Tesla FSD Makes "UNPRECEDENTED PROGRESS"!!
ฝัง
- เผยแพร่เมื่อ 17 มี.ค. 2024
- The first 100 people to use my promo code DRKNOW40 will get 40% off their first order of Hungryroot bit.ly/3OYbovC!
Ashok Elluswamy, Tesla's Full Self Driving lead, has posted that Tesla FSD v12 will make unprecedented progress in the next several iterations--as if the current state of FSD 12 isn't amazing enough already! Let's discuss why this should be happening, and how Ashok would know already.
**If you are looking to purchase a new Tesla Car, Solar roof, Solar tiles or PowerWall, just click this link to get up to $500 off! www.tesla.com/referral/john11286. Thank you!
Join this channel to get access to perks:
/ @drknowitallknows
**To become part of our Patreon team, help support the channel, and get awesome perks, check out our Patreon site here: / drknowitallknows . Thanks for your support!
Thanks to Hungryroot for sponsoring today's video!
Get The Elon Musk Mission (I've got two chapters in it) here:
Paperback: amzn.to/3TQXV9g
Kindle: amzn.to/3U7f7Hr!
**Want some awesome Dr. Know-it-all merch, including the AI STUDENT DRIVER Bumper Sticker? Check out our awesome Merch store: drknowitall.itemorder.com/sale
For a limited time, use the code "Knows2021" to get 20% off your entire order!
**Check out Artimatic: www.artimatic.io
**You can help support this channel with one click! We have an Amazon Affiliate link in several countries. If you click the link for your country, anything you buy from Amazon in the next several hours gives us a small commission, and costs you nothing. Thank you!
* USA: amzn.to/39n5mPH
* Germany: amzn.to/2XbdxJi
* United Kingdom: amzn.to/3hGlzTR
* France: amzn.to/2KRAwXh
* Spain: amzn.to/3hJYYFV
**What do we use to shoot our videos?
-Sony alpha a7 III: amzn.to/3czV2XJ
--and lens: amzn.to/3aujOqE
-Feelworld portable field monitor: amzn.to/38yf2ah
-Neewer compact desk tripod: amzn.to/3l8yrUk
-Glidegear teleprompter: amzn.to/3rJeFkP
-Neewer dimmable LED lights: amzn.to/3qAg3oF
-Rode Wireless Go II Lavalier microphones: amzn.to/3eC9jUZ
-Rode NT USB+ Studio Microphone: amzn.to/3U65Q3w
-Focusrite Scarlette 2i2 audio interface: amzn.to/3l8vqDu
-Studio soundproofing tiles: amzn.to/3rFUtQU
-Sony MDR-7506 Professional Headphones: amzn.to/2OoDdBd
-Apple M1 Max Studio: amzn.to/3GfxPYY
-Apple M1 MacBook Pro: amzn.to/3wPYV1D
-Docking Station for MacBook: amzn.to/3yIhc1S
-Philips Brilliance 4K Docking Monitor: amzn.to/3xwSKAb
-Sabrent 8TB SSD drive: amzn.to/3rhSxQM
-DJI Mavic Mini Drone: amzn.to/2OnHCEw
-GoPro Hero 9 Black action camera: amzn.to/3vgVMrH
-GoPro Max 360 camera: amzn.to/3nORGYk
-Tesla phone mount: amzn.to/3U92fl9
-Suction car mount for camera: amzn.to/3tcUfRK
-Extender Rod for car mount camera: amzn.to/3wHQXsw
**Here are a few products we've found really fun and/or useful:
-NeoCharge Dryer/EV charger splitter: amzn.to/39UcKWx
-Lift pucks for your Tesla: amzn.to/3vJF3iB
-Emergency tire fill and repair kit: amzn.to/3vMkL8d
-CO2 Monitor: amzn.to/3PsQRh2
-Camping mattress for your Tesla model S/3/X/Y: amzn.to/3m7ffef
**Music by Zenlee. Check out his amazing music on instagram -@zenlee_music
or TH-cam - / @zenlee_music
Tesla Stock: TSLA
**EVANNEX
Check out the Evannex web site: evannex.com/
If you use my discount code, KnowsEVs, you get $10 off any order over $100!
