In this video, @RebelEngineer lays out why Tesla Robotaxi is dead on arrival, if it ever launches at all on Oct 10th, 2024. He exposes the stupid technical failures and poor business strategies that have caused and will cause endless Full Self-Driving delays. The truth is clear – Tesla Robotaxi is the worst AI gamble in the universe.
Chapters:
00:00 Introduction
00:24 Never-Ending Computing & Electricity Expenses
01:34 Ride-Hailing Is Ultra Low-Margin
02:12 Eliminating LiDAR – Business and Politics Failure
07:43 Why FSD Will Never Achieve 100% Drives Without Disengagement
09:32 Increasingly Noisier Data Causes Multiplicative Edge Cases
10:18 FSD Is At Best An Average Tesla Driver
12:24 Waymo’s Goal: Batman in Batmobile / James Bond In Ferrari
14:15 LiDAR Reduces Margins of Errors
15:10 Cameras-Only Is For Losers
16:07 One Accident = Entire Fleet Shutdown (Sh*t Margin)
20:14 No LiDAR + No Sensors = No Profit
22:00 Waymo Sensors Measure, Tesla Cameras Only Estimate
23:21 FSD’s Never-Ending Mistakes
24:41 Tesla Robotaxi Event – The Stock Pumping Scam
26:33 Tesla FSD Data Totally Depends On… Its Car Sales
29:58 BYD, Baidu & Xiaomi Will Destroy Tesla FSD Margins
38:32 Tesla Tailors FSD for YouTubers To Trick The Public
39:23 Tesla’s Top-Down Management And Human Safety Disasters
40:14 Tesla Doesn’t Even Have A Robotaxi License Anywhere
43:31 Tesla Robotaxi Is Dead On Arrival (If It Arrives At All)
44:46 Tesla Robotaxi Event – The Nothing Burger
This video is NOT sponsored by any vendors in the affiliate links down below!
MY FAVORITE PROGRAMMING GEARS:
MONITOR: LG 43UN700 – https://amzn.to/3HUAlDU
MONITOR ARM: VIVO Aluminum Ultrawide Monitor Stand – https://amzn.to/3HXUybT
KEYBOARD: HHKB Pro Hybrid Type-S – https://amzn.to/3I05GFo
SUPPORT ME:
Please SHARE my videos with family and friends. Or consider becoming a Patreon member to request specific video topic to be made:
Patreon: https://www.patreon.com/rebelengineer/
SOCIAL MEDIA:
TikTok: https://www.tiktok.com/@rebelengineerrc
Twitter: https://twitter.com/rebelengineerrc
Instagram: https://www.instagram.com/rebelengineerrc/
QUESTION:
Have a question? Post in the comment section below.
THIS VIDEO WAS EDITED WITH:
Adobe Premiere Pro: https://amzn.to/3K61XZq
Adobe Audition: https://amzn.to/3Ea5oL0
Adobe Photoshop: https://amzn.to/3E2D2SP
DISCLAIMER:
This video and description contain affiliate links from Amazon or another affiliate program, which means that if you click on one of the product links, I might receive a small commission depending on the country you are in.
If you watched the whole video and liked it, then
LIKE SUBSCRIBE AND CHECK OUT SOME OF THE OTHER VIDEOS ON REBEL ENGINEER CHANNEL ABOUT TECHNOLOGY NEWS AND TECH INTERVIEW TIPS.
#rebelengineer #ai #jobs #economy #tesla #tech #musk #elonmusk
source
What are your thoughts on the Tesla Robotaxi Event on Oct 10th, 2024? Do you think it's going to be a game-changer or a flop? Drop your opinions below! 🚗💬
My Predictions for Tesla Robotaxi Event 10/10/2024 TODAY (It's 10:12AM PST). One (or ALL) of the following will happen:
1. Tesla will offer sensors / LiDAR to Robotaxi by default or as an option.
2. Tesla will make Tesla Owners "geofence" their own cars.
3. Either (2) and / or The Tesla app will suggest owners to only let the cars operate in routes that they have traveled with zero interventions for months.
4. "Cybercab" will have removable steering wheels so Tesla owners can train then geofence these themselves.
5. Tesla should have had a Robotaxi testing license somewhere.
Without any of these, Tesla Robotaxi Is Doomed / Dead On Arrival
*DISCLAIMER*: This video is not sponsored and not being paid for by any Robotaxi or LiDAR company. I am not working for any LiDAR or Robotaxi company.
Also, if there are any other topics you’d like me to dive into, let me know in the comments! I read through every suggestion and will cover as many as I can.
