Sure, my words were a bit harsh. They are on a path. What FSD is today though (IMO) is not good for the general public. It's going to take a paradigm shift to achieve something that can be safely used by the average driver.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Monthly subscription to FSD - what version do you get
- Thread starter Dr. D Flier
- Start date
Sure, my words were a bit harsh. They are on a path. What FSD is today though (IMO) is not good for the general public. It's going to take a paradigm shift to achieve something that can be safely used by the average driver.
Again, just my opinion... but I believe that most of today's FSD users are early adopters and technology enthusiasts.
FSD has plenty of dangerous edge cases that it can't handle... and as currently designed this will always be the case. If it were widely deployed, enough people would use it dangerously that it would become a problem. Dangerously in that they would just "flip it on" without understanding... get comfortable because "it's doing good", and then get into trouble and not be prepared to rescue FSD in the event of a split-second emergency.
It's not traditional software either, where they can just keep tackling the edge cases until they're all solved (e.g. IF edge_case12 THEN do_this()). Instead, Tesla will continue to churn out iterations that do better in some scenarios but worse in others... and that will *NEVER* change (just like auto wipers). At least not on their current approach to autonomous driving.
It's possible that Tesla (or someone else) will achieve the "holy grail" of AGI (artificial general intelligence) and perhaps at that time, they'll be able to create a model that can drive better than a human... but as it currently stands, that's not happening.
I don't mean this as a knock on Elon... I think Tesla's cars are pretty great... but he tends to hype up what they're doing. Even cherry-picking data and painting things in a positive light that perhaps don't deserve it. I believe this is what he did when he's stated things like "FSD is X times safer than a human driver".
Look at this article as an example:
Tesla publishes astonishing FSD safety statistics
Tesla has published a new report detailing just how safe Full Self Driving (FSD) is making the driving experience of its customers.
www.teslarati.com
It compares "FSD equipped vehicles" (was it in use, or just equipped?) to the average driver and shows that it's about 5 times less likely to have an accident... but then later in the article it shows that Tesla drivers in general tend to have less accidents, so FSD is only about twice as safe as Tesla drivers in general... however FSD requires the driver to maintain a safety score just to continue using it, so they're stacking the numbers in their favor - using only the safest drivers who are forced to exhibit behaviors that are known to reduce the probability of an accident.
Again, I don't mean to sound super negative... I think it's amazing that anyone's been able to make a car do as much as FSD can do without a driver - but that doesn't make it as good as an experienced human driver, and doesn't make it totally safe. It's dangerous.
So you don't believe that driver monitoring will suffice to keep the masses out of trouble? If the monitoring is annoying enough to irresponsible drivers, then they'll either work to defeat the monitoring or they'll stop using the system. Then it seems to boil down to it being dangerous only for the ones who actively work to abuse the system, and things revert to a cat and mouse game between manufacturers and irresponsible owners.
I'm not referring specifically to people who are not paying attention to the road.
Even an attentive driver that's holding the steering wheel won't always be able to correct an FSD mistake. I believe most drivers presented with a dangerous situation and given only a split-second to act will not perform well enough to avoid the dangerous situation.
Most people, at most tasks, are mediocre at best.
Will that driver be able to correct enough FSD mistakes to make the difference? If the accident and death rate is cut in half, I'd count us lucky that the problem to be addressed is in a piece of software that can be improved, versus a problem in people that cannot.
Fortunately, we won't have to worry about this because Tesla won't even pursue mass adoption until they're convinced that FSD's mistake rate is low enough to cut accident and death rates in half. Until then, it'll remain a toy for the wealthy, and the wealthy folks can continue to play the role of lab rat.
seenhear
Member
I love that they let me disable TACC like the MS/X have had in the refresh.
I hated when it would revert to TACC only.
The topic of software versions is super confusing to me. So there are firmware versions that do not include any version of FSDb (old or current) at all? And installing said versions would leave you with no version of FSDb even if you previously had one (effectively deleting it)? That seems odd, since FSDb should be largely independent of the surrounding software. You wouldn't expect to lose Notepad just because you downloaded a Windows update for a security patch.
All versions of the software have included FSD for a while now. Just install whatever Tesla offers you. It'll be the latest and greatest.
