CNBC Expert: Tesla FSD is "A JOKE"! FSD Beta is So Far Ahead The Experts Don't Even Understand It!
ฝัง
- เผยแพร่เมื่อ 19 มี.ค. 2023
- In a recent CNBC video about the collapse of the autonomous vehicles, Tesla is put in last place, after companies that don't even operate anymore, with one expert calling Tesla's FSD "a joke!" How can the mainstream media and supposed autonomy experts be so dead wrong about this technology? A tweet from Elon Musk and a bit of thought reveal why.
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!
Go to athleticgreens.com/drknowitall to get started on your first purchase and receive a FREE 1-year supply of Vitamin D3+K2 and 5 travel packs.
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!
**Lift pucks for your Tesla: amzn.to/3vJF3iB
**Emergency tire fill and repair kit: amzn.to/3vMkL8d
**A CO2 Monitor: amzn.to/3PsQRh2
**Check out Artimatic: www.artimatic.io
**For 30% off the amazing OneCarStereo instrument display for Tesla Model 3/Y, use this link: bit.ly/3vd7NRv
**Want to get in on WeBull's awesome Crypto and stock fun? Check out this link, and get started trading stock and Crypto!
a.webull.com/i/DrKnow-it-allK...
**If you are looking to purchase a new Tesla solar roof, solar tiles or PowerWall, just click this link to get up to $500 off! www.tesla.com/referral/john11286. Thank you!
**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 NTK Studio Microphone: amzn.to/3vjMomW
-Focusrite Scarlette 2i2 audio interface: amzn.to/3l8vqDu
-Studio soundproofing tiles: amzn.to/3rFUtQU
-Sony MDR-7506 Professional Headphones: amzn.to/2OoDdBd
-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
-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
* AI Superpowers, by Kai-Fu Lee: amzn.to/3dgRSs3
* A 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
**JOWUA:
*For an awesome wireless charging phone holder mount, check out: bit.ly/3tRMSQR
use the Promotion Code: DRKNOW5
*Tesla USB Hub: bit.ly/2PXHPit
use the Promotion Code: DRKNOW5
*Powerful wireless iPhone/Android charger for your Tesla Model 3 or Model Y, check this link: bit.ly/3b5fHSW
use the Promotion Code: DRKNOW5
**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
CNBC Video: • Why Ford And VW Shut D... - วิทยาศาสตร์และเทคโนโลยี
![[Vowel]물고기는 물에서 살아야 해🐟🤣Fish have to live in the water #funny](http://i.ytimg.com/vi/3G9MH063Cx4/mqdefault.jpg)
I wouldn't believe CNBC if they told me water was wet 🙄
CNBC posts a disclaimer after every show, disavowing any responsibility for anything their "anchors" or "guests" say.
People will only realize FSD is here when Hertz purchases it for their entire fleet and then makes an ad in which a football athlete talks about it.
*Tom Brady gets into the backseat of a Tesla, looks into the camera and says, "Let's f*cking go." And then the Tesla pulls out of a parking spot and drives off *
You underestimate the power of TikTok. Thousands of "WHEN THE #TESLA #DRIVES #ITSELF #FSD" videos going viral all at once, etc.
They will realize it when their insurance company charges them more for a purely manual car.
@@jasonwatkins7767 ...drives itself directly under a semi...
Hertz is bankrupt
Pretty much everything listed in the CNBC piece is from "Guide House Consulting firm". Like most consulting firms they don't do anything for free which means someone paid them to make up that crap and have CNBC air it.
Its cnbc you trust them good for you as they are putzes about technology
How shocked will they be when Tesla surges ahead and unexpectedly takes the lead.😱
Every time I see a version of that graph I laugh. Regardless of whether your think the technology works, to say that Tesla’s strategy is the worst either shows your bias, or that you don’t understand at all what is happening. (As explained nicely in this video)
It unquestionably shows either gross ignorance or more likely, purposeful dishonesty to bias potential buyers. Disgraceful.
I call it the gaslighting graph
I simultaneously laugh hysterically and cry out of deep despair when I see it - both because it's hilarious, and also really sad that so many people buy the FUD.
Oh well, fundamentals will win in the long run, and it'll be obvious in hindsight that everyone's been barking up the wrong tree (e.g. liderp) and that nobody's ever had a chance of beating Tesla if everyone else is doing it fundamentally wrong.
Let me lose their money 😂😂
Robotaxy business model has very small advantage over street rental business model like Bolt Drive service. Bolt drive does about 6 rides/day robotaxy maybe 8. So when you simulate that then those services are very similarandboth have strong points but i think 20% more of cheaper cars wins since it means 20% more rushhour rides. Also 20% less costly in day rides. So tesla without FSD earns only slightly less than with FSD on street rental use case when FSD costs about 7-8k then FSD starts having advantage.
