Tesla to increase cost of FSD beta software beyond its $12,000 price tag

[cbdream99]

Xpeng Xpilot 3.5 seems to be pretty decent if you watch YouTube. It's sold as L2 ADAS but it's more like an L4 with you as a safety driver!

I agreed, this is quite impressive.

 

[diplomat33]

I agreed, this is quite impressive.

The FSD visualizations on the screen look like a 3D view of their HD map. Because it is a HD map, it looks much more detailed and more sharper than Tesla's FSD visualizations. It is very cool although objects do seem to "glitch" a lot. The perception and planning also look to be good. The car seems to be able to handle busy city driving pretty well.

 

[CyberGus]

Video is age-restricted? Because it runs over children??

 

[Tam]

The FSD visualizations on the screen look like a 3D view of their HD map. Because it is a HD map, it looks much more detailed and more sharper than Tesla's FSD visualizations. It is very cool although objects do seem to "glitch" a lot. The perception and planning also look to be good. The car seems to be able to handle busy city driving pretty well.

I think the maximum it deserves is an L4 because it is geofenced to work only where the roads have been HD pre-mapped.

I don't think the gore point Mountain View fatality with the 2017 Tesla Model X would happen because this system has no trouble identifying gore points:



Xpeng was founded in 2014 while Tesla has been selling Autopilot since then, but Tesla still has gore point problems even with the latest FSD beta.

 

[diplomat33]

I think the maximum it deserves is an L4 because it is geofenced to work only where the roads have been HD pre-mapped.

Geofencing would make it L4. Pre-mapping does not make a system L4. Pre-mapping has nothing to do with the levels.

Ultimately, it depends how reliable the system is. If Xpeng wants to remove driver supervision then they need to make sure it is reliable enough. If the system is not reliable enough to be trusted as driverless then they need to keep driver supervision. My understanding is that the system is "L2 hand-off" since Xpeng requires driver supervision.

 

[Dan D.]

Xpeng Xpilot 3.5 seems to be pretty decent if you watch YouTube. It's sold as L2 ADAS but it's more like an L4 with you as a safety driver!

Terrible accident from an XPeng P7 "with its driver assistance system on."

I'm not sure quite what version of the software this car is using, whether it's Xpilot, nor exactly what Lane Centering Control is in relation to their L2 system. Not a good report for the company in any case.

News and online sources have reported on the car accident in Ningbo City of neighboring Zhejiang Province on Wednesday, which involved an XPeng P7 that smashed into a broken-down vehicle parked in front of it on the lane.

The accident caused the death of the person standing at the end of the car in front.

I recommend you do not watch the video,

Spoiler: Disturbing Video

It is worth noting that the crash happened when the P7 driver was using the Lane Centering Control, a driver assistance function of XPeng, with the speed set at a highway limit of 80 kilometers per hour.

According to the P7 driver, the warning didn't work before the crash, adding that he was also distracted at the time.

XPeng's user manual warns several times that the LCC is a driver assistance function, not a fully self-driving attribute, and that the driver still needs to keep his hands on the steering wheel even after the LCC function is activated, and should take over the steering wheel when necessary.

 

[Tam]

....Not a good report for the company in any case...

Thanks for the news and warnings.

Xpeng started in 2014 and imitated Tesla as if it were a twin by taking the offer of Tesla's open patent.

So much so that it also believed in radar and camera only as the path to Autonomous with no HD pre-mapping needed, citing the low cost.

Up to the point that Elon Musk said Xpeng stole Tesla Autopilot:

Tesla doubles down on claim Chinese EV startup stole its Autopilot source code

Tesla is doubling down on its claim that Xpeng, a Chinese electric vehicle startup, has stolen and is using its...

electrek.co

Initial cars were quite simple, just like Tesla's sensor suite.

It was not until later last year that Xpeng departed from Tesla's philosophy. It started beefing up its sensors to 32, including 2 LIDARs, probably because it now knows it can financially afford the robust sensor fusion with HD pre-mapping.

Thus, I do not doubt that older Xpeng (Tesla imitation or "stolen" Tesla double) would still have the problem with stationary obstacles.

Since the article mentioned LLC Lane Centering Control and not Xpilot 3.5, I believe this car does not have newer hardware, LIDAR, and software, Xpilot 3.5.

 

[hw13]

 

[dbldwn02]

He keeps saying "wide release"
I'm pretty sure he doesn't know what that actually means.

 

[VT_EE]

Ummm, not a chance. Good luck finding buyers for $15k FSD unless it includes HW4 and some MUCH better software.

 

[michael805]

$12k is insane. $15k even more insane. Even if it worked I don’t think many people will buy it.

 

[AlanSubie4Life]

The new price of $9k for FSD coming in September definitely does not seem worth it given current features and functionality, and even after accounting for any improvements that can reasonably be expected in the next year or two.

I think I would pay $2k-$3k for entertainment value and some of the features it provides.

 

[AndrewZ]

Looking at this timeline

-10.69 going out 20-Aug-22 to what appears to be a select group of YouTube FSD Beta testers
-10.69.1 going out to 10k people around 27-Aug-22
-10.69.2 "wide" release around 02-Sep-22
-Price increase one business day later on 05-Sep-22

I'd say there's an obvious strategy here in trying to generate excitement with the initial release to YouTubers and particularly people watching Chuck's left turn, with Elon trying to create FOMO to push people towards locking in FSD at $12k before the price increase is in effect and possibly before the wider group has had time to really assess the new release, before any broader issues become apparent, and before any of the issues are known to prospective buyers.

Elon truly is a master marketer. This isn't about getting people to buy at $15k, it's about creating FOMO and getting more people to buy it at $12k.


Rinse and repeat for the next one.

 

[VT_EE]

Maybe I'll get to test out this new awesome FSD load before I sell my FSD Beta-enabled M3 and pick up my new EAP-enabled M3. I'm sure it will be awesome. Elon says it will be mind blowing, just like all the loads since I bought FSD 4 years ago.

 

[AlanSubie4Life]

Elon says it will be mind blowing, just like all the loads since I bought FSD 4 years ago.

Still on target as far as I am concerned. If anything they are year ahead of where I figured they would be. I guess it is all about expectations (and what was paid). Haha.

 

[hw13]

-10.69.2 "wide" release around 02-Sep-22

What is "wide" release? Is it to 100K beta testers or 1M FSD owners?

 

[Supcom]

FSD will go to $20K when beta is released to all FSD licensees.

What is "wide" release? Is it to 100K beta testers or 1M FSD owners?

Wide release, in this case, is release to the ~100K beta testers. Release of beta to all FSD licensees is still just a goal for this year.

 

[srs5694]

I was just at an EV event yesterday, and I told several people that FSD isn't worth the $12k that Tesla demands for it today. $15k? That's a joke, and not a very good one.

 

[hw13]

FSD will go to $20K when beta is released to all FSD licensees.

Wide release, in this case, is release to the ~100K beta testers. Release of beta to all FSD licensees is still just a goal for this year.

Then what are people paying $15K for? There have been no improvements to existing FSD features, only beta ones.

 

[ElectricIAC]

FSD will go to $20K when beta is released to all FSD licensees.

Wide release, in this case, is release to the ~100K beta testers. Release of beta to all FSD licensees is still just a goal for this year.

2+ years waiting on that goal.