Tesla autopilot HW3

[diplomat33]

I could not read your attachment but I believe that Tesla said that AP2 would have all of the features of AP1 by the end of 2016. We are still waiting for being able to read speed limit signs.

I know that AP1 and AP2 are using different technology but we have been waiting 3 years for features that are still not delivered.

Are you sure Tesla promised AP2 would read speed limit signs by end of 2016? Here is the screenshot of what Tesla promised EAP would have by Dec 2016. It does not mention reading speed limit signs: https://www.teslarati.com/wp-content/uploads/2016/10/Tesla-enhanced-autopilot-upgrade.jpg

Tesla is 3 years late but they have finally delivered all the EAP features on that page.

 

[electronblue]

Are you sure Tesla promised AP2 would read speed limit signs by end of 2016? Here is the screenshot of what Tesla promised EAP would have by Dec 2016. It does not mention reading speed limit signs: https://www.teslarati.com/wp-content/uploads/2016/10/Tesla-enhanced-autopilot-upgrade.jpg

@S85D is right. While Tesla.com promised full EAP delivered by December 2016, Tesla promised AP1 parity by that time separately.

Tesla is 3 years late but they have finally delivered all the EAP features on that page.

In the U.S., yes. Internationally, not yet.

 

[emmz0r]

They did get sued and lost for the feature parity delay.

Tesla agrees to partially reimburse people who bought Autopilot 2.0 in $5 million settlement of class action lawsuit - Electrek

Also "Your Tesla will match speed to traffic conditions" does not necessarily mean sign reading

 

[diplomat33]

@S85D is right. While Tesla.com promised full EAP delivered by December 2016, Tesla promised AP1 parity by that time separately.

Thanks. I did not know that.

 

[ahkahn]

But for those of you who are saying that Waymo is much further ahead and already carrying passengers fully autonomously, are you also saying that the Waymo vehicles have the capability to run coast to coast (like Elon promised before) or are they strictly limited to certain areas inside of known-good roads?

 

[conman]

Aaaaand this thread turned into yet another repeat of the same discussions. HW3 discussion, anyone?

 

[diplomat33]

But for those of you who are saying that Waymo is much further ahead and already carrying passengers fully autonomously, are you also saying that the Waymo vehicles have the capability to run coast to coast (like Elon promised before) or are they strictly limited to certain areas inside of known-good roads?

I could be wrong but I think the waymo cars are geofenced. However, I believe Waymo could do a coast to coast autonomous drive now if they wanted to. They just choose not to because it's not that important.

Tesla cars are not geofenced but they are only L2 driver assist. Waymo cars are geofenced but L4 driverless robotaxis. I would still put Waymo ahead.

 

[ahkahn]

I could be wrong but I think the waymo cars are geofenced. However, I believe Waymo could do a coast to coast autonomous drive now if they wanted to. They just choose not to because it's not that important.

Tesla cars are not geofenced but they are only L2 driver assist. Waymo cars are geofenced but L4 driverless robotaxis. I would still put Waymo ahead.

Fair enough, I think it's at a point of how you slice and dice it. That said, my personal feeling is that Tesla is way ahead in the data gathering game. What does that mean? I saw a demonstration one time on how machine learning works. The more you correct it, the more it learns. Take that game times 500,000 vehicles, that's a lot of learning going on. Once that learning gets digested and pushed back out to the cars, all of a sudden they jump ahead in capabilities. That's why they run features in shadow mode, and then the first iteration of whatever feature usually sucks and then it gets refined from there. Think AP (Karpathy version), NoA, and as of late Smart Summon, and stop light recognition. Keep in mind these features are all FSD necessities that every Tesla owner in the world is training minute by minute... So once all the training is done... Drive a few million (billion?) miles with driver intervention (level 3?) and then all of a sudden Tesla is ready to "flip the switch" on FSD... true FSD... and there you have it. So they literally could jump ahead by light years with the flip of a switch having had their fleet trained by a half a million people.

 

[am_dmd]

Fair enough, I think it's at a point of how you slice and dice it. That said, my personal feeling is that Tesla is way ahead in the data gathering game. What does that mean? I saw a demonstration one time on how machine learning works. The more you correct it, the more it learns. Take that game times 500,000 vehicles, that's a lot of learning going on. Once that learning gets digested and pushed back out to the cars, all of a sudden they jump ahead in capabilities. That's why they run features in shadow mode, and then the first iteration of whatever feature usually sucks and then it gets refined from there. Think AP (Karpathy version), NoA, and as of late Smart Summon, and stop light recognition. Keep in mind these features are all FSD necessities that every Tesla owner in the world is training minute by minute... So once all the training is done... Drive a few million (billion?) miles with driver intervention (level 3?) and then all of a sudden Tesla is ready to "flip the switch" on FSD... true FSD... and there you have it. So they literally could jump ahead by light years with the flip of a switch having had their fleet trained by a half a million people.

I wish I could find out which one of that 500k Teslas that my Tesla learned from now that my cruise control and all AP related functions are broken since 32.1 update.

 

[jebinc]

I could be wrong but I think the waymo cars are geofenced. However, I believe Waymo could do a coast to coast autonomous drive now if they wanted to. They just choose not to because it's not that important.

Tesla cars are not geofenced but they are only L2 driver assist. Waymo cars are geofenced but L4 driverless robotaxis. I would still put Waymo ahead.

I would as well, put Waymo ahead. To be frank, my spanking new 2020 Subaru Eyesight AP-like functionality (not pre-2020 Eyesight, mind you) is currently more refined that my M3. Yes, my M3 offers more, but I was surprised how well Subaru’s first attempt at this works, as compared to the Tesla first attempt a few years ago.

