Cybertruck with only basic Traffic aware cruise control far exceeds the Adaptive Cruise on our Explorer which won't slow below 20 MPH. Add in some lane keeping and I'm happy.
-
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
Musskiah
DisGruntled
Reposting...for posterity
Knightshade
Well-Known Member
That is certainly one theory.
However Xpeng somehow managed to produce a comparable system anyway, in vastly less time and with vastly less data.
Note that doesn't mean getting BEYOND L2 is equally likely for both...It's possible the need for massive quantities of such data is crucial for say L4 driving but much less so at L2.
But we're talking present day capabilities-- the OP claiming nothing TODAY was comparable to V12 TODAY and that just ain't so.
Not that the humans attend much in the first place. So will it eventually be able to tell if the driver is drunk or stoned? (Ah, so that's why they want me to take off my sunglasses.
)Assuming L4 gets hung up for a while longer, I guaranteed there's some data mining there on drivers. What if Grok could tell from speech + vision, then train on it using collision data?
What if Drunk Uncle was responsible for a majority of FSD interventions gone bad (now or future)?
Car says...
"I think you should hop in the back seat for a nap, I can't move while you're sitting here, sorry."
(Driver disables FSD)"Sorry, your insurance won't allow this. Driving for all scenarios and conditions is available for ~10x what you currently pay. Interested?"
Mr. Buffett, chew on that thought for a while longer. You don't have to go down with the ship.
Musskiah
DisGruntled
The bottleneck for Autonomy is real world video data. Tesla has more of it. It's not theory.
You keep insisting that just because you call them "L2" it means they are equivalent.
Do you consider FSD V10 and FSD V12 to be comparable systems?
You keep stating that there are other systems comparable to FSD v12, so what are they? And no not in the technical ratings of said systems, I mean in real world driving experience, what other driving assistance system is comparable to Tesla FSD v12? Because I'm not aware of any such thing, not even from China?
Knightshade
Well-Known Member
No, I don't.
I keep insisting that because both provide human-supervised door-to-door L2 they're comparable.
Supercruise is also L2 but NOT comparable because it only works on a narrow band of highways.
YOU have yet to explain how they're not comparable.
I literally cited one, with a link to a whole story about it, last page, that provided L2 door to door city streets in China.
But we're not discussing autonomy. At all.
We're discussing L2 city streets- which is what V12 currently is, and which there's at least one comparable such system available in China from a company not-Tesla.
This article contain a quotes from XPeng about their "Infinite XNGP", which is their "go anywhere" system. It was first announced on February 29th of this year. So it's quite new.
Here is the quote:
This sounds like they have taken a similar approach to what Tesla did before V12. I will admit that something could have been lost in translation. So if you have evidence that XPeng's system is end-to-end rather than utilizing heuristics, please post it.
FSD V12 proves that an end-to-end system will provide superior performance to systems that use heuristics. Therefore, XPeng's system is not comparable to FSD V12.
If you have evidence to the contrary, please post it.
Knightshade
Well-Known Member
Sure.
But at least 3 times since Feb 29 you've claimed it doesn't exist (that there's no go-anywhere city streets system from anyone other than Tesla) and each time I've corrected you.
Why keep repeating an untrue claim?
If a user can go door to door safely witht he system- why would they care HOW it happens behind the scenes?
That matters a lot for potential future systems that don't require a driver-- but that's not what V12 is EITHER today.
We're talking comparable TODAY, not "which one might be much better some nebulous time in the future"
Dude.
You insisting it's true isn't "proof"
Even more insulting you then try and toss the PROVE IT back without having proved jack yoursef.
YOU claim it's not comparable performance.
PROVE IT.
Except there's no evidence. At all. That performance is "superior" TODAY on V12 compared to their system.
Superior performance how? By what measure? By whom?
Again- if you have evidence to the contrary- provide it.
You don't, can't, and won't.
Hence they are comparable.
Unless you keep narrowing what "comparable" means in a continued desperate attempt to not admit your claim is false.
You seem to have ignored willow_hiller's actual intent, and then focused on a later part of Tesla's statement...and then, even in that jump, you seem to have misread or purposely ignored key details to make a new, mistaken, anti-Tesla point.
I believe willow_hiller was pointing you toward Tesla's statement that they INCLUDE in their autopilot crashes any event that happens up to 5 seconds after Autopilot gets disengaged (obviously by the user or otherwise). This was a direct counter to your false claim that autopilot was programmed to automatically disengage 1 second before an inevitable crash, and that Tesla would then count that as a human-only crash.
You seem to have ignored that, and focused on a later statement by Tesla about airbag detection for crash counting...and you completely ignored Tesla's phrase to include "other active restraint" and "speeds as low as 12mph". Oddly, despite this selective reading in Tesla's report, you were still ready to quote just the "right" part of the NHTSA's 6-page document -- for anybody interested: https://static.nhtsa.gov/odi/inv/2022/INCR-EA22002-14496.pdf .