**For business inquiries, please email me here: DrKnowItAllKnows@gmail.com
Twitter: / drknowitall16
Also on Twitter: @Tesla_UnPR: / tesla_un
Instagram: @drknowitallknows
**Want some outdoorsy videos? Check out Whole Nuts and Donuts: / @wholenutsanddonuts5741
Source Post: / 1769351630741274682 - วิทยาศาสตร์และเทคโนโลยี
The first 100 people to use my promo code DRKNOW40 will get 40% off their first order of Hungryroot bit.ly/3OYbovC!
No thanks.
I got 12.3 yesterday. NC research triangle area. I drove around for roughly 4 hours putting the car through every situation it used to screw up on and it handled almost everything flawlessly. Super impressed with the end to end architecture and can't wait to see where this is at in a few months time.
This is what I've been hearing and seeing 'everywhere'. Even from FSD-skeptics.
@@rockycata6078 It really is a huge step up from V11. Even when it screws up there's no harsh breaking or jittery steering wheel, it's very smooth. I've had 2 takeovers so far that both involved it getting a little close for comfort to some curbs (although I still do think it would've been fine) so I do think that needs a little work there but 99% of the time it's like there's a human behind the wheel and it just does what feels right.
@NeverOutslayed also been seeing this opinion everywhere . I think we are pretty much at the point where watching FSD vids arent going to be enough to see the progress going forward. We need Tesla to start sharing the safety data for each version compared against human safety, and i really hope they do start doing this
@@GamerBoyRobby You haven't seen the quarterly /VehicleSafetyReport?
Received 12.3 yesterday in Georgia. Had FSD since 2019… this is absolutely a step change! People love to nitpick, but I have never been so optimistic!
Where in Georgia?
good to know.
John, I think you’ll find that 12.3 is FAR better than 11.x. I got it last night and have been using it today. It is a phenomenal upgrade over 11.
I got 12.3 installed yesterday and took out on a drive to the local Tesla service center on Merritt Island, FL It performed much better - no weird lane changes, take turns smoothly, turns on turn signal at good distance before turn. A little slow on acceleration after green lights. Speed offsets are a little hard to use to set your driving speed.
Need to see the missing pieces included such as reversing, parking and managing courts/cul de sacs and improvements on roundabouts. Once they are done, handling adverse weather such as heavy rain, heavy rain at night, heavy snow, heavy snow at night and ice on the road. Then its just a matter of how good before they put it into a jurisdiction somewhere for approval.
I'm an original FSD Beta tester (10.2) and got 12.3 in ATL 2 days ago. It is sooooo fundamentally different. In a nutshell the biggest change from V11 to V12 is from the pure robotic actions to very human like driving behavior.
This is so exciting. Thanks for the report.
I have a hard time explaining the difference to my wife. It’s crazy. It’s so smooth and confident.
just got it last night and wow just wow, so smooth in the streets. not perfect but certainly an improvement.
I’ve been driving with 12.3 for the last two days. This is it. It’s INCREDIBLE.
I got it Saturday. Definitely a order of magnitude better... All the mistakes I saw earlier on my personal drive have been fixed.
Thanks for the FSD update.
Thanks for the in-depth analysis of FSD from 2015 to present.
I'm just Joe Blow nobody and I got it 2 days ago. Works great when no one is on the road. But it gets shy when there's traffic on surface streets. You definitely want to be ready to take over more than ever. The car will start to take you out of left turn lane a second after it put you there.
Your análisis is spot on... you are well ahead, observation, deduction interpretation... spot on! Hats off sir!
Got 12.3 in my M3LR! So much smoother. They did break on thing. On Hwy 75 in GA it now reads Minimum speed of 45 MPH as the speed limit. But on some turns it no longer stops and shakes the wheel, it just smoothly makes the turn. It uses its own judgement regarding speed limits and now slows in residential zones. Nice.
I still think we are still a few years away. We as drivers have short term memory fsd does not appear to have this yet. I think this will be required to truly get level 4 driving.
We shall see. I'll hold back on celebrating.
My comments on future progress:
1. Navigation accuracy needs to be improved in many areas including where I live in NC for FSD to have non-intervention drives.
