Time to get real about FSD V11
Vložit
- čas přidán 6. 03. 2023
- Consider supporting this channel on Patreon: / aidrivr
Filming Equipment:
GoPro Hero 10 Black (Visualizations): amzn.to/3kGze2I
GoPro Hero 11 Black (Interior): amzn.to/3H7Xsuq
GoPro Enduro Batteries: amzn.to/3wrzhSK
GoPro Dual Battery Charger: amzn.to/3XBxnLi
Fat Gecko Triple Mount (Exterior): amzn.to/3R8J0GQ
Fat Gecko Double Suction Mount (Interior): amzn.to/3HyW0Tj
Sony A6400 w/ 18-135mm Lens: amzn.to/3RauluT
Sigma 30mm F1.4 Lens: amzn.to/3HvOU1C
Insta360 One RS: www.insta360.com/sal/one_rs?i...
Insta360 X3 (free selfie stick!): www.insta360.com/sal/x3?insrc...
Insta360 X2 (free selfie stick!): www.insta360.com/sal/one_x2?i...
Editing Equipment:
16” MacBook Pro M1: amzn.to/3j71X04
Magic Trackpad: amzn.to/403P8UV
Roost V3 Laptop Stand: amzn.to/3JgqBX4
Phillips Fidelio X2HR: amzn.to/3wvuVd7
Dragonfly Cobalt: amzn.to/3kHLPCR
Shure SM7B Microphone: amzn.to/3wxz9kK
Gator Frameworks Mic Arm: amzn.to/3XGGE4E
Elgato Wave XLR Preamp: amzn.to/3XEZpW2
I earn from purchases you make via Amazon Associates & Insta360 affiliate links - thank you!
Intro music by Scott Buckley - www.scottbuckley.com.au - Věda a technologie
I may have over-edited this video, it turned out a bit more dramatic than I expected…
EDIT: this video didn't exactly age well 😅
I don’t think so, I think it was the perfect amalgamation of edit+story
@@tejassetlur806 agree
I would have downvoted this comment if I didn't think the algorithm would misunderstand the implied non negative feedback. Oh wait, perhaps I should upvote it. Hang on...
all good, i found this was an excellent video, got the Y without sensors for a week now, i love the car, dang :)
I like the editing. The music is the only dama that I can discern, but that is fine as well.
If V11 is effectively a software architecture change, then I think it would be likely to underdeliver in the short term, or even be worse than the older architecture. However over the long term the newer, cleaner, single stack architecture would simplify development and allow the big improvements that we all want. Fingers crossed
🤞
That is usally the case with Tesla. If they find out that it isn't working, they don't hesitate to change it up unlike most other companies that DOUBLE down on the tech due to them having invested too much into it. That is one of the good things about Tesla. They are NOT AFRAID TO Go in a different direction IF IT DOES NOT WORK.
It is NOT a software architecture change. People keep saying this and it is flat out wrong. What they mentioned is a *Neural Network* architecture change. For example, going from a Convolutional Neural Network to a Vision Transformer Neural Network is an architecture change. But the underlying software architecture doesnt have to change at all.
It’s not flat out wrong, it might be a simplification or a case of differing definitions but there is at least one architecture that is changing here, possibly (I think likely) more than one - I’m not internal to Tesla so I can’t know for sure of course
how has noone realised this vid is a troll and v11 totally released already?
Your editing skills are A+++. Good video on FSD 11, thank you.
Thanks Mark, I appreciate it!
Love your FSD videos!
Appreciate it!
salut
This video was put together very well. Good job!
I myself purchased a 2023 RWD Model 3 in November, and needless to say having no parking assist functions 5 months later in vehicles this expensive, and what's also considered to be one of the most technologically advanced cars in the world, is absolutely mind-boggling. If Tesla vision wasn't ready to take over, then not only should they not have removed the sensors, but the last thing they should be doing is selling enhanced autopilot at full price when over half of the software package doesn't work as intended.
exactly my same experience with a Nov 2023 RWD Model 3. I am waiting for some news about updates to restore some basic parking assist function!
Ya, I think Elon likes focusing on the long-term vision, and that's great, but it's nice to have a good experience in the short-term, too! :)
Parking assistant in its current state is rather useless. Slow and often stops, or doesn't detect the parking space.
You're not missing out.
That's just how Tesla rolls. They make bold decisions (a good thing) but act on them long before they're ready.
