Instead of calling it FSD should we call 10% self driving?

[outie]

Tesla’s FSD order page says:
Summon: your parked car will come find you anywhere in a parking lot. Really.

I don’t have EAP or FSD. But when I had the EAP trial in December, I thought Summon simply moved the car forward or backward. But this description sounds like the car will navigate itself to you weaving through the parking lot. It Is listed as a currently available feature, not as “Coming later this year” feature. Does this thing really work?

Smart/enhanced Summon is something EM has been hyping for few months in his tweets and its release is imminent (March 15 mega update?). We’ll see.

 

[MikeS1]

Guys u think you need to fully understand that what we have for EAP is not at all what we are going to see for FSD.

One, EAP is designed for freeways ONLY so any logic or learning on side roads or cities are not even a part of the processing and that's why it doesn't detect a lot of things or is stupid.

Two the upgrade to the V3 hardware is going to vastly improve what the car can see around it and at a much improved resolution. That plus the huge amount of new logic for FSD will make the experience 100% different.

So for the love of god stop comparing EAP to FSD, it does not make any sense what so ever. It's like comparing an old Nokia phone from the 90s to a smart phone today.

 

[robski]

a couple of FSD features needed out of a thousand.

2/1000 = 10%

 

[more863-also]

Guys u think you need to fully understand that what we have for EAP is not at all what we are going to see for FSD.

One, EAP is designed for freeways ONLY so any logic or learning on side roads or cities are not even a part of the processing and that's why it doesn't detect a lot of things or is stupid.

Two the upgrade to the V3 hardware is going to vastly improve what the car can see around it and at a much improved resolution. That plus the huge amount of new logic for FSD will make the experience 100% different.

So for the love of god stop comparing EAP to FSD, it does not make any sense what so ever. It's like comparing an old Nokia phone from the 90s to a smart phone today.

Yeah, not really though. You have no idea what FSD will be capable of - whereas I'm very familiar with EAP and its limitations, like when it almost slams me into a wall when I approach a turn where dirt and slush completely obscure the fog line. I completely agree that HW3 will be better and faster, probably making microadjustments more smoothly and making phantom braking a thing of the past. But I don't see any reason to believe it will be able to perfectly navigate in areas without lane lines, or cross a busy highway to make a left, simply because we already know what the cameras see and just upgrading speed isn't going to change it. Level 3 FSD in general city driving is orders of magnitude harder than Level 2 highway driving and all I've seen from Tesla are gradual improvements to Level 2.

 

[bhzmark]

They should probably call it Not Full Self Driving.

Not A Flamethrower — The Boring Company

 

[GoSharks]

Guys u think you need to fully understand that what we have for EAP is not at all what we are going to see for FSD.

One, EAP is designed for freeways ONLY so any logic or learning on side roads or cities are not even a part of the processing and that's why it doesn't detect a lot of things or is stupid.

Two the upgrade to the V3 hardware is going to vastly improve what the car can see around it and at a much improved resolution. That plus the huge amount of new logic for FSD will make the experience 100% different.

So for the love of god stop comparing EAP to FSD, it does not make any sense what so ever. It's like comparing an old Nokia phone from the 90s to a smart phone today.

This sounds like baseless and hopeful speculation.

 

[kavyboy]

If I had a time machine I would go back to 2016 on this forum

If you do go back, tell everyone not to make any bets with @calisnow and to not be so hard on @Bladerskb.

 

[Leeclanual]

having to keep one of your paws on the steering wheel is not self driving... full autonomous driving isn't going to be allowed or capable for 10+ years.

 

[SigNC]

Way more excited about the fact I can warm my battery on the way to a supercharger than anything AP related.

 

[wwu123]

Full Self-Driving Capability
All new Tesla cars have the hardware needed in the future for full self-driving in almost all circumstances. The system is designed to be able to conduct short and long distance trips with no action required by the person in the driver’s seat.

It's interesting that page has a qualifier (boldface mine) that seems to offer them an out. I think Tesla might fall back on that and say it's achieving "full self-driving" if it's able to take you on five minutes of your journey, hit some situation that needs your manual intervention for 30 seconds, then does "full self-driving for another 10 minutes", you spot a situation that you have to intervene for 5 seconds, etc. "Well, it fully self-drove you for 98% of the duration of your trip, so that's full self-driving in almost all circumstances".

Whereas most of us would disagree and define FSD as when both no attention and no action is required for the entire contiguous journey.