2. I believe they are already doing this. If the collision detection system believes a crash will occur it can override the FSD. That ‘compliance AI’ should also continue to be improved rapidly, particularly when it comes to pedestrian detection. They can’t afford to hit a person or a kid while the system is in operation.
Don't recall who it was, but recently saw a 12.3 vid showing it trying to squeeze through a tight gap, which it was doing successfully, but then the emergency stop overrode it and stopped the car.
I drove four short trips on Saturday. Only one intervention on my part. At a light where I was turning left across a stream of traffic, the car stopped mid-intersection as the light changed from yellow to red. I forced it thru the light as it changed to red. Lane selection: very good; speed bump detection and handling: v good; lane selection: good+; turns: v good. John, let me know if you need a test as I am relatively close.
Oh dude, I love Beyond Burger. As for the version 12, I truly believe that this year we'll all be blown away by what it can do
Got 12.3 over the weekend in Northern Virginia. It is noticeably better! It no longer turns into turn lanes one block too soon. Very, very nice!! Now... if they would just fix the speed limits then I could get home without any interventions at all. Speed limits have been wrong for me in certain locations since early 2022. Other than the speed limit errors, I had no interventions in my 45 minutes of driving to multiple targets. Looking forward to more time with it!
@DrKnowitallKnows, great video and show.
To assess capabilities I think we need to be much better at defining what we mean.
For me FSD will first be out of beta when I can use it as L4. To get to L4 there is a set of defined capabilities and regulatory acceptance. In the US regulatory acceptance is more a liability issue than meeting a defined capability set of requirements. What will the insurance be when you enable FSD L4, tesla should have this covered with their own insurance.
Regarding FSD capabilities, then the question is not how much better FSD on average are compared to a human, rather it must better all the time (in 100% of the cases) or a human will still have to monitor it (L2). Tesla will have to prove that average humans would also fail in the cases where FSD fails, which some number of magnitudes.
As progress is made the known problem areas decrease. Eventually, they will get there. I see amazing progress.
I have V12 in NW Florida
I got it but there are most definitely regressions. Lots of good progress too. But still needs work. Things that were goodnow broken (getting into correct turning lanes, and not stopping when a car in front hesitated pulling into parking lot awkwardly and stop and go). I am in Norcal. But all that being said, the “naturalness” overall is really really good
I fully agree with you.
'Cambrian Explosion /Revolution ' Yes!!!
I'm still waiting on 12.3 too.... You are not alone!!
I got 12.3 Sunday on my MY in Atlanta. My son, who got his MY 3 weeks after me does not seem to have it yet. So far so good, but autopark does not seem to be showing up any more.
NHTSA needs to investigate the safety of FSD, the day they find FSD to be just 1% safer than the average driver, it should be mandatory on every new vehicle produced.
Its already 10 times safer why are you waiting? I now have 12.3.
@@tedmoss You can go to sleep in the back seat? It is only safe with a competent driver atm.
What criteria determine the safety of a driver? How do they determine the "average" driver? Statistics can be skewed heavily based on a relatively bad group of drivers imo.
V12 is a big change but a big step backward on narrow winding roads. It drives to the right (unlike v11), however, off the side of the road. I assume it's the lack of training on such roads.
I’ve had FSD started 9/2019 and I have not gotten it either. I guess I’m a lower class Tesla owner😖
I keep having the same problem nearby at a bridge in Allenstown NH over the Suncook River where the bridge was under construction for a bit over two years mostly spring into fall. Now mu model Y keeps stopping short of the fully fuctioning bridge and has been for many months now. My black Y named Bear balks at going over the bridge without stopping. Reminds me once of an old pony and cart I had and took my very young children for a ride to the school a couple miles away. Our beloved pony Kaleb suddenly stopped amid stride and refused to cross a yellow line going across the road. A definite "no go" despite my encouragement. We turned back. Maybe a name change is in order to Kaleb!
Maybe the pony and car know something you don't.
In terms of generalisation, do you think the foundation model can be used for LHD countries?
The YT channel from Hans Nelson has a good explanation of how using this type of software can result in super-rapid changes. The use the example to teaching computer to play Go. Not sure if I should post a link to some else but you might want to look at it. The process seem s to be a lot like an Hegalian dialectic to someone without any real software background.