Fatalities is an important metric, but interventions are what people perceive most. Within a short time of my daughter learning to drive, I did not need to intervene to not die. From this metric FSD seems like a very new driver. I am trusting it more, and intervening less with every update. I think it will be over the intervention hump soon.
Will FSD graduate from high school?
@@FrunkensteinVonZipperneck the dumbest people I know are college graduates
That is my hope, but as you said, not there yet.
I think it's also trusting it, making decisions you would not have taken. That it doesn't do things like you would, that it doesn't do them in a way you are comfortable with, doesn't mean it's wrong.
@@Nelis1324 You make a good point. There is that aspect of FSD doing something in a different but still correct way. I have experienced that feeling with FSD. The remaining problem is when it does something illegal or dangerous that could cause an accident.
I'd love to see Tesla's stats on how many 'dangerous' disengagements they get on a highway when doing under 40mph, while following a lead car, without changing lanes, on a clear day (no snow, rain, glare).
U already saw the number, well u saw no number, exactly
Doesn't happen.
❤️Hi..
glad to hear from you.
I appreciate your support
Where are you watching from…
My friend hit a cyclist on autopilot because the car software crashed didn't see it went straight into the back of the cyclist then carried on and hit a motorcycle before finally resting with a pedestrian on the bonnet. It's so advanced it's life-changing literally. Don't buy Tesla it will never be finished❤.
I nodded out while driving in stop and go traffic in Miami Beach a few months ago. I was driving a rented Model Y from Hertz. The car was in autopilot. After a few seconds I BOLTED UP… wide awake. The car had kept me in the right lane.. in an appropriate distance from the car in front of me!! If autopilot had not been engaged…. No doubt I would have been in an accident !! TY Tesla.
Your mind was responding to Florageezermosis, which sedates visitors into a snooze after Florageezers finish SeniorSupper at 4:30pm.
That’s not a feature unique to Tesla though. Lots of cars have lane keep and traffic aware cruise control. Also, with the cabin camera in Teslas they have started watching your eyes and if they see you’re ignoring the road it’ll warn you before disengaging autopilot so you could’ve been woken up by cars honking at you to move instead 😂
You wouldn't have nodded out if you actually were driving yourself - or are this negligent of a driver? Jk, but not really
@@braindecay9477 JK you just sound very uneducated when it comes to driving. Nodding off is common in places where people work long shifts and have long commutes that invole driving on highways, especially at night
How much did the self driving features of the car contribute to you getting extremely bored and nodding off?
And what about the next time? Now you trust that the car will keep you from crashing so, you "rest your eyes for a moment" more likely because what could go wrong?
Until it goes wrong.
FYI, Out Of Spec did a video about 3 weeks ago showing a driverless Waymo trip in Phoenix in a Pacifica mini van. It was worth a view. In the Phoenix area they do not have drivers in the geo fenced area.
Waymo is similar to a driverless subway train. Don't need a driver on rails.
You are SO RIGHT, and share the frustration with you about the poorly informed and un-educated in formation of the media. KEEP the videos and info coming.
The big difference between all those other competitors is they use HD maps, they cannot go ANYWHERE. This is the big difference and what amazes about Tesla. They apply the concept to any road in almost any condition. When you show me another one of those competitors who can do this at 85mph then i'll eat my own hat
Your analysis is exactly why FSD has already failed. Tesla made its FSD to have no redundancy. It is nothing more than a gimmick that will most likely never be approved for level 3 let alone level 4. Mobileye and Cruise are so far ahead of Tesla in terms of strategy and execution. Mobileye is so close to profitability and has an excellent business model. Mobileye’s SuperVision platform is everything that FSD wishes it could be. The combination of cameras, LiDAR, and HD maps makes these businesses the most sustainable in the long run. I think Tesla could pull a 180 in strategy, but Musk is kind of an idiot and has repeatedly stated he doesn’t believe in the necessary redundancy systems.
When MSM says product is bad, its must be really good.
As a Tesla fan, the major problem Tesla fans and videos is trying to deny the fact that Waymo not only commercialized in 2020, but operates and is expanding - now in both Phoenix and San Francisco. I have used the service in Phoenix, being the only human in the vehicle sitting in the back, and it worked very well, dropping me off in the parking lot of Starbucks, my destination, and then driving away. Saying Tesla will "soon" (I was convinced it was happening in 2018, 5 years ago, and many were convinced years before that, but we're still waiting) be able to instantly drive everywhere / entire states, etc legally with no human in the car, instead of more likely taking similar baby steps like Waymo, seems like we're all in denial.
-Talking about remote drivers - why doesn't Tesla use this as well to launch self driving now? Because they are lagging behind in the PR race with Waymo literally years ahead now.
indeed. One is a commercial solution, the other is basically still L4.
For Waymo, the big question is whether they can ever make it scalable and profitable.
For Tesla, the setup is already scaling and is profitable. The big question for them, whether they can ever make it safe enough to be used as a robotaxi with no one behind it.
