Elon: FSD Beta tweets

[JulienW]

Intervention = disengagement when comparing to a driverless cross country summon, or as near as makes no difference at least from out here in the cheap seats.

Any drive with either is just boring, so what, call me when it's actually zero.

Actually it is disengagement always = Intervention but not always the other way around.

Many times my patiences is challenged and with the Go pedal right under my foot I can't resist even if FSD would do about the same thing in another second or two. Maybe it's me that needs an intervention.

 

[OxBrew]

It depends on the intervention. An intervention to pass a slow semi right away rather than waiting for FSD to figure it out, or to get around someone traveling super slowly in the fast lane is not necessarily the same as a disengagement, especially when thinking about it as cross-country summon.

Many intervention types could be disqualifying (may as well be disengagement), as listed in examples above.

Plenty of optional ones, plenty of mandatory ones. They have to be dealt with case by case since there are so many possibilities, but some are categorically one or the other of course.

Sure, but for a coast to coast test case, the driver needs to understand that any intervention equals fail, just like disengagement, or people like me will dismiss it as it will be a staged, curated stunt that does nothing to answer the question: can the car drive unattended across the country?

(excluding plugging in the charger. But not excluding parking at the charger)

[edited: strike through and last parenthetical part]

 

[JulienW]

Sure, but for a coast to coast test case, the driver needs to understand that any intervention equals fail, just like disengagement, or people like me will dismiss it as a staged, curated stunt that does nothing to answer the question.

....but it IS an L2 system and NOT L5. Even L4 might not be able to do this.

 

[OxBrew]

....but it IS an L2 system and NOT L5. Even L4 might not be able to do this.

so what?

"In ~2 years, summon should work anywhere connected by land & not blocked by borders, eg you're in LA and the car is in NY"

Elon, 10 Jan 2016

So anytime anyone says anything about driving NY to LA, this is the base line. It is technically capable of making the trip. I'll give 2 caveats, because it's hyperbole: legality, and plugging in. Legal, because obviously, plugging it, because robot plug in is a waste of resources. No one actually needs to summon their car from that far away, Elon was just saying how capable it will be. Prove it, with zero interventions.

 

[Mark II]

Less than 200 miles and those roads shouldn't be too busy.
View attachment 896101

Why no one is talking about California to New York on a single charge! The pace of battery innovation is staggering!

 

[Daniel in SD]

Why no one is talking about California to New York on a single charge! The pace of battery innovation is staggering!

Tesla Semi with a Tesla Megapack could probably do 2500 miles. That would be a fun delivery stunt.

 

[JulienW]

so what?

"In ~2 years, summon should work anywhere connected by land & not blocked by borders, eg you're in LA and the car is in NY"

Elon, 10 Jan 2016....

I'm stating facts and you are quoting aspirations. Elon has said literally 100s of things that he expected to happen that haven't came to fruition including things like SpaceX landing on Mars in 2018.

Whole Mars' main goal is YouTube views, period.

 

[DanCar]

Tesla Semi with a Tesla Megapack could probably do 2500 miles. That would be a fun delivery stunt.

Good one. I have a new product suggestion. Tesla megapack trailer that can be hooked up to Model 3 or whatever.

 

[aronth5]

Tesla Semi with a Tesla Megapack could probably do 2500 miles. That would be a fun delivery stunt.

Might be a bit of problem since a single Megapack weighs 51k pounds. That plus the weight of the semi and trailer would exceed the total weight allowed.

 

[aronth5]

This year? 10s to 100s of interventions and disengagements for cross-country trip.

Using NNs for planning is going to lead to a lot of regressions. Even after it being over a year in development up until now, it is not reasonable to expect interventions and disengagements will go to zero - because that is not the design intent!

Yeah lots of disengagements and interventions there and that doesn’t count the freeway ones.

It’s just an incredibly difficult problem and not reasonable to expect that there will be workable drives cross-country that have fewer than 100s of interventions and disengagements any time soon.

I expect as low as single-digit disengagements and 10 to 100ish interventions. Some of the interventions will be effectively disengagements (for example pushing the accelerator to make the car proceed, or speed adjustments for safety and courtesy), and will be counted as such.
...

Glad to see you revised your original projection of interventions from 100s to now 100ish. And disengagements to single digits. Big difference.

 

[AlanSubie4Life]

Glad to see you revised your original projection of interventions from 100s to now 100ish. And disengagements to single digits. Big difference.

“As low as….”

Anyway, a lot. Would not be surprised if the normal result was 10s of disengagements. Is lower possible? Sure. It’s a distribution.

Definitely 100s of interventions is likely/certain.

So no difference.

It’s hard to put anything other than an order of magnitude guess on it. It depends a lot on the driver (and for interventions, what we count - very poorly defined here). Since it is Omar, will be lower, but the exact value will still vary a lot. I am just saying I expect it to be low 10s and low 100s. But lower results in some conditions, carefully chosen, seem possible.

Regardless, not even close this year.

 

[Daniel in SD]

Might be a bit of problem since a single Megapack weighs 51k pounds. That plus the weight of the semi and trailer would exceed the total weight allowed.

The promotional video had it carrying 11 jersey barriers which apparently is 44,000 pounds so it's close. I suppose they could get overweight permits which is probably expensive but if it's just a one time stunt might be worth it.

 

[aronth5]

“As low as….”

Anyway, a lot. Would not be surprised if the normal result was 10s of disengagements. Is lower possible? Sure. It’s a distribution.

Definitely 100s of interventions is likely/certain.

So no difference.

It’s hard to put anything other than an order of magnitude guess on it. It depends a lot on the driver (and for interventions, what we count - very poorly defined here). Since it is Omar, will be lower, but the exact value will still vary a lot. I am just saying I expect it to be low 10s and low 100s. But lower results in some conditions, carefully chosen, seem possible.

Regardless, not even close this year.

Please make up your mind and don't change the goal posts. Is it 100ish or 100's?

 

[AlanSubie4Life]

Please make up your mind and don't change the goal posts. Is it 100ish or 100's?

No goalposts have been harmed thus far!

Low 100s. So more than 100 and less than 400 interventions, as I define interventions (most straightforward way: signaling with stalk, scroll wheel speed adjustment for any reason, accelerator).

But as I said I think if “optimized” and willing to put up with a lot of non-optimal results, it could perhaps be less than 100. Might be tough and frustrating (and certainly dangerous!) but seems possible.

 

[Mark II]

Please make up your mind and don't change the goal posts. Is it 100ish or 100's?

We talking Tesla here. Changing goal posts is not only allowed, it is expected.

 

[DanCar]

Elon reviewing V11 next week for possible release.

https://twitter.com/x/status/1616635070156255232

 

[JulienW]

Elon reviewing V11 next week for possible release.

https://twitter.com/x/status/1616635070156255232

You know the old saying:

Fool me once, shame on you. Fool me twice, shame on.........well you? Fool me 3 times, shame on.......you again?..... Fool me 8 times, shame on you.... Fool me 12 times, shame on you......

Now pass me that Kool-Aid so I can take another hit because it "makes it fun to be thirsty!" for Elon's next it's a comm'n Tweet .

 

[DWtsn]

Elon reviewing V11 next week for possible release.

https://twitter.com/x/status/1616635070156255232

 

[hugh_jassol]

Elon reviewing V11 next week for possible release.

https://twitter.com/x/status/1616635070156255232

I remember Elon saying like 2018 for possible release

 

[thewishmaster]

Elon reviewing V11 next week for possible release.

https://twitter.com/x/status/1616635070156255232

I guess code reviews at twitter are going well so he’s reviewing v11 code next week?