Here in the UK I’m still on AP not FSD, today it refused to change lanes on the motorway (highway) although it was a 3 laner just refused point blank. We then went through a bit of ‘narrow’ lanes due to road works, it got really nervous jerky braking / acceleration getting really close to the large trucks in the lane next to me, I took over and drove for a bit until we got clear of the narrow lanes section.
Then the sun hit the car, it came up with a warning to say the right camera was obscured in some way ? Sunshine shouldn’t be an issue, these cars work fine in California surely they can manage to operate in our weak, infrequent bright periods ?.
It’s not a great advert for upgrading to FSD is it, don’t get me wrong, love the car, ten years driving EV’s and this is the best, range, equipment etc but I really upgraded to a Tesla for the charging network (excellent) and autonomous driving in the not-too-distant-future as I’m 70 now so thinking to the future when I will need a clever car to do the heavy lifting, right now that isn’t a Tesla which is a great pity as this is my dream car….meanwhile my squirrelly AP awaits me !!!!
I'm waiting for end-to-end improvements in highway performance. Most of my FSD miles are on road trips, and mostly out west where there's lots of open country. v11.x has consistently had too many phantom breaking events to allow even using it on most bright/hot days. Looking forward to v12.3, but even more to v12.x that fixes this show-stopper issue.
Haven’t heard people complaining about phantom breaking since V11. But don’t have it myself.
@@ken-mb5cpHaven't taken a road trip in hot weather for 6 months, but it was a serious issue then. Maybe fixed???
I’m tired of waiting for the FSD V12. I did pay the full amount for it and I did register as a beta tester. Good grief
It sounds like you've had experience with Nepal! If you enjoy Nepali cuisine and find yourself meeting your folks in or around Springfield, VA, let me know. I and the family would love to have you over.
You predicted CyberTruck would take maybe 6 months to get FSD (or autopilot) and I thought that seemed very long. You said it would need to get a lot of miles on the CT to train a completely NEW NET. With Elon saying he wanted to license to other MFG Cars this seems strange. So every new car from every mfg would need to drive 6-12 months to train a NEW NET for every single car model? I have 2 months now with my CT and not have AUTO STEER for trips is a bummer with about 6+ years owning Tesla's.
Just got it. You and I have been very similar timing for version upgrades, so, any minute now for you.
Looks like v12.3 roll out is slowing. Maybe hope for 12.4? I'm in the same situation having V11.4.9.
Does anyone know if they’re retraining the entire model or they are stacking LORAs for specific use cases?
Got it!
New training and way of doing things will likely mean less or no regression with each update. In comparisons in the past we would see some regression on new updates along with new behaviours.
If you did the 2024 dash icon size update (like me) you are going to have to wait for a new release on the current branch. I'm normally very early on releases - first 200-400 or so. As far as being too optimistic, I think Tesla pushing back the next gen/robotaxi until late 2025 or 2026 is a good indicator of when they think L4 FSD approved by regulators will happen.
I got 12.3 two days ago
Your channel should have 12.3x more subscribers!
A limitation may be the hardware version. Is only hardware 4 currently using FSD 12? Agreed though that end to end neural nets will result in magnitude improvements in short time. Tesla can daily access millions of miles of data to mine & train for their best drivers and specific scenarios (eg “road closed”).
OK, great progress - I won't be able to comment on my FSD until mid-June when I'm back at the house that has our 2022 Model Y.
But in the meantime - as Elon said in his FSD White Papter 5+ years ago
"FSD is ready for final release when each of the 50 states certifies it fit for use on their roads."
Don't panic - Elon never said that because he never wrote a White Paper about the most complex.
coding task in human history.
The sad truth is Elon never told us when FSD is "ready," all we can do is jump up and down
after every release and say, "It's closer." Kind of like each new calculated digit of Pi gets us closer to the actual value of Pi.
I’m in the uk so really interested in how long it will take for us to get it here . As we drive on the left I presume they will have to train a completely new model also if a rule I’d different for the same thing in two different countries,again I presume you will need to start with a blank model. For example in France you have right of way onto a roundabout ( trust the french to be different to everyone else). Where as in other countries cars on the roundabout have the right of way .