I got my Tesla Model S in December 2016 after they decided to drop MobileEye and make their own Autopilot. It took 6 months for the car to obtain basic functionality like automatic headlights (day/night detection), windshield wipers (you could have them either on maximum speed or off - nothing in between. It took 2 years before rain detection came in, and that still doesn't work well enough to rely on). Not exactly what you expect for a $113k car.
Autopilot appeared 6 months later, with a maximum speed of 30mph. It took a year to get to a usable 70mph. Despite having paid for FSD, having all the hardware upgrades (FSD HW3, Intel MCU, latest cameras) - I still haven't got FSD Beta.
Yet I still love the car and would buy another 😅
Nice Job putting the whole story into perspective - thank you!
Thanks for watching! More V11 stories once it’s released 😉
And just now its released ... a few hours later :D
Glad to see this reality check on FSD. I, too, for a while had been overly optimistic about it until I actually tested it on my new Y using the latest FSD version (purchased on a monthly subscription for a taste test.)
But I’m still bullish that this hard problem of autonomous driving will eventually get solved in the next few years.
Appreciate the honest reviews!
FSD won't be "solved" in a few years. That's crushingly obvious by now. It would require miraculous updates to go from the current driver that can't average ten good minutes in a busy city to be averaging a few decades of good driving in cities. If they really solve general driving in a decade with this approach it would be quite an achievement but still unlikely.
I used to think like that. I’ve been driving beta for over a year. I honestly don’t think any tesla on the road today with hardware 3 will be capable of level 4 or level 5. I think it’s a scam. Tesla knows this, and they are trying to throw a couple more cameras and different cameras into the mix to keep people interested. My bet, is FSD won’t be FSD for at least 7 years, if not over a decade. Every Tesla on the road today won’t even be driving by that time. 😢
Hey at least FSD beta has improved from a totally drunk 12 year learning how to drive, to a 12 year old texting his friends while learning how to drive. 😂
Elon conned you too?
I also tried on a new MY with a month subscription. It was fascinating. It is a very different experience being in the car than watching other Tesla's on CZcams. It could be super impressive and super annoying in the same 15 min drive. But I also think they will be able to figure it all out ... eventually
4 NHTSA items should not really be hard for them. I agree that it is good to hear that NHTSA is asking for a few specific laws-related items. Slowing for flashing, even though you should, will really just keep us pressing the accelerator. Glad to hear they only asked for those 4.
Unknown. You'd think things like choosing the correct lane and responding to a speed sign would be the first things done. The fact that they appear in an NHTSA document today might indicate that it's harder than we might expect.
Love your vids. Keep them unbiased and informative and i'll keep watching.
You always have logical and lucid observations; great to have you on the team!
Epic video! Your content are adding so much value. Thank you! Keep up the good work!
Love the new style and editing, looks great and was just overall a great video on this topic 👍
Imagine all the things it is still likely hopelessly clueless about - the rules for passing a schoolbus, following diversion signs, following cop hand signals, NOT following crazy guy's hand signals, navigating junctions where lights are out, driver courtesy (e.g. stopping when someone is performing an obvious manuveur that needs space), following directional arrows in car parks, etc.
Parking lot driving, pulling over from a busy street to the side and then pulling out again into traffic, reading road markings, remembering where dangerous road areas are and driving with extra caution there, handling heavy traffic, seeing everything like potholes and speedbumps that it needs to see, recognizing what odd objects are on the road, .....
The only thing it can do most of the time is go on straightforward roads in light traffic with simple lane changes and through protected simple intersections. Since that's most of driving for lots of people, they think it's close to solving driving :)
This video was amazing! Please continue with more like this.
I really like this channel, been watching for some time now. You've got a good thing going on. Hope you're finding some enjoyment in it all lol
Another excellent video from AI DRIVR - this is my favourite channel for watching FSD beta videos, always contains honest and insightful feedback
Thanks Tim, really appreciate it!
As always a great video! I am really looking foreward seeing your V11 videos and your thoughts about it. As a German we can only see videos and can not experience it in Germany so far. I think your fsd beta videos are the best!
Great video Cris!
Great points! I agree and think you covered just about everything
I think you're probably spot on!
BTW it seems to be rolling out now to customers, not only employees. Hope you get it soon so we can get your input on it!
Love the content. Keep it comin!
Pretty good summary. I agree with pretty much all these points
really good video, perhaps your best so far!
Since V11 is a big software change, it is expected to have some fall back in the progress. Three steps back, but 10 steps forward eventually.
Great Video as always. Truly the best channel about FSD. Can you please make a video “reacting” and explaining FSD v11 video clips? Just so we can get a sense of if and where it improved/regressed.
