, as each time they replace that obligation with the newer L2-only promise they've already delivered (once summon works on no USS cars anyway)

I think you're overthinking it. I think they just want to flip new cars to existing owners that happen to own FSD and won't buy a new one because of their "investment". Turns out that selling the future in advance has drawbacks.

 

[AlanSubie4Life]

 

[tivoboy]

Likely they'll keep running it until they've filtered as many of the pre-3/2019 buyers they still owe L4 to out of the mix as possible, as each time they replace that obligation with the newer L2-only promise they've already delivered (once summon works on no USS cars anyway)

I think you're overthinking it. I think they just want to flip new cars to existing owners that happen to own FSD and won't buy a new one because of their "investment". Turns out that selling the future in advance has drawbacks.

I think at this point, BOTH things can be true.. certainly new sales, a D is a D at this point are priority for cash flow and getting D up. But, I do think there is some either carried obligation or overall liability risk that the pre 3/2019 buyers of some classes represent and moving them out of the liability bucket into the new sale bucket is a positive. There are some holder outers that have just been happy to maintain what they have without taking another 12-15K hit just to keep playing the FSD guinea pig game. And certainly some who think they are holding an asset that has some additional value to market price due to the obligation the company still has.

Personally, I’m not willing to part with my 30K 2018 M3 RWD LR for less than FSD transfer, $5000 discount and 4-5 years of free supercharging (MAYBe with an annual cap).. Overall, the liability to the company is probably between $15k-$20K per unit. Some could make the argument that with the objective of RT, it could be much higher. I’m not in that camp. Frankly, I don’t really think any of these cars are going to be RT. At least not more than a limited geofence version of it - which might be fine to keep the narrative alive.

 

[JB47394]

If I buy a new Tesla and transfer my FSD, does the contractual meaning of FSD change? That is, folks were more or less promised essentially L3 for a long time, and I wonder if Tesla is trying to move people off of that contractual obligation and onto something less capable.