Again traffic signals are different, we don’t have left turn on red at all , and no flashing orange or reds ( which that seems like something that would be of benefit to us ) .
Very interesting times ahead.
Just flip the training videos!
...
🤭
@@user-gv4cx7vz8t that’s an interesting idea wonder if it would work.
Don Browning has it!
I got 12.3 yesterday, and I'm not fanous like you !!!!😂
have you roasted him yet ? do it... I dare you... :D
One way that unmanned, fully autonomous self driving cars might make their debut is making late night and early morning deliveries (say from 2 AM to 5 AM) when there is not much car or pedestrian traffic.
In doing this they can collect much data as well and maybe even be used by police departments to provide added security during those wee hours of the night.
Also maybe that might be a way to self drive a Tesla semi truck, those new ones with the sleeper cabs where the human truck driver drives by day and the self driver drives by night while the human driver sleeps.
That could be a win win in that the human driver can make his run faster but still has a job. And the self driver runs in the wee hours of the night where there is less traffic.
And if the self driver sees something that does not look right it can awaken the human driver to take a look see on what that is and how to handle it.
And as for range limits, just as aircraft trade range for payload maybe the Tesla Semi can do likewise and for longer runs carry extra batteries in its trailer, thus giving up some payload for extended range.
And the sleeper cab could have a monitor so the driver need not even get out of bed to see what the situation is and even instruct the self driver on what to do then go back to sleep, though maybe say half as much as when awake. For the idea is not so much to save money but to make quicker runs.
Also,, since the traffic is less at night the self driver could be more cautious in that even if that slows it down a bit, the reduced traffic might offset for that so one can in addition to trading range for payload one can trade (average) speed vs safety.
And the threshold trigger for waking the sleeping human driver might be adjustable as well so that any driver that values his sleep and trust the self driver can set it high while any driver that does not trust the self driver can set it low, and thus not complain when it wakes him more often, for that is what he (or she) told it to do.
And yes, with this the human driver can be paid for the time he (or she) is asleep while the truck is in motion.
And here is a bit more elaborate fully self driving scenario.
Unmanned self driving Tesla trucks deliver Amazon deliveries to distribution sites around a city late at night/early morning. These could be like a Dempster dumpster like Container sitting in a parking lot.
These might be such as to cover say one or two mile radius for last mile deliveries.
Then Tesla Optimus Robots riding Cargo E bikes can make the day deliveries within that one or two mile radius and then dismount from their e bikes and place the merch on the customer's doorsteps , remount their e bikes and move to the next delivery site..
As for using e bikes for last mile delivery one might Goggle
It's time to replace urban delivery vans
And BTW, a e-bike can be made to go really fast with a big enough motor
e.g. Goggle
Top 5 FASTEST ELECTRIC BIKES In The World You Can Buy
So these being lighter poise less of a danger to other traffic but mostly would be traveling on roads with 30 -40 mph speed limits and not freeways and such.
So Tesla can make the Tesla Trucks and The bots and maybe even the cargo bikes as well and sell (or lease) them to Jeff Brazos.
Now there is a unique ability that a humanoid robot has over its human counterpart, it can instantly be reprogrammed to do some new task while it might take years to retrain a human to do a new tasks.
And given a certain degree of dexterity a bot can change form one task to some other with just a different software load, or mode change, or maybe one should say a role change, or one might call that changing hats.
Thus when a humanoid bot is task with some role, say last mile delivery for Amazon and then can ride an e-bike, get on and off the bike, and pick up and place a package plus navigate and such, it could then perhaps suddenly (at the drop of a hat, pun intended) change hats and do something entirely different if the situation arouse.
For example if we had that network of bike riding bots delivery Amazon packages all through a city and someone needed a defibrillator really fast, the network could then identify the closets bot, and if it carried a defibrillator in its cargo e-bike, it could then change hats from a Amazon delivery Bot to a first responder EMT.
And doing so it might hand off from an Amazon controller net to an EMT net and then act both autonomously as well tele-controlled by an centralized EMT human or doctor.
So there is this huge potential for rapid role changes that humanoid bots intrinsically bring to the table simply by being human like and thus adaptable to new task, just by assigning a different role to them.