Great video as always. Thanks.
Thanks for watching and dropping a comment!
Haha I loved the editing. Thanks for talking about the truth. I'm just glad I didn't cancel my subscription to FSD beta
Great takes on FSD Beta.
Thanks for the honest updates. Do you believe version 11 will be the FSD software most have been "expecting?"
The removal of those sensors makes even less sense if you think about the markets where FSD Beta is not even available (i.e. practically everywhere outside North America). I'm pretty happy my MY still has ultrasonics and can only hope that they won't be deactivated because of an unreliable vision-only software stack.
I’m starting to suspect that the control must be fully handled by neural networks, rather than traditional C++ code, in order for self-driving to work. Tesla seems to be moving in this direction. Will be fascinating to see further developments.
I agree, and this release was supposed to be the beginning of that, which is why it was a bummer to see the same indecisive path planner that V10 has. Maybe I'm wrong about that though. I need some seat time with V11
Tesla has said just this. It's pretty obvious really, you need something intelligent to handle the huge variety of situations.
Great video all the way around, presentation, info, editing, etc. great work! I’ve said for a long time I didn’t feel HW3 had what it takes to do what they’ve promised. Cautiously optimistic about HW4
A splendid spread of facts, opinions, great visuals and pleasing voice. As usual though.
Really, really good video 😊
I know nothing about it, but my guess would be with inclusion of NN, you would have to train it again and run many scenarios. It might be like starting over, but you have a lot of data from previous, but can all be used again with NHTSA involvement and concerns? The NHTSA would also have to approve the changes and who knows how long that might take...
I did however enjoy your take on it ❤
I was hoping we’d see something brand new like you describe, but so far the videos release show familiar path planning and patterns. I need some seat time with it to make a good judgement though
Elon says training computer for fsd will 100X in speed within 2 years.
this is a great video, I agree with all of it. Great visuals as well!
Thank you Ahmed, appreciate it!
About the speed limit adaption: Here in germany with the old NoA/AP combination, the car adapts to changed speed limits much more quicker, like within maybe 10-20 seconds, you realy can feel it decelerating. But still, adapting before the sign would be great to have!
dang i didnt realize they had removed ultrasonics too... i love that red line when i am backing into a tight spot.
Yep it really helps. Hopefully vision can match it
this is what happens when a manager who hasn't programmed in 10 years makes stupid predictions and ignores his software engineers. I've been working as a software engineer and consultant for over 10 years in enterprise world. Rewrites ALWAYS takes 4x longer than you think. That's when you ask the engineers to give a reasonable estimate. Add a bad manager to the mix giving BS estimates, it makes the rewrite more frustrating and painful than it needs to be.
I saw the lex interview when it came out and knew immediately it was a completely BS estimate. There's a saying in software development "under promise, over deliver". In the consulting world, this is an unspoken rule for consultants. When you're talking about rewrites, you need a complete regression suite to make sure the new version doesn't introduce bugs that were already fixed.
The difficulty with ML models is that building a complete regression suite is considerably harder than normal unit/integration testing. Mean while Shenzhen china has autonomous taxi already and it handles more complex situations than Tesla FSD. I feel for Tesla FSD team, they have to work under crazy pressure and deal with Elon's BS.
Loved the video and the editing was excelent. Might just be personal prefrence, but i'd prefer if the static type sound you used in cuts would be just a bit quieter. Watching this video just as I woke up and find myself getting startled by the sound :D
I feel like more constant, smaller updates would inspire more confidence and allow for better progress. But I also see the single stack as a necessary hurdle on the path.. not sure if it's the type of thing you could branch and work on multiple parts simultaneously..
Nice vid, I agree
Great take, I've been pretty pleased with what I've seen so far with v11 so hopefully the final version which is going out to employees right now (v11.3.1)
🤞
I don’t know if I got a background fix to the yellow light problem but I had the car abruptly stop for a yellow it could have easily made. The van behind me had a difficult time stopping and avoiding a crash. This is opposed to running stale yellows a few days before.
Of course, another excellent video.
Interesting! Is it consistently slowing quicker for yellows now? Or was that just a one off?
in south atlanta you will get rear ended if you dont run the first 2 seconds of a red too...
@@AIDRIVRso far just a one off. I just got back from a short drive and didn’t encounter any stale yellows. It did handle a few stoplights well though. I have had several hummm…. moments when new behavior happens on occasion then is not repeated. I makes me feel like they a trying something in shadow mode then revert back to what it usually does.
