Safety Score

[dckiwi]

I'm pretty sure the code has been written already to send to all cars who have a 100 score. Would be a lot more work for them to segment that up.

Code? More like someone pulls a report, sorts by score, exports as csv, and the VIN numbers are copy/pasted into some config file.

I have nothing to back this up. Just too much experience in IT.

 

[jebinc]

Code? More like someone pulls a report, sorts by score, exports as csv, and the VIN numbers are copy/pasted into some config file.

I have nothing to back this up. Just too much experience in IT.

@dckiwi Delimited or undelimited csv?

 

[EVNow]

Code? More like someone pulls a report, sorts by score, exports as csv, and the VIN numbers are copy/pasted into some config file.

I have nothing to back this up. Just too much experience in IT.

More like exported to AWS S3 buckets and used by their release application. They are quite likely to have a sophisticated application for releases since it is mission critical.

 

[dburkland]

If there are too many 100s they should rank based on who bought FSD first. Elon then can make good on his promise of early adopters getting early access

 

[S4WRXTTCS]

Been stuck at 98% since last week. I drive over 100miles a day. Last few days score has been 99-100, still no change in the overall average. Unsafe following is what’s haunting me. I had to disengage AP at highway speed twice last Tuesday, this is my issue. Oh well.

The way to fix unsafe following (assuming it happened the same day) is to drive on the freeway WITHOUT AP. The reason for this is AP miles won't be counted towards following distance miles so it just gets stuck. I drove completely manually for 60 miles and it greatly reduced the unsafe following amount.

 

[fiveten]

Ok, I got two cars sitting at home at 100s for their scores. One car has around 600 miles on it since this all started, the other had 1556 miles in it. Lots of “cheating” (driving in round about, doing safe braking over and over again, and even worse things ) to get here…we will see what midnight tonight pacific times brings…

 

[g1000x]

I’m interested to see what happens. I’m expecting many “I have a 100, but didn’t get the beta” posts tomorrow.

 

[minis2003]

Just wondering who might have the lowest safety score so far and how many miles they drove during that time??????

does anyone think it would be possible to a score of "0"? just wondering

 

[MarkS22]

To add to the speculation, I believe they were monitoring the number of vehicles with a 100 SafetyScore. About six hours ago, the Safety Score wouldn't load in the app ("Page failed to load. Try again.") for 45 minutes. It's very possible they hit, for example, 1500 beta testers and took a snapshot at that time. Perhaps the surprise is that "everyone who had a 100 as of 2pm PST has been invited."

I don't think it's unreasonable for them to have time to pull the records during the workday before launching.

 

[Nakk]

I think Tesla is going to roll out FSD Beta to a certain amount of cars and then pause rollout for a certain amount of time. That just seems the most likely to me.

 

[Ramphex]

This is the "end result" ? Can you get anymore hyperbolic ...

To date it is. Can’t say end result is anything else because then it would just be speculations. Also think of all the people who dropped $10k years ago and have since sold or crashed their cars and NEVER got a chance to even play this try out game.

You can dislike my post all you want, all in all I think that I wasted 10k on this and that’s my personal opinion, whether you like it or not

 

[fiveten]

To add to the speculation, I believe they were monitoring the number of vehicles with a 100 SafetyScore. About six hours ago, the Safety Score wouldn't load in the app ("Page failed to load. Try again.") for 45 minutes. It's very possible they hit, for example, 1500 beta testers and took a snapshot at that time. Perhaps the surprise is that "everyone who had a 100 as of 2pm PST has been invited."

I don't think it's unreasonable for them to have time to pull the records during the workday before launching.

There’s been other times during these past two weeks where scores were inaccessible, but today I think was the longest time they were/

 

[S4WRXTTCS]

I don't understand these kinds of posts - specifically, posts that come across as if FSD beta is an actual public release/final product and posts that criticize Tesla's utilization of safety scores for determining eligibility for the beta group.