I think what is confusing for analysts is that for Tesla, there is no financial "push" to rush out L4 with crutches (backup driver, hi-def mapping, geofencing), as their current business model is already profitable. They work on the foundation model to make it as robust as possible.
While for others, their whole business model is about providing fsd as a service. Cruise cannot afford not to provide it, even if it is currently still not.profitable.
The problem with Waymo and other level 4 cars is the large cameras on top of the car that use lidar and other tools to detect when there's a car or person and measure their velocity and direction of movement their not integrated into the car or anything and also they can't understand visual ques given by humans which is what Tesla want's to achieve with their FSD Ai so it can pick up peoples hand movement gestures and react in a way that a human driver would and be able to work anywhere in the world with different traffic laws they want to make an AI driver pretty much that feels natural and is extremely safe.
The graph is based on the number of cars they currently have on the road without drivers. It ignores the fact that the other approaches won't scale. Tesla's approach is the only one which could lead to full general autonomy.
I watched that video and I thought that comment itself was a joke ... I'm not an expert but I saw Waymo testing around my neighborhood all the time, with tester sitting inside, and I saw lots of Tesla driving around with normal owner (including myself) so who software will require less "supervision" eventually is rather obvious ... their systems simply grow up with tons of supervision.
Elon Musk has stated that his goal for going full Autonomy is not "better than a human driver", but rather "10X lower fatality rate than a human driver".
YUP, and I think stats show it is already 8-9 times better than USA average on a per mile basis!
@@mohammadwasilliterate8037
That's Tesla's self driving tech (FSD or Autopilot) PLUS A HUMAN DRIVER, not Tesla self driving alone.
That's not enough though. If you have many kids speed fender benders then it's still no good.
@@castortoutnu I would agree in theory, however in practice, I thought that avoiding fender benders was actually easier than avoiding fatalities in the sense that fender benders come from known (solved) issues and fatalities come from the truly unexpected.
@@ParaSpite Your logic doesn't make much sense. If a driver normally make 10 crashes, but reduce to 1 with fsd, obviously fsd avoids 9 crashes.
is there a link to the original video?
Great video. You really know how to explain ai well and should prioritize this domain :)
I'd disagree with one part. You can code probabilistic decisions in procedural code. My guess is that replacing that with neural nets is more about handling unexpected situations and easier code maintenance.
How do you maintain code you do not understand? Neural nets "code" is inherently undecipherable for humans.
Level 3 and 4 are at least a decade away. Rain, snow, poor road maintenance, and a ton of other variables will keep pushing it out.
I drive Tesla FSD and progress is slow. It’ll run over speed bumps, crash thru gates, brake late on curves when traffic ahead, miss lane changes, etc. few years away from next level
Anyone who thinks FSD will never happen or take decades, has not tried GTP-4. AI goes from a baby to super human, when it suddenly hits a tipping point.
GPT
Nope. If That Were True It Already Would Of Happened .
Two different things bud. Chat GPT is not "true" intelligence. Its a language learning model. A language model that cant convincingly tell you completely false lonely information because It doesnt actually think for its self lol. Not to mention self driving has peoples lives at risk.
@@Gcanno You clearly know nothing about machine learning. Tesla is still gathering the massive amount of data needed to train the neural net. You didn't think it made sense for AI to solve chess before Go?
@@VictorDelPrete Correct, that was a typo.
It all comes back to the talent of the team and the commitment of the business to its mission. The mission has never wavered, and top talent appears to be continuing to line up to be a a part of it all. If the talent sticks around, they will find a way through each of the challenges, turning the impossible into late. Seems like a 90%+ probability right now. Once the exponential curves of the NNs start to accelerate, they will likely repeatedly surprise to the upside. There's a stubborn threshold we still have to get over, but then it's gonna get crazy good. Then insane, ludicrous, and ultimately plaid...
"turning the impossible into late", really like that :)
I agree, although I do wonder what will the AI do if the cameras get blocked by snow or mud or what ever?
A human driver can step out of the car and wipe the windshield if she can't see, but an AI can't really do anything.
@@toffotin The front cameras do automatically clean themselves with the windshield wipers and spray. The side ones and rear are more exposed to failure, but I think they are a little less likely to get blocked in the first place due to the angles -- but if they do, the vehicle seems generally willing to drive with 1 or 2 cameras down. Additionally, they are training the neural nets to ignore/see through a certain amount of occlusion. That said, it it a real problem without a perfect solution. Maybe when the robotaxi network is up and running, owners/fleet managers will need to carefully clean them every day to reduce the likelihood of blockage. I also wonder if heat can be used to melt snow and ice around the cameras.
Plaid isn't the right term to use
@@toffotin Yes it can and it is obvious what it can do, ask you to get out and clean the lens.
I actually had a driver look at my Tesla Model Y and say 'Nice job' as I paused and did not block an exit at a strip mall for exiting drivers. Car was not in any auto mode but other drivers must just assume all Teslas are.