 

[diplomat33]

Aaaaand this thread turned into yet another repeat of the same discussions. HW3 discussion, anyone?

I think we are all waiting for our HW3 upgrade.

 

[conman]

I think we are all waiting for our HW3 upgrade.

Besides, the extremely small number of pilot upgrades, have there been any other reports? Alternatively has there been a single report of a HW2->HW3 upgrade, or only 2.5?

 

[diplomat33]

Besides, the extremely small number of pilot upgrades, have there been any other reports? Alternatively has there been a single report of a HW2->HW3 upgrade, or only 2.5?

No, not that I am aware of. But if Elon's tweet is to be believed, it sounds like upgrades could happen in "a few months" when Tesla is ready to do upgrades en masse. But who knows how long "a few months" is so we may waiting awhile still. But I guess the good news is that if Tesla does do the upgrades en masse, then we will definitely know about it for sure since we will get reports of not 1 or 2 but hundreds getting invites to the schedule upgrade.

 

[ahkahn]

I scheduled for other service and did put on my request to upgrade my car to AP3. They said that I would have to come into the service center for the other service, and they "think" they may be able to hit the AP3 upgrade at the same time. I'm going in the end of the month... we will see.

 

[S100D]

I am just saying that if Elon won't give early adopters free autopilot on their next purchase because Tesla is losing money, he's probably not going to give early adopters a refund on EAP/FSD. In case you are wondering, I am referencing this tweet: Bruce Mittleman on Twitter

This is an easy fix without Tesla losing money. If a Tesla owner already has EAP/FSD, Tesla can take it off the trade in car and put it on the new car. You might want to leave EAP and only move FSD since that is the most expensive.

Hopefully this would be without Tesla tanking the trade-in because it does not have FSD. If most cases you do not get the full value of options when you trade a car in.

 

[dgatwood]

This is an easy fix without Tesla losing money. If a Tesla owner already has EAP/FSD, Tesla can take it off the trade in car and put it on the new car. You might want to leave EAP and only move FSD since that is the most expensive.

They could also do that with unlimited supercharging. Hint, hint.

Yeah, I know, snowball's chance in you-know-where.

 

[S100D]

Are you sure Tesla promised AP2 would read speed limit signs by end of 2016? Here is the screenshot of what Tesla promised EAP would have by Dec 2016. It does not mention reading speed limit signs: https://www.teslarati.com/wp-content/uploads/2016/10/Tesla-enhanced-autopilot-upgrade.jpg

Tesla is 3 years late but they have finally delivered all the EAP features on that page.

Your .JPG link is broken and cannot be viewed on Teslarati.

 

[diplomat33]

Your .JPG link is broken and cannot be viewed on Teslarati.

 

[Bladerskb]

To the non-Tesla point. I hope you realize Supercruise relies on Lidar mapping, not just GPS map data.
Cadillac to Increase Super Cruise Compatible Highway Network

Uhm no because you can remove the requirement and enable super cruise on local roads.
Nevertheless neither propilot, bmw's DAP and others have that requirement.

Just because safe guards are put in place or features are disabled doesn't mean the product is inferior.
For example, BMW disables auto lane change in local roads but you can enable it yourself. Or the fact that BMW doesn't enable assisted driver view on cars made in 2018 (only cars made after July 2019) even though all the tech is exactly the same. However you can enable it yourself on 2018 made cars. I could go on and on...

Whereas Tesla can drive in unmapped areas like race tracks and dirt roads.
Tesla's Autopilot Continues To Improve, Even Performing Well On Dirt Roads | CleanTechnica

You call failing basically 100 times in a video and trying to kill the driver basically 100 times as 'can drive'?
You Tesla fans are hilarious. So a car being able to drive in a city and be reliable up to 1 million miles per disengagement/accident sucks.
But a car attempting to kill you seconds after you turn it on is showcasing advanced leading tech and clearly leading in the race to remove the driver from the car? What type of warped logic is that?

Amazing!

 

[mongo]

Uhm no because you can remove the requirement and enable super cruise on local roads.
Nevertheless neither propilot, bmw's DAP and others have that requirement.

Just because safe guards are put in place or features are disabled doesn't mean the product is inferior.
For example, BMW disables auto lane change in local roads but you can enable it yourself. Or the fact that BMW doesn't enable assisted driver view on cars made in 2018 (only cars made after July 2019) even though all the tech is exactly the same. However you can enable it yourself on 2018 made cars. I could go on and on...

You call failing basically 100 times in a video and trying to kill the driver basically 100 times as 'can drive'?
You Tesla fans are hilarious. So a car being able to drive in a city and be reliable up to 1 million miles per disengagement/accident sucks.
But a car attempting to kill you seconds after you turn it on is showcasing advanced leading tech and clearly leading in the race to remove the driver from the car? What type of warped logic is that?

Amazing!

Why are you bringing up BWM in response to a post regarding GM Supercruise which you brought up in your response to a post regarding Tesla FSD which you brought up as a response to my post regarding areas that Tesla is ahead of others in terms of units and specific features?

Does BWM have a parking lot summon feature that I am unaware of?

Were any of these items actually incorrect?

• Number of vehicles in the wild, in the public's hands, that are self navigating and moving while empty.
• Number of vehicles in the wild, on the road, self navigating and changing lanes without active driver permission.
• Number of vehicles on the road, owned by the public, with OTA SW updates that enable new self driving features after purchase.
• Number of vehicles produced quarterly with the above features.

I am not aware of these hundreds of near death retakes involved in the recent AP lap around a track (which I unfortunately was not able to re-find a link to), any data to support your claim regarding that?