I completely understand the issues with the safety score app and I understand why people are frustrated. However, at the end of the day, Tesla is merely using the safety score as a means to expand its very, very limited pool of beta testers to a slightly less limited pool of beta testers. It's not using safety scores to weed out irresponsible drivers. It's not using safety scores to rip off people who paid for FSD. It is simply a means of determining who gets to try out this early, early stage software in furtherance of improving the software prior to widespread distribution. And, most importantly (at least from Tesla's perspective), the safety app gives Tesla cover in case a serious issue (fatality) arises with the beta FSD software. After all, this is a huge endeavor and all eyes (regulators, other car manufacturers, potential customers, etc.) are going to be squarely on Tesla for leading the pack into this virgin land.

Tesla just as easily could have picked 1000 people at random for its beta group or not provided its FSD customers with any means of becoming a beta tester. Personally, despite the inherent flaws with the safety score app, I like the fact that I have had some degree of control over the process. I like knowing that there is objective criteria that, if I meet, I can become part of an exclusive group with respect to a cutting-edge movement. And, in the coming months, everyone who paid for FSD will have it.

We have the illusion of control over it.

This makes it substantially worse because if you didn't get it you can't just blame Tesla, but you have to blame yourself too. If you didn't stop for that yellow or if would have just driven your other vehicle for 2 weeks while the Tesla sat in the garage with 1 mile of perfect driving.

As you see in this thread people have resorted to adjusting the way they handle situations, and doing weird things to try to undo marks against them. On a few occasions I've helped people with advice on how to reduce the marks against them. The person who posted the spreadsheet helped us all out a ton, but the entire existence of it along with the usage shows how obsessed people got over it.

This Safety Score Beta is a disaster even without how its tied into the FSD Beta. They just tied it to convince thousands of people to have their driving judged. If it wasn't tied to it the beta score would simply be entertaining.

The entire purpose I had of buying FSD back in 2018 was to experience it as it grew. I didn't expect autonomous driving, but I expected to at least be able to experience some elements of it. The hardware necessary for autonomous driving simply isn't there so the only question is how well L2 level human+car interaction will work as the SW gets better.

There is also a bit of FOMO out going on. Where there is the very real possibility that some event will happen with the first 1K people that will stop the rollout for the rest of us.

At the end of the day I would have preferred a regional specific rollout as then I could simply see when my region was scheduled, and not worry too much about it.

I'm certainly not the only one who wasted countess hours obsessing over it. I spent the last week driving my car in a way I've never driven a car before, and I've been driving for 30+ years. Like I said before I had to pretend it was my RV.

If I had a spouse I'm pretty sure she would have turned it off, and told me to stop it.

 

[tivoboy]

I'm pretty sure the code has been written already to send to all cars who have a 100 score. Would be a lot more work for them to segment that up.

I am fairly certain that there is not automation for this execution.

 

[MilesAR]

So how many days are they going to use to factor in the score?
Past 7 days?

 

[Sporty]

I don't understand these kinds of posts - specifically, posts that come across as if FSD beta is an actual public release/final product and posts that criticize Tesla's utilization of safety scores for determining eligibility for the beta group.

I completely understand the issues with the safety score app and I understand why people are frustrated. However, at the end of the day, Tesla is merely using the safety score as a means to expand its very, very limited pool of beta testers to a slightly less limited pool of beta testers. It's not using safety scores to weed out irresponsible drivers. It's not using safety scores to rip off people who paid for FSD. It is simply a means of determining who gets to try out this early, early stage software in furtherance of improving the software prior to widespread distribution. And, most importantly (at least from Tesla's perspective), the safety app gives Tesla cover in case a serious issue (fatality) arises with the beta FSD software. After all, this is a huge endeavor and all eyes (regulators, other car manufacturers, potential customers, etc.) are going to be squarely on Tesla for leading the pack into this virgin land.