Wow, I was just thinking that neural nets really benefit from having a lot of data, but Tesla still has a lot of hard coded parts of the control system.
There was an expert explaining how you can't just throw in a NN, but when Tesla does make enough progress to incorporate neural nets throughout then the improvements would really accelerate.
“expert” also said you can’t land an orbit rocket. Ignore the “experts” if they don’t work at Tesla.
Great video! It will be awesome to see Tesla deploying FSD with one press of the button.
Corporate media provide broad but shallow coverage of the world we live in. These organizations are prime candidates to be replaced by AI.
@Are Gee Bee Perhaps. Too bad the victims wouldn't know the difference.
I've got a model 3 with fsd, and everytime I try it out I'm disappointed, (although less disappointed with each update). If feels like a student driver. Driving to the mall, going to turn left about a mile ahead. It get's into the right lane. :/ Then at the last minute it crosses two lanes to get to the turning lane. It signaled, but I would rather have some better planning. Then turning left, following a car, the we were following waited for a car to drive past and then went. My model 3 moved up into the middle of the intersection and waited. I don't know why. Then going through roundabouts it's like it's never seen one before. It's got a long way to go.
Unfortunately what I see in the Tesla FSD community is the community only intensely follows what Tesla is doing. There seems to be a lack of awareness of the capability of what Waymo and Cruise are doing.
It's understandable but I think it causes the Tesla community to discount these other companies technology and overestimate the ability of FSD.
I'm not saying FSD is behind, but I'd suggest judging for yourself. There are videos out there of these other companies showing what their technology can do in some complex situations without a driver behind the wheel. I get that it's geo fenced, but there is no where you can drive a Tesla after all these years without a backup driver. You'd think at some point Tesla would at least be able to geo fenced some area, or at least begin to work with regulators to open up that option.
Also, the description of how these other companies have developed their software is incorrect in this video. You can search and find lead developers of these other systems talk about their implementation. They are not hard coding decisions in if then statements to implement their software.
I'm impressed with his presentation. Throughout the video, he's keeping his cool while providing a counter-argument. Me? The moment I heard that statement from the CNBC guest, I lost it.
AI is way smarter than these CNBC "guests."
Dr KNIA question: do you know if Tesla is going to use multi modes of data gathering in future vehicles to achieve the “vector space” required for the necessary “situational awareness for the calculations required for FSD? I understand Elon M’s tweet about using NN to go from image space to vector space but wonder if just visual data is enough. I live in Canada and have driven in some dodgy white out conditions in winter and at night. If LIDAR or RADAR are integrated in addition to cameras isn’t that a more efficient way to 3D vector map surroundings?
Your description is filled with links, hard to find the link to the video!
IIRC, a while back this channel said something about not being surprised of Tesla achieved FSD by the end of the year (whichever year that was). If you're a software engineer you know that the last 5% and last 1% are the hardest. I've been saying we're not close to FSD and won't be for quite a while at least. Even though it might look like it. Being able to drive autonomously 99% of the time does not mean you're 99% of the way there. It's still a _long_ way off.
I think a better target for the near term instead of full FSD would be if the AI was geared toward avoiding accidents. The AI can make quicker decisions and doesn't panic. So the driver would still drive normally as always but the AI kicks in when an accident is imminent and steers the car away from trouble, or brakes, or whatever avoids the accident. Computers and humans compliment each other. Computers are good at catching human mistakes humans are good at catching computer mistakes.
As a software guy who lets his Tesla drive 99.9% of the time in FSD mode, it's hard to be confident what a percentage means. In my case, I'm speaking in terms of minutes I don't spend driving. A lot of it is subjective too, but your numbers are way off if this percent is the baseline of what is considered safe.
So while yes, I agree, the hardest software challenges will be closer to the closer to end side, there may never be an end, just better and better tech. The real challenge I think Tesla will run into will come from its limitations of the camera to see details, and that will take better hardware to overcome.
@@MrElektrokution Depends where you demand that FSD should be working. We have a bunch of roads where Tesla will not be able to drive autonomously without braking at least 5 laws. They do not look out for the Police, which you must do before driving those roads correctly without endangering anything including yourself. Those situations are such extreme edge cases that will never be handled by automatons.
99% of the rest of the driveable roads should be possible to drive autonomously this decade. Everywhere. Even in Siena or Calcutta.
Would be great that the upcoming roadster have self driving track mode at some point. Basically letting the car get you around a track at near maximum speed capacity with perfect trajectories, bracking points and so on. Just imagine the ride, it would be tremendously scary and exhilarating at the same time.
imagine if it was a well known track with a history of laps taken by certain racers. Imagine having FSD take you around Laguna-Seca at a record-setting pace by Lewis Hamilton using his lines.
@@pierrenorman421 Forget Laguna Seca, how about a sub 7 minute run around the Nurburgring. A dry cleaner could make a fortune at the finish line.
James Bond mode.
search for BMW M5 with Clarkson.