The full quote from Tesla's description is:
"We also receive a crash alert anytime a crash is reported to us from the fleet, which may include data about whether Autopilot was active at the time of impact. To ensure our statistics are conservative, we count any crash in which Autopilot was deactivated within 5 seconds before impact, and we count all crashes in which the incident alert indicated an airbag or other active restraint deployed. (Our crash statistics are not based on sample data sets or estimates.) In practice, this correlates to nearly any crash at about 12 mph (20 kph) or above, depending on the crash forces generated. We do not differentiate based on the type of crash or fault (For example, more than 35% of all Autopilot crashes occur when the Tesla vehicle is rear-ended by another vehicle). In this way, we are confident that the statistics we share unquestionably show the benefits of Autopilot."
An "active restraint" would include seatbelt tensioners...which can activate in much more minor events that don't require airbag deployment. Active restraints like seatbelt tensioners are going to activate in way more than your highlighted "18% of police reported crashes" that require actual airbag deployment.
Tesla also clearly states that they assign statistical "blame" to Autopilot even when the other driver is at fault. Tesla can often have full video of accident events, and they could use that to readily argue something as not Autopilot's fault...but they don't do that for this data. To be fair, the general "miles between collisions" data likely also count it as a collision for both drivers no matter the fault...but Tesla is obviously showing effort here to not be unfair with the data.
And, we can also see here that Tesla's systems automatically notify them of crashes. At least in the US, we all know many minor accidents are just never reported to police or insurance companies if the parties agree to just handle it between themselves. Police involvement and the threat of higher insurance rates is a large dissuading factor. So, it's pretty reasonable to assume that Tesla might actually be recording a higher percentage of crashes in Tesla vehicles than are ever recorded ANYWHERE for cars from other manufacturers.
Finally, anybody with any background in science learned about "uncertainty" in about seventh grade. Uncertainty is not bad, it is a recognition of the fact that no data is perfect. Tesla tells us how they collecte their data...it is not and will not ever be perfect. And that is fine and expected.
concur the lack of lane keeping is an issue ... but with the number of people driving the adjacent lane doing 70 mph and snapping photos of the CT agility to get out of the way is critical...
the amount of love this truck is getting is mind blowing .... went by an intermediate school today past eh school yard during gym class and i could hear the kids screaming Cybertruuuuck!!!!I have never claimed that XPeng's system does not exist. I have claimed it is not comparable to FSD V12. And unless XPeng has made some AI breakthrough that nobody here knows about, I am absolutely correct.
I notice you haven't answered my question.
Do you consider FSD V10 and FSD V12 to be comparable?
I don't know the details of the case, and I am far too dumb to understand the implications...but isn't Xpeng the company that may have had access to stolen autopilot code from a nefarious engineer that left Tesla? Lawsuit/court case started back in 2019 on that...
Thoughts about that definitely come to mind for me when I see a statement like "Xpeng somehow managed to produce a comparable system...in vastly less time and with vastly less data."
Quick google work does seem to show that XPeng's source code doesn't look like a copy of Tesla's...and correlation is not causation...but it's interesting there is a correlation between a door-to-door L2 competitor to Tesla and this case. Even just hiring former Tesla engineers, while totally legal, is a decent way to shortcut things a bit and changes my evaluation of how impressed I am with a company for developing a similar product faster.
Last edited:
And then I read a full page here of noise. Seems I could be buying more TSLA today even.
What's with the feeding frenzy?
What's with the feeding frenzy?
You keep stating that there are other systems comparable to FSD v12, so what are they? And no not in the technical ratings of said systems, I mean in real world driving experience, what other driving assistance system is comparable to Tesla FSD v12? Because I'm not aware of any such thing, not even from China?
I'm watching this company:
Wayve: Reimagining Autonomous Driving with Embodied AI Technology
Learn how Wayve is leading the way in automated driving with their advanced Embodied AI technology.
wayve.ai
You can see videos online of them navigating UK streets. Just closed $1+ bil funding with MSFT/NVDA/others. Maybe you just need better AI vs. videos? I've stated before that if I was a top Tesla engineer, working here seems like a no brainer in terms of impact/upside.
Yes, and the engineer who did it had to pay damages to Tesla. It is still not known for sure if XPeng actually received or used the stolen data.
Knightshade
Well-Known Member
Based on what evidence?
"ONE USES AI!!!" is not evidence their capabilities, today, aren't comparable.
It's a claim without proof.
Again if both systems can safely get me from A to B, always needing supervision but rarely needing intervention, on all roads in the country-- why would I as a consumer care HOW it's doing it behind the scenes?
The systems are certainly comparable.
One might well be a better precursor to a FUTURE non-supervised system-- but that wasn't your claim.
YOU made the claim, YOU keep trying to insist another system with the same current capabilities isn't comparable- the burden to actually support those claims remains on you.
Do you have data showing significant differences in disengagement rates? Do you have documentation that their system doesn't "really" work in some places that FSD does?
Or are you just taking it on complete blind faith Teslas MUST be better -right now- because they use the phrase AI more?
Similar threads
- Locked
- Locked
- Locked