The only people who refuse to install newer versions are very... particular in their requirements. They'll consider one or more of the latest and greatest changes to be rubbish and refuse to install any new software. You can do that, but you'll lose out on all changes going forward until you again choose to install the latest and greatest.
And yes, the infotainment software and FSD appear to be joined at the hip. I've yet to hear why that is. Not that it matters; I don't want to have to juggle two versions of software on my car. Nor do I want to build the software from source code. Nor do I want to spend time configuring tuning parameters. Or any of that other rubbish that makes computers such a pain to use these days. That's hobbyist stuff, not consumer.
seenhear
Member
What hardware are you on? M3? MS? Year? HW2.5? HW3.0? Which MCU do you have? Cameras? All of these play a role in what firmware you will get.
I have an Oct.2017 MS75. It was built with MCU1, HW2.5 (NVidia AP computer, updated camera suite). It wasn't even possible for me to subscribe to FSD. The app indicated I needed a HW update first.
I did end up getting that update, (story posted here) both the AP/FSD computer, and the MCU were upgraded. The car then downloaded a much newer firmware than I had been running. I was able to "subscribe" to FSD, but initially nothing happened when I subscribed. There was no switch for beta. I already had EAP, which is identical to FSD without beta. I forced a reboot a couple times, no luck. Then one day when I arrived home from work and pulled into my garage, the car self-initiated a reboot as soon as I put it into park. After the reboot FSD beta switch showed up in the AP menu, and all was golden. (Well, as golden as FSD v11 can be, which is more Pyrite than gold, IMO, LOL).
seenhear
Member
Except that FSD v12 is only available to those with an older firmware. It's not available on 2024 firmware yet, AFAIK. So if you've got or want FSD v12, that's a good reason to not upgrade the firmware to 2024.
seenhear
Member
Did you ever have Enhanced Autopilot? EAP included "Navigate on Auto Pilot" which was basically FSD on freeways only. It will navigate fwy interchanges and exits, even handle merging back into traffic after changing from one fwy to another. It will stop functioning at the end of a fwy exit ramp that goes to surface streets.
So if this is the "drugery" aspect of driving you enjoy FSD for, then you aren't really utilizing what FSD has to offer. NoAP has been available to EAP and FSD owners since I think mid/late 2018, long before FSD beta was available to anyone outside Tesla. What FSD-beta offers that you can't get with any other software, is navigating on surface streets. But you say you don't use it for that. I don't blame you as while entertaining, it is pretty bad at it. With v11 I have to take over around 1x/trip (where a "trip" is anywhere from 0.5mi to 9mi of surface street driving).
If all you're using FSD-beta for is the long boring sections on expressways and freeways, you're not really using FSD's unique features as that can be handled by EAP (navigating on freeways) and base AP (just lane/speed keeping on long boring stretches of any road.
Personally, I would characterize FSD v11 as just barely more than a parlor trick. It's occasionally beneficial, but more often I engage it for the entertainment (I like technology and like to see what FSD-beta will do in various scenarios). I will not be renewing my current subscription at the end of this month. I will probably subscribe again when v12 is available to me, just to check it out.
I wish the FSD team well. I hope they achieve their goals. But full autonomy is ages away IMO. Even true / safe level 3 autonomy is ages away. v11 is just not safe by any means right now.
seenhear
Member
seenhear
Member
Here's the thing with FSD - are people going to engage it "so that they are safer" or are they going to engage it in order to offload some of the stress of driving. These are two very different cases, IMO.
Before I had FSD, I had EAP with nav on AP ability. I used this all the time for freeway driving, and for longer drives found I was much less fatigued at the end of the trip. I now have FSD-beta. For freeway driving, I'd say it's the same story. I'm less fatigued at the end of a trip. This is because it's capability is essentially unchanged from EAP+NoAP.
However, when using FSD in other scenarios (navigating through residential, business, etc. surface streets) I'm actually MORE fatigued by the end. I'm MORE engaged in a surface street trip with FSD-beta engaged, than I am just driving myself. HUH? That sounds wrong. But I think it's true. When I'm driving, I'm paying attention to everything around me and intuitively piloting the car through every scenario. When I'm playing back-up to FSD-beta, I have to do double duty. My brain is thinking of how *I* would do something, then observing how FSD does it (or doesn't) and then having to decide if I need to take over. Even when it's doing well in an "easy" zone, I'm still constantly on the lookout for it to do something stupid (which it occasionally does do). When I am driving, I'm not constantly afraid that I am going to swerve unexpectedly into another car. The OTHER car might do that to me, this is just a fact of life when driving, but *I* am confident in MY abilities and intuitive decisions.