I believe this would be called Level 6, putting Transformers at Level 7.
Shareholders should launch a class action against the media for mis information…..win, lose or draw….it should restore transparent honesty!
I live in Marietta & a Life long Marietta resident. I watch you for your content and you know about Ga chargers & issues here.
Elon hit the nail on the head when he said that FSD is not a sensor problem, it's an AI problem. We need to remember that a human driver has just 2 cameras, 2 microphones - and very powerful intelligence.
While true, humans would be much better drivers if we had evolved biological radar.
@@Juttutin like bats.
Not sure cameras only will give a car depth perception. The added USS would be better.
The difference is "Human Cameras" can adjust the angle in real time to see 360 degrees (when using mirrors). Teslas cameras still have blind spots. Tesla cameras can also not clean themselves which is a huge issue. I have camera issues in low light and anytime its muddy, snowing, raining, etc.
That being said I still use FSD almost every day and love it!
It costs less and is far more reliable to use multiple cameras than to put 2 cameras on a neck that can turn and use multiple mirrors. Tesla needs to get rid of the blind spots and needs binocular vision or >180 radar at the front of the vehicles. Obstructions and secondary T-bone crashes from the left are inherent to driving on the right. The real fix would be car to car communication. Then, only user errors would cause most collisions.
Tesla FSD is a freakn Joke... except its already drives me to work and home no problem. :-) Thanks Dr. for putting in the time to explain your perspective. So interesting and obviously super exciting to see what happens.
It doesnt though. Your hands are on the wheel, feet on the pedals and you are paying attention to the road at all times, right? You are driving the car.
I take a freakn joke that saves the lives of my relatives any day over the "normal" from anyone else. And twice on Sundays!
@@Withnail1969 If the Tesla see a need to decide: break or turn the wheels and then break, it will stop in less than a 1 second. Your reaction time 1S ++ decision time,,, action 1s ++!
Thanks for this video. When I saw the CNBC video, I rolled my eyes so hard. They have no clue what they're talking about. And they don't describe the "lack of business model" with any facts, data or information.
How do Cruise and Waymo get regulatory approval? What kind of data do they have to submit?
Great video! Thank you for countering all the self-serving FUD in the mainstream media. I'm planning to subscribe to FSD for my Model Y in a few weeks when we make our annual trip south to Boca from the Philly area. I can't wait to experience it first hand!
I suggest getting practice with first.
Yes, highly recommend getting plenty of FSD Beta seat time before your long road trip. That way you have a good idea of how it operates and what to expect. Last thing you want is have unrealistic expectations during your trip, get booted from the beta for doing the wrong things, and having an even worse experience. Most people that give FSD or even autopilot the worst reviews are people who come into it thinking they just turn self driving on and pretty much go to sleep or not pay attention. Then they hear about GM’s super cruise and it being “hands free”. So they think that’s better than FSD/autopilot
@@FlipBoxStudio Thank you for the great advice! The other issue I'll have to contend with is possibly not being able to use V11.3.2 because I currently have V 2023.2.12 instead of 2022.x.x
What most don't get, is the fact that Tesla will not move from Level 2 to Level 3, and then Level 4, and then Level 5. Tesla will jump straight from Level 2 to Level 5. Overnight they will move seemingly from last place to first.
Really good point! I totally agree. Hope we are right.
@@impc8347 that will absolutely not happen with present hardware. I wish it could be true but we will not be seeing level 4 anytime soon. Much, much more work to be done and more hardware advancements or implementations.
My 2018 Mercedes already does traffic jam self driving really well. It does fine on open highways and the Interstate as well, but it is considered a driver aid, not self driving. It lane keeps fine provided the road markings are visible, but you do need to pay attention to what's going on in case it decides to wander. It requires you touch the wheel every 30 to 45 seconds or it will alert you. On a clear piece of Interstate with no traffic around I went ahead an ignored the prompts to fondle the wheel and after several ignored prompts the car rapidly braked to a halt in the lane.
Overall, I'm OK with the implementation as it does reduce driver fatigue on long drives and allows you to have a bit of a sandwich or a drink without keeping a hand on the wheel. The adaptive cruise control will follow traffic from highway speed to a stop and then restart again within limits. In some cases you have to give the accelerator peddle a tap to start the car rolling again.
However, it will not detect a stopped car in your lane in time to stop when traveling at highway speeds - works up to about 35 mi/h but above that speed differential it won't start braking in time to avoid a collision with the stopped car.
But then again, Mercedes never billed this as full self driving and lists the restrictions. Accordingly, there haven't been loads of Mercedes fan boys and girls in the news killing themselves while taking a nap in the back seat or whatever like the Tesla (non)drivers.
Great vid...question. If I subscribe for the $199 Full FSD on my 2022 Model 3 Long Range, will I get to experience this? Reason I ask is I would like to test it out myself and determine how fill of poop these journalists are (lol). Thank you.