4:40 Your assessment is valid.
I suspect a non-trivial amount of the autonomy team's resources are being diverted to Optimus development. This could include labelers, the best software engineers, compute time, etc. In the long-run, building a comprehensive vision stack for vehicles and robots will make everything stronger, but I bet it is slowing down V11 to some extent in the hear and now. Just one more thing to add to your list of reasons.
Totally agree with all your points
Thanks for watching and leaving a comment!
Bravo, bravo!
Nice video.
I apprechiate
I appreciate you watching & leaving a comment! Thanks :]
These expectations need to be realistic.
As someone who has Asperger’s, multitasking is difficult for me. That includes driving.
I can focus on one thing at a time,
If I focus too much on my speed, my navigation can suffer, vice versa.
These computer guidance issues are in driving aids from the 2019’s
The NN was fine. There only needs to be a system in place to adapt to frequently traveled areas’ quirks.
That was a great take on the NHTSAs regulation as a good thing, as they actually used it and are providing clear guidance, not banning. Funny timing of the video as 11.3.1 is out for wide release just hours later.
I
I wonder how it'll react to the new radar in the next hardware?
Always entertaining your videos. Lol is he reallllly showing footage of v11?
Thanks!
Thank you so much Burhan! Truly appreciated
Excellent and well balanced. Why more FSD users are of a different mind set is beyond me.
Perhaps some of these folks did not spend decades in engineering development and production, but even so, the warning signs about FSD have been apparent for years.
Or perhaps because they were told 3 years ago that their expensive add-on option would actually work.
Elon recently admitted that the FSD will never be fully implemented onto the current platforms. And will only come to full feature with Gen4 hardware that will not be retrofitted or offered on Gen3 hardware cars. Meaning only Gen4 cars that are not even build yet.
Lawsuits are pilling up, and its only the beginning.
Personally, I would like my Model3 with FSB to simply stop trying to kill me on the highway, by insisting on changing lanes into a concrete divider. Like it did three times today on a 30 mile drive.
Nice timing for this video haha, it's being released to early beta testers as of 40 minutes ago.
I think you are right on target. I for one, hope that V11 gets rid of the dual path of software releases. I find it a bit irritating that they release some new feature that the FSD Beta testers then wait weeks or even months to get.
The beginning of this video is awesome 🤣
I agree with your points :)
Haha yeah I had a good time making this one. Appreciate it!
Good video as always. A completely agree with your view on the NHTSA complaints. Yellow light and stop sign behavior is not flawed in my experience. Lane selection and speed limit behavior is where big improvements are needed.
I would like to know your opinion on Tesla's use of sparse maps for lane selection and speed control. Do you think that maps play an unnecessarily strong role in lane selection and speed control given that the quality of maps varies and being outside the control of Tesla, as least for the present? Is the answer better maps, or less dependence on map data?
That's a tough one. I do think there is currently an over reliance on map data currently, for example there's intersections around me that were recently re-striped and the car can not navigate them on the current SW version. I think both are needed - better maps and also less reliance on map data. I know that sounds a little crazy, but I think it should have an idea about what will be ahead (kind of like the basic maps we have in our head about which lanes go where, etc) and also be able to abandon the plan when it sees something different, like new striping or arrows painted
@@AIDRIVR Yes, I agree, and I anticipate that Tesla will eventually have to build and maintain their own maps not just to address the spotty quality of the OpenStreetMap data, but also to map parking lots, potholes, road closures, hazardous conditions, etc. Fortunately, Tesla has a large vehicle fleet capable of collecting massive amounts of street information. Didn't Elon say that maintaining their own maps will likely be necessary?
Damn that was funny 😂.
Thanks Stig... Eeeeeh... AI DRiVR
I'm pretty sure (the truth is that I hope) that V11 is delayed to include European roads ... You're right, merging highway code that is almost perfect with street code is a great challenge. Not only because it should give same comfort than before but because it will break current code gap between USA/CA and others.
Almost.......but far, like a horse shoe or hand grenade.
Full FSD Model3 Owner here.
Don't hold your breath.
There's a lot of differences between the US roads and the European. There's even large differences between countries within the EU.
@@jsjs6751 Not sure because I'm looking at many FSD videos and it works on roads that are very similar. Common idea is that all USA roads are great city roads but many FSD CZcams creators show multiple scenario. We're not so far and it can be strange but I confirm Elon words: "EU roads should be 90% covered".
@Teslandicap Well, I too do hope it arrives soon, but remains sceptical.