Thus they could change a hat from a delivery bot to an EMT or a deputized police bot and such.
And whether this leads to a dystopian or utopian future is up to us (and it probably will be a bit of both) there could be many benefits from having this ability to suddenly change hats.
Take for example the problem of Swatting.
e.g. Goggle
Inside the alarming trend of 'swatting'
If before a human SAT team arrives on a scene a Humanoid deputized bot could arrive first. And it would have no fear so it might be far less trigger happy (if it even has a trigger at all in that it need not be armed) and thus reduce the risk of accidental shootings.
And so one could imagine any number of other examples of how this ability to rapidly change hats might prove to be beneficial.
So BTW, does anyone want to make a bet or prediction when the first baby will be delivered by a bot, albeit tele operated?
Will they have different FSD models trained for different countries? Each country has its own style of roadsigns, different highway codes, drive on different sides of the road, etc
If a human can do it fsd can too. In time of course.
The US has some really badly designed roads e.g 4 way stop signs, in other countries like Australia they are almost gone replaced by roundabouts, give way signs and traffic lights. Plus turning while lights are red, not a thing at most intersections in other countries that use green/red turning signals ( arrows ) as well. But in places like Australia and New Zealand the Tesla fleet isnt large so accumulating data will be slower for right hand drive vehicles. My bet is after US release they will try to do Norway and the UK. Competitive pressures from BYD might decide when they do China.
So now that it's NN end-to-end, will our cars get better at driving without updates? Are they learning on their own?
FSD does not teach itself. Are you sure you'd want that? Imagine buying a used car poorly trained by a previous owner, or teaching a car your own bad habits.
@@user-gv4cx7vz8tOur NNs teach themselves. How can one like FSD be it be limited to only cloud-based learning? What limits it? Who is going to keep the genie in the bottle?
@@andyfeimsternfei8408 The version of FSD you're driving with does not change on the fly. It requires an update from Tesla. That is how quality is enforced and chaos is prevented.
Interesting that you never mention Wayve in your videos. Those Tesla engineers who went to Musk suggesting an e2e approach did so after seeing Wayve’s results
Can anyone say wether or not Tesla is currently using Dojo? What’s the status on that? I’ve been so interested in Dojo since they began talking about it but haven’t heard anything since….
Tesla is using pupu!
Love the Hopium. Currently holding 356 shares at $229 avg.
@@ken-mb5cpit exists just as it did 6 years ago. I’m still seeing it do things that could result in accidents without interventions - on relatively short drives. It’s nowhere close even if it’s much better than before.
I got a navigation update not FSD 👎😩🤷🏾♂️
I guess that insurers and Tesla use similar metrics to identify good drivers. How about using pulse rate, blood pressure, facial expressions etc of human passengers to provide additional driver performance measures?
Though Tesla does not tell us, though I strongly suspect they utilized MIT's VISTA RL system with 19 neuron-like modules (modelled after a 300+ neuron worm, using ODE, "liquid time constant" system), just an "educated" guess, for FSD 12.
I see the FSD ver 12 as being like alpha go. I look for it's capacity to explode as Tesla brings more computer power to their training program. When you look at how far it's advanced in 15 months or so and you know that Tesla will soon have orders of magnitude more computer power to train with its easy to imagine it also increasing orders of magnitude in it's ability. With end to end AI increase training by 10x and you should get 100x improvement. It's going to shock the whole world by years end if Tesla gets the computer systems they have discussed.
Nice to be loved :)
it should be Tesla's ChatGPT moment!
NN will pump exponential improvement of FSD in next few months leading to commercial version stripping off Beta from the version by July 2023
They need more compute to crunch the data so waiting for Dojo to ramp. Elon said by end of year Dojo would be mega. Like the Semi, it's all promises.
Suspect that your Model Y HW4 priority is lower than HW3 minimum operational FSD platform target goal. I wish I could be an alpha tester so I could interact with the development team about how some of the legacy decisions put into Version 11 for patching my corner cases are generating new unsuitable behavior on the early Version 12 limited preleases, even though they show massive improvements in alot of other ways. 😂
It’s not fair that those with legacy Teslas got moved to software stack 2024.7.2 and won’t get FSD V12 for who knows how long.
