Summon seems like a silly party trick to me

[alsetym]

Because most of this thread is based on one user's posts, I just thought everyone might want to know that if you go back through
@SandiaGrunt's past posts, not one single one of them is anything aside from scathingly negative. Many have been moved to 'snippiness'.

Regardless of their motivation, it's clear they're on this forum just to spread negativity. Don't feed the trolls.

Fair enough, I'm out.

 

[Rockster]

Or she could get stuck with a damaged car that may or may not be drivable
Of course I don't wish this upon her or anyone but that's the risk.

At 5 mph? I doubt it.

 

[SandiaGrunt]

You know, I can see how I appear trollish. Here I am, saying over and over, "but look at all the evidence, it mostly looks bad."

But... it really does look mostly bad. I'm having a hard time trying to weigh the pros and cons when these are the stories:

https://jalopnik.com/people-using-teslas-new-smart-summon-feature-are-alread-1838593877
https://jalopnik.com/theres-more-and-more-tesla-smart-summon-problem-videos-1838703032
Tesla owners furious after ‘Smart Summon’ feature that automatically picks up drivers causes string of crashes
Smart Summon Feature From Tesla Is Off To A Rough Start
Tesla cars can now drive themselves to their owners with 'Smart Summon' but video shows the feature wreaking havoc
Tesla's new Smart Summon feature is drawing scrutiny from US traffic safety agency as people use it in crowded parking lots

And I have a hard time when I come here and am told "yes, I know it's bad today, but it will improve." This is not acceptable in a product that is used in public parking lots and around vulnerable people.

It's a 4,000 lb robot, and it cannot reliably detect curbs, trash cans, garage walls, grass, reversing vehicles, stop signs, parking space lines, or semi trucks. This is a bad idea, folks.

 

[S4WRXTTCS]

I don't necessarily disagree with any of the OP feeling that Tesla should focus on genuine utility, but in this case Elon tied their hands years ago. Tesla/Elon put smart/enhanced summon in the list of promised features for Enhanced Autopilot, and then when that was dropped they put it on list of promised features for FSD. Even going as far as declaring it already done well before it shipped.

So it's really irrelevant whether it's a party trick or not. People paid for this feature so it needed to be worked on, and released. It was a significant part of the promise for enhanced autopilot.

As to the utility of it? It really comes down to needing to have a bunch of things in place, and smart summon is just one piece. It's going to be a lot more useful when it's been improved, and when it's capable of the reverse. The whole finding a parking space, and parking itself.

In a lot of ways it feels premature, and out of order.

I would have approached it from a freeway/highway perspective first, and then worked my way towards the parking lot. Where the parking lot was the LAST thing I added.

 

[Tam]

You know, I can see how I appear trollish. Here I am, saying over and over, "but look at all the evidence, it mostly looks bad."

All the pieces of evidence so far point toward property damages but not endangering the safety of human life.

In theory, if the car could not recognize the grass and it damaged the tire rim then how can it recognize a 14-month-old toddler?

That does look bad and that's a fair concern but it's hard to convince that killing grass and tire rim is the same as killing a human!

 

[SandiaGrunt]

In theory, if the car could not recognize the grass and it damaged the tire rim then how can it recognize a 14-month-old toddler?

Right! Grass and curbs are common, static elements present in just about all parking lots. If they can't get those right, how could anyone expect them to get toddlers right? Toddlers wear all kinds of different clothes, move in all kinds of unpredictable ways, etc.

I mean, seriously, would anyone here in this forum willingly lay their toddler on the ground in front of a Tesla "summoned" by someone hundreds of feet away who can't even see the car?

That's the ultimate test for acceptability here. Other people are going to accidentally do this. Other people who didn't sign up for this *sugar*, and can't opt out.

 

[Tam]

...In a lot of ways it feels premature, and out of order...

Like someone said before, it's just like the same movie all over again!

When AP2 came out in January 2017, it was really bad.

Several months later it's better but it still veered left or right suddenly when I wanted to go straight!

It's still very bad but people still want it!

When Navigation on Autopilot came out, people were screaming in this forum that it's the worst but other people still used it.

After that, Consumers Report also picked that up and agreed with this forum that "Using the system is like monitoring a kid behind the wheel for the very first time. As any parent knows, it’s far more convenient and less stressful to simply drive yourself."