Yes, but... Wait until version 11 of FSD goes wide to all owners. I think they are holding release to new customers until then.
YES, the $200 subscription unlocks the full FSD suite.
Version 11.3.3 is amazing. It takes me for long drives.
Last night, I got the latest software code from Telsa 2022.45.11 is loading now. Yes I got FSD and I use it all the time, And its working pretty damn good. This evening I will be testing the new software 2022.45.11… Im doing my part to push Beta in every way…. You fools TESLA IS SO SO CLOSE ITS UNBELIEVABLE!!!
!
My question is if you have a 2018-2020 model 3. Will it work with fsd or will the new cameras on the new model take over and the 2018-2020 would be out of date and unusable for fsd
Does anybody know about dojo? When is that going to be online and connected to the fleet?
Another great video, thank you! Like you, I can't wait to get the new update and see what strides the Tesla team has taken!
Couldn't agree more with your analysis. When I started to understand this on Tesla's AI day it blew my mind. It's a completely different approach to autonomous and understanding the world. Tesla is truly playing a different game from the rest of the competition.
Anyone that tried this would know Tesla is the only game in town.
I saw a podcast with Elon stating the money that will come from autonomous driving will be bigger than the whole car industry. Becasue once they master it with cars the technology will then be used for robotics.
You believe Elon? Didn’t he say FSD should’ve been available years ago? Such a scam. FSD is nowhere close to being truly FSD.
Very informative and interesting!
Thanks Doctor
Just wondering what your take would be on the Ga EV tax that is paid annually when the auto license comes due. I've been paying that $200+ fee since about 2016. I'm retired, drive my Tesla around 10K per year. I get what the tax is for; I don't buy gas therefore not paying fuel tax like ICE cars, however is every EV owner in GA paying this same amount? It doesn't seem right if so; either I'm paying too much, or other EV owners or probably paying too little depending on annual mileage.
Well, a company in the U.K. is working with cameras, radar and AI as well. Rather successfully so far, I might add. Yes, they are tiny by comparison and don't have a massive fleet out on the roads yet. This may well change in time by cooperation with big manufacturers. And of course, they came up with their own ways amassing huge amounts of data to feed the AI. Time will tell how far their excellence will carry them.
Are you referring to Wayve AI based in London? Wayve is using a generalized AI approach to autonomous driving and the YT videos I have seen of their cars driving in some difficult situations in what appears to be a city in the UK.
Wamo is also geofenced. Tesla's motivation is based on philosophical and philantropic; profit is just a tool to get to the objectives. That is a big difference that affects the stamina of Tesla via the employees motivation.
Do you consider suggesting Randy Kirk to upgrade his camera to as HD as yours?
FWIW I see Waymo vehicles operating in my area on a daily basis in the west side of Los Angeles, and every time I am able to observe the driver, a) there is a driver (it's not driverless) and b) that driver is indeed directly operating the vehicle. I imagine they are just gathering data for the neural nets, but I'd not count those as driverless miles.
Large Auto: We have the best engineers of the world. If we can‘t solve autonomous driving, no one else can.
Elon: Hold my beer…
I know a few cruise employees. They really do believe that Tesla FSD is "a joke" internally. They recognize/concede that if Tesla can operate without 3D maps and lidar, that their product would be non-competitive. They just don't believe Tesla will ever reach the necessary miles per disengagement targets. The waive away FSD progress as "too large a gap to close". I think they're suffering from massive confirmation bias.
And trying to keep their jobs as lifeguards on a sinking ship.
I don’t know if you guys remember the 2005 DARPA Grand Challenge. If my memory is correct, MIT hardcoded and Stanford use a NN.
John, Please respond to this comment.You inferred in this video that once FSD is accomplished, anyone who had purchased FSD could benefit by converting their vehicle to a robotaxi. That's what I had planned for my 2020 Model 3 which has hardware 3, however, now I understand it must have hardware 4 to qualify. I'm also at the point where I would like to upgrade to a newer model. Since I have received no real benefit over the past three yrs. of FSD ownership, it only seems fair that my FSD should be transferable to the newer Tesla. Can you stand behind me on this conclusion?
Just buy a Tesla
Done! (2023 Model Y) 👍
Ya
@@usahome smart, also the most popular
❤️Hi..
glad to hear from you.
I appreciate your support
Where are you watching from
@@Elommuskdz9ve Portsmouth, UK. Tesla Stock owner Great Video.
You only need to watch the FSD videos by Whole Mars, John Locke and Chuck Cook to see how amazing FSD Beta already is. Tesla’ s fan community is the tesla PR department. CNBC should see the fan community mode videos in order to see that Tesla should take the MobilEye position of that infamous chart. Upside though is that The greater surprise it will be for these companies when they suddenly get shoved aside by FSD and people will buy it for the advantages it has.
Congratulations on the veery good flow of the speech - IMO the message gains much when avoiding filler words, and is left yo just flow. Maybe slowing down pace a tiny bit would help further more efficiently passing the great, invaluable and relevant info, insight and options that you got us used to.