Additionally, in some (maybe all?) countries there may be needed legislative updates to allow FSD.
May take years.
Some countries may even ban FSD as we know it today.
@@jsjs6751 European regulatories updated since last year, some others as Google are pushing autonomous delivery trucks so I think that Tesla does not want to be too late, I hope that it will be there end of this year. And there's another great issue: FSD pack can be bought and if Tesla does not push it, it could be hard to avoid customers to group and give lawsuit. But you're right, the only thing that we know is that we know nothing clear ;)
Great video, honestly, felt very neutral and I appreciated that! Also love the SpaceX badge :p
My parallel parking has not been working in the past few months wasn’t sure if I was doing something wrong but today it worked so still not sure what is going on.
I think you misunderstood, the stop sign recall is regarding the fact that the car is too slow to cross the street even when there aren't other cars around, and yes this is actually almost the opposite of the previous recall
Really enjoyed watching your video. :) First release of V11 will not fix all issues until they include a version that contains the lane mapping, speed limit, and road hazard information. The issues which you noted in your video could be easily fixed by sharing this information with the fleet. Tesla is likely collecting them now as we speak with their V10* software.
I'm dying to know if V11 made stop-and-go traffic on the highway useable again? 🙏 For some time now it's been really aggressive on the acceleration/braking - making it a bit embarrassing.
Well this didn’t age well
Lol and now it just released, nice timing
Having their team focused on two version is a recipe for slow progress. Going to v11 will hopefully allow the team to track closer to the users and allow improvement to be visible.
You make the best FSD videos 😂
Appreciate it!
Wow, so I had only sort-of noticed I don't get parking assist offers anymore when I want them. Are you saying that's completely shut off for the moment?
Yes, seems to be missing from my model Y too. I believe it’s only some vehicles. (with no radar and with fsd?)
I believe the self park feature was removed when they stop using the ultrasonic features. The feature return with fsd 11
Hey man, wanna go ahead and drop a vid talking about how v12 won’t be ready for a while. Hoping to get it wide release like today… jk but it is still funny to me the timing of this video lol.
I just want my damn park assist
I believe the reason FSD 11 hasn't been released to the public yet is that it's currently less safe on the highway. The extra functionality adds more complexity for the vehicle's decision-making, and those decisions are made at high speeds where mistakes can be severe. I don't think Tesla is quite as comfortable saying, "It's the driver's fault if they let FSD do something stupid." More to the point, I think Tesla is afraid of regulator and public reaction.
FSD v11.3.1 looks different but you still need a driver at the steering wheel.
maybe explains Karpathy's sudden exit ?
Correct me if I’m wrong here please, but I swear I remember someone sharing that V11 will using AI to drive. What I mean is currently v10 uses “if then” statements so to speak. It’s hard coded in on what to do for different scenarios. I heard that V11 is solely going to be AI based on the Neural Network that has been mapped out by all the cars. Again, correct me if I’m wrong on this.
this is why u have a feature/product ready, then deliver... not promise and then realizing ops... this was harder then we thought. I know companys do this for investors but in the long run this will be negative for the company. I hope they come out and say, the fsd stuff is not going as planed, we have a plan to get it where it should be and once its ready we will realese it.
I think this vid was a troll as it released at almost the exact same time v11 did. You can look it up, people already have it
I live in Europe and the thing is that there is not that much freedom in terms of full self driving what i think is that the us is al little bit more relaxed about all of this stuff and now as there is more and more coming and problems it seems like that its gonna be more restrictet now or in specific parts of the full self driving
(6:40) You are suppose to stop completely at stop sign ... it is not a yield sign for a reason.
I am holding out for V20 on HW10.
😂😂😂
After version 11 is better that the 2 stack version, it's a bummer. But after that, it should be really a steady climb to at least level 4 in few months. Dojo should now be at a pace as well. Maybe version 11 was the first where most networks are trained in dojo... Can't wait to see 11. 5
I believe from the way they phrased things at investor day that dojo is not currently being used. Hopefully soon though
I wonder when we will see a real breaktrough in that technology.
Do you also know how other car manufacturers are doing with their self driving software?
Good analytics …
Hey man, why do your scene transitions always travel through the Breach?
A glimpse inside the matrix
I don’t know ??? Maybe the recall is an excuse to delay again the V11; it’s simply not ready. I don’t think the recall is very difficult to solve
This is how Windows got so bloated over the decades; piling software on top of software to patch the holes in the updates that they updated.