People just forgot that all these features are beta. I don't expect much from beta so it's a pleasant surprise for me that I can access it prior to its finalization.

 

[diplomat33]

Half a million smart summons already according to Musk.
Elon Musk on Twitter

Yet, we've only had a few reports of incidents. Only a few incidents out of 550,000 smart summons is an excellent reliability rate.

@electronblue I would say 550,000 smart summons is a non-negligible sample to start making a judgment on reliability. Maybe my optimism was warranted after all?

@SandiaGrunt If Smart Summon was so dangerous, don't you think that we would have had a serious incident or injury in 550,000 smart summons?

 

[Daniel in SD]

Half a million smart summons already according to Musk.

That number seems way too high. On the other hand I probably racked up a couple dozen summons doing Tesla jousting with @AlanSubie4Life (which by the way is a hilarious glimpse at our FSD future! I won.).
The question is what is an acceptable number of deaths/injuries/property damage for what is basically an entertainment feature? Unfortunately I think the answer is nearly zero.

 

[SandiaGrunt]

Only a few incidents out of 550,000 smart summons is an excellent reliability rate.

Actually, no.... this is rather bad.

There are 268.8 million cars in the United States.
There are 50,000 parking lot crashes per year in the United States.
That's a 0.02% chance of parking lot accident per car-year for the entire population of cars.

There are ~500,000 Teslas in the United States.
There have been ~10 Smart Summon accidents in just five days.
That's ~730 Smart Summon accidents in a prospective year.
That's 0.1% chance of Smart Summon accident per car-year for Teslas.

Even with just a few days of data, it looks like it could be as much as 5x worse. This is NHTSA is investigating it already.

 

[alsetym]

Half a million smart summons already according to Musk.
Elon Musk on Twitter

Yet, we've only had a few reports of incidents. Only a few incidents out of 550,000 smart summons is an excellent reliability rate.

@electronblue I would say 550,000 smart summons is a non-negligible sample to start making a judgment on reliability. Maybe my optimism was warranted after all?

@SandiaGrunt If Smart Summon was so dangerous, don't you think that we would have had a serious incident or injury in 550,000 smart summons?

Oooh, numbers, I'm back!

Numbers look good.

I'm out.

 

[diplomat33]

Actually, no.... this is rather bad.

There are 268.8 million cars in the United States.
There are 50,000 parking lot crashes per year in the United States.
That's a 0.02% chance of parking lot accident per car-year for the entire population of cars.

There are ~500,000 Teslas in the United States.
There have been ~10 Smart Summon accidents in just five days.
That's ~730 Smart Summon accidents in a prospective year.
That's 0.1% chance of Smart Summon accident per car-year for Teslas.

Even with just a few days of data, it looks like it could be as much as 5x worse. This is NHTSA is investigating it already.

I think your stats might be flawed. 10 smart summon accidents out of 550,000 smart summons is a failure rate of 0.0018%. Put differently, Smart Summon currently has a reliability of 99.998%.

 

[SandiaGrunt]

I think your stats might be flawed. 10 smart summon accidents out of 550,000 smart summons is a failure rate of 0.0018%.

Most people drive through parking lots multiple times per day. Think about that a bit, and the "5x worse" I stated makes a bit more sense, yeah?

 

[diplomat33]

Most people drive through parking lots multiple times per day. Think about that a bit, and the "5x worse" I stated makes a bit more sense, yeah?

No. I think your method is still flawed.

There are ~500,000 Teslas in the United States.
There have been ~10 Smart Summon accidents in just five days.
That's ~730 Smart Summon accidents in a prospective year.
That's 0.1% chance of Smart Summon accident per car-year for Teslas.

You are extrapolating that the number of smart summon accidents will be steady for the whole year. In other words, you are assuming that there will be an average of 10 smart accidents every 5 days for the whole year. I don't think you can make that assumption.

I think my method is better because I am just taking the total number of accidents divided by the total number of smart summons attempts.

 

[boonedocks]

I too, appreciate being part of the journey, and its fun to watch the baby steps and continual improvement and growth. That said, I find zero features of V10 that I will ever use. (well perhaps Joe Mode, but that's it). Can't wait for v.11. (back to the closet to don my fire suit for all the forthcoming Disagrees....)