Thank you for all the great work and value you deliver here in YT!
Karpathy (before leaving) shared that they were going to drop the use of the radar in favor of a camera-only system.
You know, one of the big reasons that driving autonomy based on vision is such a big thing for Tesla might maybe... not be related to Tesla at all...: Imagine a robot vehicle (Mars Rover) riding around on Mars: at this moment, ANY robot needs a carefully planned path, going slow to allow it to be controlled from here. Bring in self driving from Tesla.... the robot does not need a litterally remote driver on another planet: you basically give it a mission to search for something (will be able to, as it can actually recognize things) and it can start driving around. On its own. An call back home when it found whatever you need it to find... Maybe that is also part of the business model... I think it's best not to see all of Elon's project as seperate things: Tesla - Electric cars => need it on other planet, need it here on earth. Boring company: solve traffic problem here, also solve a big puzzle on another planet when you want to colonize it (impossible and dangerous to construct things on the surface), Space X: doesn't need any explanation. Battery solotions: need it here to solve energy problem, need it on other planet to power things. Giga factories and die casting: need this technology here for cars, may need it on other planet to efficiently build things... I could go on like that for an hour. All this guy does is related and one thing is part of the puzzle to solve another thing. It's all part of his vision and he did put his mind on colonizing Mars. The tech he's developing here: he'll use it to get there. And that might be much sooner than we think 🙂
If tesla route planner picked routes that avoided places where disengagements have happened, how much more reliable do you think FSD could be? at what average % cost to length of trip?.
Lane-specific-geo-fencing or Lowest-Risk-Route-Planning is my prediction for how Tesla will launch L4 autonomy.
Disengagements are a good thing. Learn from your mistakes. Everytime you you report one your showing the team where it needs work
@@saff226 yes super valuable in L2 driver assist mode. Just wondering how much the nav would change if it tried to plan a L4 autonomous trip avoiding historic spots (they've already recorded) that aren't solved yet.
That chart has been around for a couple years! Why not create a new one to counter it based on number of NNs and miles traveled! I will help promote it!
What do you think about nvidia drive? Don't they have a similar approach?
these 'experts' do understand they are just being paid enough to tell it as a lie.
You and others in the Tesla know need to start referring to Waymo and other similar companies as Fully Autonomous ‘Trolleys’ and explain how Trollies 2.0 are restricted to a set of carefully mapped roads in a small area around which those 2.0 Trollies can drive without a driver - like original Trollies could on local rail lines on specific roads. Only that messaging will help the masses understand the difference between them and Tesla’s FSD.
Nice job. The difference between the neural network probability approach VS hard coding is HUGE!!
If there was remote driver, why did they block emergency services and not know what to do? I call bs
Is the audio misaligned or is that just my player failing? Can't get it to sync.
I think they said the same thing about a couple brothers a 100 years ago from North Carolina. Giant leaps in technology never happens in a few years it takes decades.
Great breakdown. On the one hand it's lame to have to refute these silly graphs, but on the other hand it's another chance to refine our understanding of the strategy and consider different possibilities.
The graphs are correct. Tesla FSD is garbage. Mercedes is way way way ahead.
I was going to make comments about FSD and NNs, but you covered them before the video ended. Thank you.
can you please put links to stuff from the video at the beginning of the description rather than the end. i thought you'd forgot to add it until i scrolled all the way to the very end and saw it. you aren't a grocery store, just put the link where it makes sense please :)
im suprised they did not put GM at the top of the list... You did it Mary!
Not even mention of Open Pilot? They have more than 40 million miles with the system actually driving. Many people claim it is better than the Tesla AP.
I own a Tesla Model Y with FSD. I drove 4.2 miles to Walmart to pick up groceries. I had 5 interventions, 3 phaeton turn signals, 2 times car stopped 1/2 block away from the next turn and the biggest intervention was when the car was making a right turn onto highway 10 east bound. It pulled out in front of on coming traffic traveling at 65 MPH. Thank God I intervened. I watch many videos of FSD Teslas and am wondering how do they get their FSD to perform that good. Or maybe all the FSD people like me are not reporting the interventions they have to constantly make? On top of that I constantly get the warning the “FSD is not available” or “FSD is degraded” FSD is still years away from being ready for prime time. BTW I live just west of Minneapolis Minnesota in Elk River.
As an executive that worked at the so-called leaders, my Tesla fsd is way ahead of 2nd place.
I have FSD visualization. I saw a large orange container parked on the side of a road. FSD could see the orange cones around it, but not the container itself. 😅
Does it need to see the container? Once it’s noticed the cones it excludes the space as somewhere it shouldn’t drive and moves on to spend cpu cycles on recognising something else…
@@gothmog2441 Well, containers normally come without cones…
My point is once it decided that space was bounded by cones (don’t drive here) FSD doesn’t need to parse what is within the space - that’s wasted cycles. If the cones hadn’t been present, it might well have parsed that there was a large orange container (don’t drive here either).