SUPPORT ME:
Please SHARE my videos with family and friends. Or consider becoming a Patreon member to request specific video topic to be made:
Patreon: https://www.patreon.com/rebelengineer/
SOCIAL MEDIA:
TikTok: https://www.tiktok.com/@rebelengineerrc
Twitter: https://twitter.com/rebelengineerrc
Instagram: https://www.instagram.com/rebelengineerrc/
QUESTION:
Have a question? Post in the comment section below.
THIS VIDEO WAS EDITED WITH:
Adobe Premiere Pro: https://amzn.to/3K61XZq
Adobe Audition: https://amzn.to/3Ea5oL0
Adobe Photoshop: https://amzn.to/3E2D2SP
DISCLAIMER:
This video and description contain affiliate links from Amazon or another affiliate program, which means that if you click on one of the product links, I might receive a small commission depending on the country you are in.
If you watched the whole video and liked it, then
LIKE SUBSCRIBE AND CHECK OUT SOME OF THE OTHER VIDEOS ON REBEL ENGINEER CHANNEL ABOUT TECHNOLOGY NEWS AND TECH INTERVIEW TIPS.
#rebelengineer #ai #jobs #economy #tesla #tech #musk #elonmusk
Simple fact: nobody knows what announcements Elon has on 10/10. Nobody knows what direction Tesla stock will go, no matter how much explanation/justification of bulls/bears opinion!
Clueless, guy. You smart than Elon.
Your Chanel is dead in irriving😂
You don't need lidar. Have you ever been in a Waymo? Sometimes the drive in circles and on the wrong side of the road.
Total BS. Lidar is not needed. Radar? Neither. You have two eyes and drive every day only looking what is in front of you, and sometines looking at the mirors. Tesla AI cars cost only $1200 per car. All cars are equiped with that,and they have 8 cameras looking in all directions 100% of the time. Data does not collect all the data. Only about 10% of "good" data.
You have no idea how the system works . Try renting a Tesla with the FSD supervised ver 12.5, and then you may talk with a hands on experience.
Data is based on human drivers but the system works many times faster than a human driver can.
You can not compare all FSD versions. Only those released since November 2023 up to now.
I really doubt you are an engineer. I am. More likely you are paid by Waymo.Tesla does not want to drive like a human. Tesla wants to drive better than 10,000 drivers, then more that 100,000 drivers, than more than 1,000,000 drivers.
As of now and is public ; Tesla has two additional version of FSD supervised. An accident will mean releasing a new version via OTA updates. Not expensive at all. One waymo accident and the hardware, and software had to be revised.
You know what would be better? A fucking train! Also Bikes for short journeys, trains, trams, trolley buses for middle & long journeys. All of which can be electric without batteries. Cars are for those in between trips that don't fit. And you don't need your car to drive for you. Huge waste of time. Are you going to do work in the car on your way to work?
You need to go back to school and learn something pay attention, thirsty
Wow… a half hour of nonsense.
RebelEngineer has many good very practical points. But Elon worshipers and their hate can be ignored. I have driven in San Francisco and seen Waymo does currently better than Tesla FSD. The practicalities of Robotaxi has many other problems when comparing it to Uber or Taxi experience – won't succeed and the least to cover the immense revenue/profit growth it would require to save TSLA. That's why I sold my multi-million TSLA position.
Like chess and go, compute will play against itself till it is infinity better than humans at driving.
This is a solution to a problem that does not exist, a business proposition that most car owners will likely not want, and an enterprise that no insurance company will cover.
I think the RoboTaxi will have Lidar. I know most don't expect it because of Elon's past comments, but without Lidar it's DOA imo.
The first mistake people make is to consider that Tesla have a RoboTaxi system and that it is a market leader in vehicle autonomy.
Waymo, Cruise and Baidu all have Robotaxi system and are two years or more ahead of Tesla and even those three well known companies have a number of less known competing Chinese Robotaxi systems that are operating today. Unless they have been hiding a level 4/5 system, Tesla have a level 2 driver assist system.
If the Tesla RoboTaxi launch day follows the company's playbook, there will be a lot of CGI, smoke and mirror, doctored video, underwhelming live demonstrations, hype, future promise, blatant lies and over promises.
Tesla's vast collection of live FSD data is a red herring. For a start its usefulness has been limited because it has relied on a team of human operators to review and label the data. The collected data will also have a shelf life, the accumulated data often cited includes the very first FSD releases, as soon as a new FSD is released, the previous data collected is of limited use.
If Robotaxis are to be a profitable business, it probably won't be achieved by Tesla.