V12 will run off the side of a narrow winding road. V11 never did. V12 is also much slower and unsteady on such roads. My car drove hundreds of feet, 2 feet off the edge of a single-lane mountain road leading to my house. Be careful!
At this time, according to TeslaFi, it appears 12 is only going out to the southern half of the country. TeslaFi has only a handful of installs in states above the Mason-Dixon line. I wonder if they are still cautious about how it will do in snow.
BlackTesla has it and has reviewed on his channel. He is in the North East
soon you will have FSD autonmy and GROK as the assistant talking with you abou the road, or anything you want. haha
Could someone explain how training on the best human drivers can make FSD better than those drivers. If all FSD knows is how the best drive then how can it extrapolate beyond that (and it needs to be demonstrably better for the regulators).
The best class of human drivers does not have each driver near or at the top of that class in every situation. Nor do any of them perform at their individual best 100% of the time. FSD ought to be capable of both.
A weak analogy is hearing a live concert of a very good singer vs an album representing many recorded takes edited together, plus with autotune focussed on fixing the edge cases (pun intended). You could also merge the best vocalization techniques of many top singers to get the superhuman recording out.
And FSD has 8 eyes constantly looking and never gets distracted. It beyond what any human can ever do.
@@feedvid Right. And we didn't mention superhuman reaction times.
I still have to get used to the compressed audio from the Dr.
That's not the way he talks?
@@user-gv4cx7vz8t he for sure gets more words through per time unit than most
I want me my V12!! I think my wife will relax a lot after the next download. 🙂
I assume, but don't know, that Tesla has extensive regression tests for each new release. The big concern with an AI approach is that some form of "mental illness" might creep in at some point and not get caught until it is experienced in the wild.
I’ve been driving my Tesla for over 3 years now. Before I purchased it, I really looked forward to FSD. The vast improvements keep coming. But it seems to pale in comparison to the driving abilities a 10 year-old can master in less a quarter of the time, if not much less. It seems that for years we’ve heard the tremendous leaps being made. Not saying it’s impossible. Just saying it’s exhausting to hear about the great progress being made. Not blaming anyone for the lack of progress. But perhaps highlighting the arrogance in stating that true FSD is coming very soon. And, frankly, the folly of paying much attention to this topic. The last video I watched on Tesla’s “amazing”progress was about 9 months ago. And the one prior to that one was 18 months ago. And it sounded the same. Lots of progress but not there yet. It’s still in an infantile stage, like a 10 year-old, one or two months into their daily driving lessons. I’ll check again in January of 2025!
Maybe you aren't getting v12 because Tesla doesn't have much training data around Athens. Does anyone around Atlanta have it? If robotaxis require training in each supported locale, FSD is sounding more like Waymo.
I have it in ATL. I also live in midtown (middle of the city).
Don’t you think that the progress on a single stack proves that TSLA has less advantage over future competition?
It will be interesting to see where it’s at the next 6-12 months. I’m optimistic at least. Glad to see improvements. The multitude of nuances in daily driving is just unfathomable to imagine but as dumb humans if we can do it, ai certainly can.
I’ll be satisfied when it can find parking in an indoor parking lot with multiple floors
Hear me out: train a separate version of FSD with rally cars. If FSD spidey-senses imminent danger; use the rally skills to GTFO.
FSB doesn't have the ability to fail quickly like other AI neural net developments. Failing would mean lots of car crashes. Since it can't 'fail' in the real world, how does the neural net learn? And learn quickly?
Tesla should release to all.
I hope you're right. BTW, nice JIF.
Is having some sort of memory also part of this AI that is driving? I mean like when a human being drives a lot himself and encounters many different situations on the road that may be called complex or tricky and then solves those by driving a certain way, e.g. gives gas/accelerates quickly to solve a certain situation on the road, and then this succeeds, then the human being remembers this and can take it to all future situations on the road that are similar to it, so then it probably will do these things quicker then, more automatically/instinctively in the future. So it will have learned and will recognize similar situations in the future and remember what it has succesfully done in the past and then do it again. Does this AI of these Tesla cars also have things like that? How does it work? What are some tips and/or resources I can look up to learn more? :)
You will get the Version 12.3. soon. after I get it I think. Who know maybe a random generator that is broken heeeheeee
Once FSD proves to be safe I think Ford will work a deal to install system on their vehicles. It's a no brainer and then the rest of the automakers will follow. You can't beat the extra safety and anyone know that the population is getting older and so a little help with FSD goes a long way toward safety for driver and innocent by standers
I have it and I'm nobody!