Starting from scratch led to a better cleaner more efficient experience. I think V11.3.x will be something like that.
Oh god they've gone from backing up their product to denial to straight up lying about the existence
AI DRIVR: Do you have any thoughts on their new HW4 and it seems to have no AP for the first batch of customers. Them not having software ready for HW4 seems to be horrible planning.
I’d bet the recall is the main reason for the tardiness on the releases. I’m sure they had a version working with HW4, but probably didn’t address the recall issues. Not sure though
@@AIDRIVR they came to realization the way they doing FSD HW3 would not work, HW4 now requires complete re-training you can't just switch..
@@teslascoop2177 I don't believe this is accurate. HW4 will do more processing potentionally but the main change is increased bandwidth to increased resolution cameras that need to be stitched into the same "view" of the environment around the car. HW3 will have a lower resolution to extract details from but I suspect the same ai model will interpret both datasets. The question becomes what extra models they put on those extra ai cores.
Love your videos, I keep wondering what the problem is and think up weird ideas about it, like why don't they have more cameras if it's a vision problem, since the computer puts it all together it doesn't make sense to not just have 16cameras around the vehicle with some up higher, or higher resolution cameras that see more sharply, and I don't understand how those radars are a good idea if the car can see properly. But it seems like Tesla are getting their at a snails pace and that is all that really matters.
Stop signs are called stop signs for a reason, and just because we take liberties with lots of aspects of driving, an automated car has to adhere to the law especially whilst human drivers exist on the road. As and when all cars are autonomous, the laws and regulations can be changed to better reflect the ability of the car to improve traffic flow.
Backprop trained networks will always have a loooooooong tail of edge cases that can be very dangerous if controlling driving of a vehicle on the road. The only way you can teach a machine to drive is by first building a machine that can experience and learn about the world like a human. That's not to say it needs arms/legs/etc.. It just needs to learn about the world and all of the things in it. You will never be able to plan for all of the things a car will encounter, ever, and then train a neural network about that list of things in terms of vision alone - it will always encounter things that don't fit into the list of things you programmed into it. Those little models of cars/vans/trucks/cones etc that you see on the in-car-display are human-decided things. What about all the things it wasn't told about? How is it supposed to classify them? A machine must learn about the world before it can drive through the world like you can expect a human to. Trying to force a machine to classify everything as only the things that a human decides upon is a huge mistake. This is going to be a never-ending rabbit hole because nobody understands the problem entirely, so how can they find the solution? Yes, they can develop FSD that is *statiscially* safer than a human, but it will *always* make huge glaring mistakes that could cost lives, mistakes that no human would ever make, and that's what's always going to turn people off from it as they keep hearing more and more about FSD killing people in situations that nobody ever dies in otherwise.
Great comment. I agree, the rabbit hole is very very deep. You can never create the equal of a human brain, with electronic chips and software. In 100 years time, when they've decoded how biological intelligence works, these 2023 FSD efforts will be head-shakingly primitive.
The humble house-fly, zooming through all the rooms of your house, processing genuine 3D vision fields with reaction-processing times of less than 0.1 seconds, is probably 10,000 times more advanced. With a processor the size of the tip of a pin, weighing 0.000001 grams, and needing 0.000000001 watts of power.
I think it has came a long way and inventions are great. We need to respect the obstalce the team and staff are solving. I'm happy we are pushing forward, I'm not here to rush the tech and AI, it is brand new and never been done. So many years ago the car was first created and look at where we are now. Great work Tesla and other companies attempting to solve problems we can not solve ourselves
If only the cameras took a glance farther ahead every one or two seconds, it would make it so much easier for FSD to determine where the car should actually go. So often when lines are sporadically erased the FSD drives exactly like a person that is not looking far enough. If FSD did take glances ahead, it would easily be able to fill in the sporadically erased line, to figure out if the widening road is simply parking spaces that disappear as soon as the road narrows down again 500m ahead. Often FSD insists on changing lanes in between two cars when it would be obvious that there was lots of space in front or behind these two cars. Preventive driving would make everything more smooth and safer too. Can the cameras zoom in and zoom out?
Aged like day old milk
There is no reason NHTSA recall should stop V11 if it’s complete. An update with recall fixes could be sent when ready. At 4 months on 69.25.2 we need some improvement.
The speed sign issue needs to be addressed for non-beta cars as well. Super annoying. And I don’t see that I’ll get access to beta any time soon in Sweden. :(
No park assist sounds less frustrating and more lawsuity
S/X don't have Ultrasonics either. Sad.