There needs to be more of this. A someone with limited eyesight who cannot drive and has been waiting decades for an actual self driving car, I'm annoyed that Tesla is actually setting things backwards. As an engineer I can see precisely what is wrong with vision-only systems. You can only trust a neural net based recognition system so far. You need to back it up with reliable measurement. The "with human intervention" part is utterly reckless and unethical.
Subscribed just to soak in the pain from the fanboys 🙂
Daimler has the best system by far.it's not even close.
You say you have been an engineer at Tesla… but you also seem to hold a grudge against them.
Were you maybe fired?
Wooow, I think you hit the nail! Seeing all this comment! Offcoars Ponzi will tell you it's for next year 😂 Wonder if his Biden style walking robot will assist Ponzi on stage 😂😂😂
This is an excellent video👍👍👍👍
Thanks for confirming you’re an idiot.
You are so uninform about about the tech buddy. You sound really dumb. Lidar is not the answer.
This for this video, it's a good rundown of the diss-advantages of the Tesla and I respect your background. However I think it's worth pointing out a couple of key differences between the operating models of Tesla and Waymo (possibly others) is that Tesla has been operating as a for-profit company with a positive cashflow. This means it's had to monetise their R&D as best they can as they go. Waymo, by contrast doesn't need this, they have an endless source of funding from Google and can therefore follow the traditional R&D route of overengineering in the beginning and simplifying later (though they haven't really started simplifying yet). The knock on affect of this is that anything that interrupts Tesla cashflow, e.g. expensive or delaying sensors get from the standard build and the ultrasonic sensors are IMO case of this. I doubt they were expensive enough to warrant changing the build to remove, more likely the lack of them was holding up production and hence cashflow. So Tesla will ultimately compete with Waymo on cost, not on capability. So where Waymo has seen their risks in terms of system capability, Tesla has seen their risks more in terms of financial sustainability. This financial sustainability risk management carries over into other areas as well, and the end result is that while we know Tesla can make EVs profitably, I haven't seen any data that suggest anyone else can. I've seen the BYD might be profitable and we know that everyone else isn't. On paper, most legacy auto are going out of business and are unlikely to survive if the global economic trends continue.
As you say, continuing down the route of Lidar may have been preferable with vertical integration and cost reduction making the costs more palatable, but it may not have been financially realistic at the time these design decisions were made.
Good video
Holy mother of cherry picking data and making a Sunday.
Waymo losing $2bn a year and that´s on only 800 taxis, Taxi´s ringfenced to specific mapped area. Cost of lidar and radar sensors makes Waymo not profitable. After 5 years it has 2 main cities and 50 taxis only in LA. City maps have to be constantly updated.
When Musk said he was going to reuse rockets that launch, place his star link satellites in space and then return and land on the launch pad everyone just burst out laughing. Now they are not laughing. Recent versions of FSD have become a lot better since the ditched the hard code and switched to Neural nets and AI learning. He will use similar systems to train the robots.
Sorry, don't bet against Elon!
you literally dont' know what you are talking about
Recent releases of FSD have shown remarkable progress. What is amazing people is that testers are driving to a busy shopping center, park the car and go shopping. It´s a busy car park with lots of cars going in or out lots of people walking from or to their car with trolleys of goods and no actual pavements.
The owner and family come out of the center and it´s started raining so they summon the car. It´s starts it´s engine drives towards them avoids pedestrians, halts for other cars or drives around them and gets to the family and stops.
This version was only released very recently. People are amazed to see a Tesla driving in a very busy carpark with no-one inside the car. The car park has not been mapped, the car is an ordinarily Tesla model 3 or Y and has no Lidar or radar.
The car can park itself in the car park and drive out again. It can drive into your home garage and be linked to the house system to open and close the garage door.
Tesla spent 8-9 years trying to do this with code. They realised human coding was never going to be viable so ditched most of that and went to Neural nets (out of my skill set) and training from showing the Neural nets footage of good real driving and the neural nets learn how to drive, the same way as humans do.
Weymo started with SF, I suspect Tesla will do the same plus possibly Austin. Even Weymo had to start with a driver in the car, several Uber drivers use FSD for their current Taxi rides. It will be likely Tesla will start with Model 3s with FSD and a driver (maybe ex uber drivers)to sit in the front keeping an eye on the road. They´ll create or use a current service center for cleaning and maintenance. Use AI to select areas best to park the cars when not in use, maybe some parking slots on Tesla charging stations. They can show the regulators miles of FSD driving around the city and outside.
That is the start and they then have a year to sort out the 2 passenger robotaxi which covers 80% of the taxi rides. If some one selects more than 2 passengers they get a bigger Tesla standard car. The driverless cars could start with 1 on 1 supervision from HQ and as things progress drop this to eventually 1 on 100 supervision with AI doing supervision on every taxi and intervene in an emergency and call back to HQ for a human intervention.