You have observed FSD apparently not recognising a large object that it didn’t need to recognise to drive safely. It’s not a person: it has no interest in the container, merely in the boundary of the space within which it can safely drive. Which was defined by cones, which it did recognise. So … might just be working as planned?
@@gothmog2441 Yours is an assumption, we don’t know cpu time save was at play there, or it simply missed the elephant in the room.
Always great.
Thanks ~
I like your thinking doc. But the clincher for my subscription is the surfboard on the wall. I hope you have time to use one. 👏
George Hotz has been saying that NNs all the way down was the way to go for a while now. On a related note, I'm surprised 'comma' didn't even make it to that dumb chart.
I am confused about what you said, regarding all of the electronics for FSD. I just picked up my car yesterday, my third model, S, and I believe that it only has seven cameras and no other hardware for FSD in terms of radar and LiDAR. In fact, one of the seven cameras is the internal camera Pointing into the cabin.
I really don't see full autonomy for several years. While I don't have V11 yet, there are so many times the car freaks out. For example, apart from bad weather, it has no idea what to do on a two lane road with a lane closure and a worker turning a stop sign to slow in order to cross a double yellow line and proceed into the oncoming lane.
When your Tesla "freaks out," do you sedate it with valium?
Try 11 and you might rethink that. Tesla is making exponential improvements
I agree, but these are edge cases. However, FSD needs to be able to handle these situations. And that will take time. Going from 99% to 100% will take more time than go from 0 to 99%.
As soon as they merge to master, there will be a 10x increase in data collection: 400k -> 4M+ vehicles vacuuming up the roadways.
PHOTONS !!! It takes in photons, not pixels. It turns photons into bit strings into a video game that plays itself.
I did 380 miles last weekend - and about 95% of that was on autopilot.
Meanwhile human drivers kill around 1.5 MILLION people a year.
Excellent info on a topic you clearly know and love. Well researched and enjoyable FUD-busting. Thank you!
John, Please respond to a good ole Georgia boy. This is my second comment concerning a problem that will affect many Tesla owners who have purchased FSD. You mentioned again in this video that we would be able to convert our existing Tesla into a robotaxi when FSD was finally approved. That would be great. I could convert my old Tesla into a taxi, make excellent money, and purchase a new Tesla for my own personal use. I hope that this will be the case, however, I have heard that only hardware four would be allowed to function as a robotaxi. Now, if that is true, doesn't it make sense that Tesla should allow FSD to be transferable to the new Tesla?
Whenever I watch a couple of Tesla's self park side by side on Cape Canaveral I wonder how close Waymo is to doing that. As far as I know Waymo has yet to achieve lift-off.
I have a couple questions about using neural networks for this purpose. I'm a software engineer but I don't know a lot about NN except the basics.
First, NNs depend on the input data to set up the weights between nodes (the training) and then output a result. They do not use logic like typical software programs do. So how can you know that the output is what you want? It seems to me that the only way is to test. There's no way to reason that the output is correct for sure. And as software engineers we typically like to say that we should not "test in" quality. We should build in quality.
The second question is that if the NN can only deal with situations that were included in the input data then it possibly cannot deal with new situations like humans can, true? Won't this mean that FSD will never get totally there because a NN can never deal with new situations?
Nn are capturing a model not a 1 to 1 logical process for input x equals y. They also simulate data to test with. Like masking an input or adding noise. The weights capture the model.
Look the last Video of Dirty Tesla Channel. There you see, how bad the FSD is doing after some updates this year. In common situations it was getting worse. In my software, before release, I run many tests to find out new bug, which I made. My tests are also based on bugs from the past. However, it looks like Tesla dropped FSD and don't improve it. It looks like it degrades. And funny is the windshield wipers. They are not able to make a proper mod for this 😱🤣. How should they make an FSD?
I think the graph is from a "research firm" JLG and other FUDsters cite. Can't remember the name right now.
Control ANNs are key. The will be Bayesian Nets. No more classical C core code for a Rule based engine, but a prediction error experience optimizer which in effect is a low level PFC. The second Trillion is mine 😂
Thank you very good update.
The difference between multi million dollar major news media outlets and a guy in his garage. Thank you.
I keep watching fsd videos and there are still way too many interruptions to be able to call it autonomous. There will have to be hundreds maybe thousands of non interrupted drives before it can safely be considered autonomous.
So, a single day then?
I challenge Waymo and Cruise to come drive me to work like my Tesla Model Y does. I will mention that I live in Paso Robles, CA.
You may be a little too optimistic. I predict H3 will not be able to copy with - possibly - L4 and, for sure, L5.
H4 is more likely due to high camera resolution and faster CPU. The camera resolution is a big issue when trying to see from a long distance read red/green turn arrow in traffic lights (AU & EU case - not sure about US)