Nobody has it?
Funny with all these so-called long term investors, that grabs any chance to pump the stock short term. Why is that? Why not let the stock stay low as long as possible so you get to buy at good low levels for as long as possible?
Hungry root is great!! They have made healthy eating easy. And I actually save money.
FSD is still FAR from "an average driver:" Without human intervention, the mean-time-between-accidents for FSD would be hours or days. For the average driver it is years. (I have 12.3 already - better, but not really human.)
No FSD 13 will NOT be coming, so "The Next Step" is a commercial-quality platform[...X.FSD?]. BTW, as #DOJO gets more robust at 'graphic-compute' and end-to-end communications, FSD will gain in acceptance. The 'march-of-nines' will trample every driving test NHTSA or anyone else can devise. Yeah, please be advised to still PAY ATTENTION to what is happening whenever/wherever you drive, but also how FSD drives. The 'limiting-factor' is when to put your children in the vehicle and know they are as safe or safer than with you driving. This is more psychology than science, but you put them on that school bus every day with only a prayer.
FSD in Indian Metros will never be a success unless it is assisted by some kind of AGI.
I have v12.3 and I don't even own a Tesla yet.
Yeah, too pessimistic. It will quickly be better than the best human drivers because better vision and faster reflexes and more driving experience.
You should buy and hold tech stock Crypto investment
There are many interesting stocks in many industries that you might follow. You don't have to act on every forecast, so I'll suggest that you work with a financial advisor who can help you choose the best times to purchase and sell the shares or ETFs you want to acquire
You should know it's easy and hard to make that much profit. I say EASY because it's very possible to make that much, and Hard because you'll need professional assistance to do it, I'd suggest you get assisted by a market advisor.
Same here all thanks to Mr Jasper Strowman, he has always been there to advise and guide me through with detailed analysis and recommendations that I wouldn't have access to otherwise.
Indeed, most people downplay the roles of financial planners until burnt by their mistakes. productivity is optimized and keeping up to date strategies and analysis makes it more lucrative. I've been able to navigate the volatilities and scaled up 19k from 5k with professional guidance
Oh stop whining! 😉
I didn't get 11.4.9 before others were getting 12.1 because I had dared to buy a HW4 car..... and to be honest 11.4.9 was so bad in my area of Oregon that I don't use it.
11.4.9 has a lot of issues on 55 mph two lane curvy roads.... the majority in Oregon. The car still slams on brakes routinely for shadow covered roads.... we have a lot of trees here.... and for oncoming trucks, especially ones coming around a curve from the right. It has been so bad here that I felt it was a safety hazard and just don't use it. I use TACC but it is almost as bad until you get to straight roads or four lanes.
I'm desperately waiting for 12 because I think it might resolve the issues I had in my area and I will then use it constantly.
I'm in the same boat you are-HW4. So far, according to TeslaFi, there are almost no cars left with 12.2.1, and only 1% of 12.3 installations are going to HW4 cars. So I think we have a long wait ahead of us!
I GOT IT, I GOT IT, I GOT IT.... I AIN'T GOT IT. Immortal words of chauffeur in Mel Brooks' High Anxiety, as he lifted suitcases - and always dropped them in the end. Basically words of Musk and KnowitAll every year for the last 10 years - and for the next 10 years. Total failure to understand that AGI/FSD is a whole different kind of problem requiring a whole different paradigm and not just an extension of narrow AI/Driver Assistance.
You still use some misleading expressions regarding autonomy. There is no getting to “it” or “solving” autonomous driving. There only is a progression of improvement (March of Nines). We already have autonomous driving, it just isn’t reliable enough for Robotaxi. Not sure why someone as smart as you continues to use those terms…
Please don't buy TSLA next week. I want to buy more.