Tesla just as easily could have picked 1000 people at random for its beta group or not provided its FSD customers with any means of becoming a beta tester. Personally, despite the inherent flaws with the safety score app, I like the fact that I have had some degree of control over the process. I like knowing that there is objective criteria that, if I meet, I can become part of an exclusive group with respect to a cutting-edge movement. And, in the coming months, everyone who paid for FSD will have it.

They weren’t clear that it was only going to scores of 100 from the start. Knowing that would have helped. It would have set expectations up front and I could have decided it wasn’t (or was) worth it days ago.

 

[Sporty]

To be fair, it could be the end result, at least for a very long time. But we all hope it is not. Hopefully everyone will be extremely careful.

I hope I’m wrong but I suspect it will be weeks, or months, before anyone else gets it. How can you decided ‘in a few days’ that things are going well so move onto the next 1k? That doesn’t add up for me.

 

[JosiahLuscher]

Hello all, just today i got flagged with a FCW of 56.9avg on my daily commute this morning. I was never even close to anyone on my drive, and my warning never went off?? I was at 99 and this bought me down to 98.. does anyone know of this happening and if there is any way of having it investigated????

Thank you Brad

So, @Huksie13 you can set the forward collision warning sensitivity in the car's menu; BUT the safety score always reacts to the medium sensitivity level. So yes, your safety score can get hit by an FCW that you don't know about if your sensitivity, or whatever it's called, is set lower than medium.

Alternatively if your FCW is set to high sensitivity the car could beep at you, but it might not be 'enough' of an FCW to trigger a hit to your safety score.

 

[Knightshade]

I hope I’m wrong but I suspect it will be weeks, or months, before anyone else gets it. How can you decided ‘in a few days’ that things are going well so move onto the next 1k? That doesn’t add up for me.

Why does it not add up?

Tesla can get performance data from the fleet in just about real time.

If in a few days they see things like disengagement and intervention rates, or even attention-monitoring warnings, spiking, they might slow down to find out why. Maybe even boot some people behaving poorly.

If everything still looks great and the new testers don't appear to be having any more problems than existing ones, they might decide it's fine to add another bunch and review again after a few more days.

 

[gearchruncher]

There's a deep irony to this safety score, and FSD beta only going to those with 100's.

While it looks like it's good cover for Tesla to say "we only sent it to the very safest drivers" - that is WAY worse if there is any kind of accident with it. It means that even given the very safest drivers, it still leads to accidents, so Tesla needs even more development and testing before trying again. As is always said, FULL SELF DRIVING CAPABILITY is an L2 system- it is fundamentally the driver that is responsible, as they are the only safety critical element in the system.

If they sent it to random drivers, they could say "well, yeah, some drivers kind of suck, but that's not average, and one accident per 1K people is just statistics" But in this case, they are claiming only the very best have it, way better than average. If they see any issues with this group, they flat out can't expand it downwards, because those issues will only get worse.

This hand picking of people that meet some other metric means that this is not a valid test of reliability nor is it even that useful for many kinds of training. This is absolutely not the way any kind of statistician would run a test to start to see if it can be widely deployed. That doesn't bode well for it going wider soon. I think some of you have it right that any accident with this group means the whole thing will need to be pulled. If the very best can't handle it, who can?

And, most importantly (at least from Tesla's perspective), the safety app gives Tesla cover in case a serious issue (fatality) arises with the beta FSD software. After all, this is a huge endeavor and all eyes (regulators, other car manufacturers, potential customers, etc.) are going to be squarely on Tesla for leading the pack into this virgin land.

There will be accidents when FSD is in use, eventually. The question will be the severity and whether FSD played any significant role (since the driver of course will still be driving, it could be driver error).

I expect the rollout to continue or fsd beta testing stopped (i.e. revoked / withdrawn) in case of a catastrophic event.

This Safety Score Beta is a disaster even without how its tied into the FSD Beta. They just tied it to convince thousands of people to have their driving judged. If it wasn't tied to it the beta score would simply be entertaining.

There is also a bit of FOMO out going on. Where there is the very real possibility that some event will happen with the first 1K people that will stop the rollout for the rest of us.