Driving surface streets with FSD-beta on is a lot like driving with a 15 year old behind the wheel, who's still learning the basics and developing their reaction times, spatial awareness, and ability to interpret complex dynamic situations instantaneously. If you've ever taught a teen to drive you know. LOL.
So, back to my first statement:
...are people going to engage it "so that they are safer" or are they going to engage it in order to offload some of the stress of driving.
If you purposefully engage it to improve your safety, and are willing to be vigilant and put the extra work in, then it will likely enhance your safety. It's like another pair of eyes monitoring the situation at all times. But if you are turning it on so you can relax a bit and trust it to mostly get you there with minimal effort from you, (as NoAP does today on freeways) then you are actually going to be LESS safe, and indeed with current v11 FSD, will certainly get in an accident.
IronCoffee_Max
Member
Yes, I own EAP (bought it when I got my car - it was offered in September 2020. That's one reason why FSD is only $99 for me. And no, NoAP is nothing like FSD for me. NoAP is like 5 year old tech at this point. Compared to real FSD it's a huge step back. On a recent road trip of 4000 miles I switched to NoAP half way and it was constantly phantom braking, speed limit issues, and so many old fashioned issues. Went back to FSD and except for the nag, it was a huge difference.
Actually, FSD on city streets is actually pretty dang good. It's not my primary use case though, so FSD for city streets doesn't drive my subscription/purchase decisions.
Incorrect. FSD-beta handles phantom braking (ie. none), curves around tight highways, and lane switching so much better than EAP does. It also handles truckers, speed limits, maneuvers, and exits so much better. One of my biggest issues with EAP is how exits/lane changes are slow and start too many miles ahead, how it's not confident in anything. FSD beta completely blows EAP NoAP out of the water…except for the stupid nag.
To be clear, I've got about 20k of road trip miles on EAP and 20k on various FSD betas.
Level 3 autonomy has little to do with capability. It's all about the legal responsibility the government and mfg take for autonomy. Of course that has something to do with the capability because as Tesla gets more comfortable with the capability of it's software, the risk goes down and it might make a decision to take responsibility…but from a purely technical POV, the levels of autonomy are not related to capability.
BTW - I have a background in AI, ML, reinforcement learning, teleoperation, etc… so FSD development is something I follow quite religiously
Also, I am currently not subscribed to FSD. I was for a bit a month or two ago - and it was quite nice, but right now I'm not doing any significant driving so it's not worth the extra cost.
Last edited:
IronCoffee_Max
Member
…………
I 100% use FSD Beta specifically to offload stress of driving. I am vigilant to watch what's going on. But in 90% of my road trips from California to Tennessee and Texas, the roads are near empty and/or very light. 20k of driving on FSD beta (various forms). It's far far safer than NoAP or driving on your own. I can unequivocally say that has been my experience.
seenhear
Member
My experience is that it does not offer me significantly more than what EAP/NoAP does for freeway driving / road trips. Yes, it offloads the stress of driving in the freeway use case, but not noticeably more than NoAP for me. While FSD might be slightly better in some aspects, both are sufficient for freeway driving.
Therefore the only thing that is significantly different with FSD, is the surface street navigation, and that is, IMO, higher stress than driving it myself. It's amusing sometimes to use it, but FSD on surface streets almost never reduces the stress of driving for me, and usually increases it.
So all in all, for me, FSD is not worth the $99/mo over EAP+NoAP. There's still no summon, self park still sucks, and everything else is close enough to not be worth the money. But I'm happy to have the option to subscribe to it on occasion. Subscribing for a few months every year is WAY less expensive than buying it outright. So that's cool. If / when v12 is available for my legacy HW3 car, I would definitely try it out. I suspect that it might never come for cars like mine though, unless they offer/require an interior camera upgrade for monitoring awareness. Ironically, if it were truly "self-driving" (level 4 or better) the camera would be unnecessary, LOL.
Similar threads