Joe Mode is absolutely the #1 best feature of V10. #2 will be on 10-16 when my wife’s Spotify account will be renewed for family plan for $2 more a month and I can cancel my $9.99 Slacker account. Yep. 2 TWO useful features out of V10.

For those of us that charge overnight and spend 2.5 - 3 hours a day commuting, sitting in our car to do and watch things that we can do in the comfort of our own home makes 95% of V10 absolutely useless. I have tried summon 3 times and unless a parking lot is empty it has been 100% useless and actually a nuisance to others. To each their own though. (Still a party trick)

 

[PACEMD]

Right! Grass and curbs are common, static elements present in just about all parking lots. If they can't get those right, how could anyone expect them to get toddlers right? Toddlers wear all kinds of different clothes, move in all kinds of unpredictable ways, etc.

I mean, seriously, would anyone here in this forum willingly lay their toddler on the ground in front of a Tesla "summoned" by someone hundreds of feet away who can't even see the car?

That's the ultimate test for acceptability here. Other people are going to accidentally do this. Other people who didn't sign up for this *sugar*, and can't opt out.

Oh my absurd redundant god............

 

[SandiaGrunt]

Hey @diplomat33, first, I want to say thank you for being willing to look at numbers with me. Other people in this thread have accused me of just having negative opinions, but numbers are real.

You are extrapolating that the number of smart summon accidents will be steady for the whole year. I don't think you can make that assumption.

Why not? What's going to change? If anything, I'd expect Smart Summon users to become more complacent, and accident rates to go up.

I think my method is better because I am just taking the total number of accidents divided by the total number of smart summons attempts.

Neither method is better, they are just different. We can find common ground if we agree on the methodology. Would you like to be so... diplomatic?

You're talking about each individual parking lot traversal, each individual use of Summon -- i.e. there's only a 0.0018% chance of an accident for each Summon. I agree with this. Your number is plausible.

I will need to convert my data so that is also in terms of the chance of accident per parking lot traversal.

Americans take 1.1 billion car trips a day.
Each trip likely involves two parking lot traversals, so Americans traverse 2.2 billion parking lots per day,.
That's 8.3 trillion parking lot traversals per year in the United States.
There are 50,000 parking lot accidents per year in the United States, as I've noted above.
That's an extremely low chance of 0.00000622% of an accident per parking lot traversal.

That means Smart Summon is possibly almost 300x worse than human drivers, in terms of the probability of an accident per parking lot traversal. Even if only half of car trips involve a parking lot traversal, Smart Summon is still 150x worse than human drivers.

The fact that Smart Summon keeps failing in ways that humans usually don't -- hitting curbs, driving on grass, etc. -- is telling.

The fact that so many people videotaped their first attempts at using Smart Summon, and crashed their cars, is telling. If Smart Summon were really very very good relative to humans, the first attempt would have had an almost vanishingly small chance of an accident.

I invite you to check my math, and I will check yours. Maybe we will both learn something.

 

[Fossil Fool]

So Smart Summon (beta) isn't perfect. Name the car currently being made by another manufacturer that comes close to this feature.

 

[PACEMD]

We know Smart Summon is mostly safe for toddlers. Tesla has been testing this for over a year. The Fremont factory, in fact, has a toddler test lot. It's part of their employee daycare program. Every day they take a hundred or so toddlers, aged 10 to 30 months, and let them loose in the parking lot, to test against Smart Summon. So far a reasonable percentage have survived...........but it's still Beta, not perfect.

 

[SandiaGrunt]

So far a reasonable percentage have survived...........

I know you're trying to be funny, and your joke sort of lands. I guess you're not a parent, though, because it's hard to listen to a joke like yours and not think of my baby girl's bright blue eyes, the way she runs around the living room going brrt! brrt! brrt!, and the visceral horror I would feel as someone's 4,000 lb zombie car turns a corner and catches her under its fender -- you know, where there is no sensor coverage -- and then continues to pin and kill her right before my eyes. I would strain and kick at the car itself. I would scream and shout and try to drag her out of from under it, but it would not stop. The owner, 400 feet away, is holding his fat thumb on that smartphone button, while his buddy shoots a terrible portrait-mode video they can't wait to post to YouTube.

This is serious ****, my friend. Please have some decency.