 

[kavyboy]

AP1 really was good enough. Tesla went off the rails with FSD and its promises, in my opinion. I hated to see it happen. Make great EVs and EV support systems and let somebody else, the people have been working on that problem domain intensely for years, figure out the self driving stuff. I'm happy to see Tesla pulling back on both the promises and the expectations regarding FSD to something reasonable.

 

[OPRCE]

we already know what the cameras see

Not entirely, as currently both resolution and frame-rate is clipped on AFAIK all AP-relevant cameras to avoid overloading the APE. This limitation will be lifted with HW3.

Further, the expanded NNs designed for HW3 are reputed to provide a significant boost in recognition performance.

So better inputs and more sophisticated processing, done faster, will surely bring a decent leap forward in the accuracy and fluidity of those AP features we already have, if not the new FSD ones immediately.

It's interesting that page has a qualifier (boldface mine) that seems to offer them an out. I think Tesla might fall back on that and say it's achieving "full self-driving" if it's able to take you on five minutes of your journey, hit some situation that needs your manual intervention for 30 seconds, then does "full self-driving for another 10 minutes", you spot a situation that you have to intervene for 5 seconds, etc. "Well, it fully self-drove you for 98% of the duration of your trip, so that's full self-driving in almost all circumstances".

Whereas most of us would disagree and define FSD as when both no attention and no action is required for the entire contiguous journey.

Yes, it's an open admission they never expect to exceed a nagless L3 with the current sensor suite & HW3.

 

[notAnExpert]

It's interesting that page has a qualifier (boldface mine) that seems to offer them an out. I think Tesla might fall back on that and say it's achieving "full self-driving" if it's able to take you on five minutes of your journey, hit some situation that needs your manual intervention for 30 seconds, then does "full self-driving for another 10 minutes", you spot a situation that you have to intervene for 5 seconds, etc. "Well, it fully self-drove you for 98% of the duration of your trip, so that's full self-driving in almost all circumstances".

Whereas most of us would disagree and define FSD as when both no attention and no action is required for the entire contiguous journey.

That "almost all circumstances" line seems to have existed in their website description since 2016.

 

[9erDog]

having to keep one of your paws on the steering wheel is not self driving... full autonomous driving isn't going to be allowed or capable for 10+ years.

I’d suggest you stop letting your pets drive your car. I found my drives much less stressful once I did that.

 

[MikeS1]

This sounds like baseless and hopeful speculation.

It's not, when your car slams on its brakes because another car crosses the road into in the distance is a clear example that there is a *sugar* ton of logic in FSD that does not at all exist in EAP.

You can claim my assumptions are baseless but the evidence does not work at all in your favor. That evidence being there is zero additional logic for local streets that are currently in EAP.

 

[diplomat33]

Seems more accurate. Far from full self driving. Or perhaps call it a couple of FSD features needed out of a thousand.

No. Based on the description on the website, Tesla aims FSD to be a system that can fully drive you on highways and on city streets, with driver supervision at first and then later with no driver supervision. That's not 10% self-driving. FSD will be closer to 80% or 90% self-driving. EAP is 10% self-driving now.

 

[GoSharks]

You can claim my assumptions are baseless but the evidence does not work at all in your favor. That evidence being there is zero additional logic for local streets that are currently in EAP.

What evidence? Are you telling me you know the code inside EAP/HW2.5, and have insider knowledge?

You presented nothing to back up your claims in post #22

 

[MikeS1]

What evidence? Are you telling me you know the code inside EAP/HW2.5, and have insider knowledge?

You presented nothing to back up your claims in post #22

That's because you make assumptions that are not logical in any sense.
It's absolutely clear there is zero FSD logic in EAP. Thats proven by your lack of faith in the autonomy because you are drawing an illogical comparison to the functionality of FSD.

I don't have insider knowledge but I do have a background in electrical / computer engineering. And the most logically answer is you do not see any of the code or machine learning for FSD.

 

[diamond.g]

That's because you make assumptions that are not logical in any sense.
It's absolutely clear there is zero FSD logic in EAP. Thats proven by your lack of faith in the autonomy because you are drawing an illogical comparison to the functionality of FSD.

I don't have insider knowledge but I do have a background in electrical / computer engineering. And the most logically answer is you do not see any of the code or machine learning for FSD.

Isn't that more because NoA doesn't work in the city?

 

[SunCharged]

Yeah, not really though. You have no idea what FSD will be capable of - whereas I'm very familiar with EAP and its limitations, like when it almost slams me into a wall when I approach a turn where dirt and slush completely obscure the fog line.

This just happened to me yesterday. The snow and slush build up against the concrete highway divider. AP drives you into the slush which is outside of the lane. Then the car becomes extremely unstable and yes, tries to put you into the wall!