Tesla has always been buy the rumour sell the news so us retail investors will be expecting this but I think, watching the very recent progress a lot of people are going to be impressed, even wowed or OMG. Or the naysayers claiming it´s going to take decades if ever to become a reality will be shocked. Just remember those SpaceX space link rockets taking off and returning to land on the take off launch pad to be re-used within a few days, now a daily or weekly event.
Ok thats your opinion waung chong!
I am here to see bearish info, Everything he says is like the opposite of what I think 😂, Like man, getting rid of lidar and use Ai is the best thing to do.
Dude, you the man! Subscribed. Thankyou for this video and continue doing what you do. I see a lot of Muskrats in the comments sections. That's typical when you the rattle the nest, which contain Musk's girlfriends. Just ignore the cult. Reality is FSD is vaporware just like every other Musk project. I would never pay for it letlone their disposable cars. Also their aging lineup is hurting the brand (asside from Musk's antics) and the best Musk can give everyone is a silly 'truck' for rich people. Tesla will be heavily diminished of a company or go bankrupt by 2027 at this rate. And as a side note, self-driving cars aren't going to solve anything, not even traffic. What will will be removing or reducing the need to drive at all, through changing our zoning laws, mixed development, giving people more options to get around, such as more HSR, bike lanes, and just walking. All this hype around EV's (especially Tesla) and FSD is pure nonesense.
I feel all Robotaxis are prone to carjacking. Once thieves realize how easy it is to force robotaxis to stop in the middle of the road and subsequently rob the occupants, it will mean robotaxis can only operate in upper class markets thus limiting potential rollout. Even just a few robotaxi carjackings a year will scare off a large portion of potential riders. I will never ride in any brand of robotaxi for this reason, when you're inside, you are a potential target. There is a use case for personal FSD, but robotaxi's in their current form are easy targets for criminals.
Claim: Cost of training will go up. Based on what? Once Tesla solves FSD there is no need to retrain the model every day with new data thus cost of training might actual drop. Further more training and development of Teslas FSD is fully paid by customers buying or renting FSD
Claim: Teslas need LIDAR. Based on what? I as a human use my vision (eyes) to measure distance and objects moving. Tesla can do the same without LIDAR. Yes LIDAR is more precise, but if Tesla can be >99% precise without LIDAR, why use LIDAR? Tesla does use LIDAR on their test cars but only to verify that the vision system works as expected.
Claim: Other car companies will not license FSD because they are heavily invested in their own tech. Based on what? There are so many historical examples like HD-DVD, BetaMax, NACS vs. CCS, CDMA 3G etc.
If Tesla has a working system that costs less then the competition and works better, why would anyone develop their own more expensive, less capable system?
Claim: Waymo is ahead because it has less interventions. Based on what? You do know that you are comparing apples to oranges? You are comparing a fenced system that only drives on premapped roads to a system that can drive everywhere without premapping. And on top of that interventions are registered different on Waymo and Teslas systems – you simply cannot compare the two.
That is the BS I saw in the first 3 minutes of the video… I do not have the time or energy to write a longer post.
This video has so many wrong statements based on clearly wrong assumption that watching it is a waste of time for anyone except for people without a working brain.
There is more than one way to succeed in FSD. Huge collection of Data has so many edge cases that happens rarely. Those edge cases can be corrected and thus increase it's safety. Real edge cases not simulation data. The more compute the faster the FSD becomes safer. Texas will gain another 100,000 GPUs this year in Texas to work on FSD.
Tesla has a grading system and knows which drivers are the gold standard. What if Tesla used data from these drivers and weight the decision of FSD based on the best drivers with safety score above is 90%
I am a race car driver with SCCA. I set the lap record at sears pt race way and Laguna Sega. To set the lap record I have to make almost zero mistake turning onto each turn. Should information be taken by trained race car drivers with exceptional skills.
Bla bla bla, before all this nonsense just try a Tesla with the latest FSD no need of lidar and data collecting is Tesla strengh and Mojo is coming. I think Elon is making big mistake in politics but the FSD is absolutly amazing
no way, you are dead wrong!!
If Robotaxi is such a good idea, why doesn't Musk launch his own fleet of taxis?
Gee, do you think the current interventions with Tesla FSD has anything to do with lack of Lidar? In other words, would adding Lidar fix ONE SINGLE issue with current issues? Doubt it….. Waymo works because they are geo fenced into a specific area. FSD NOW would work as well or EVEN BETTER then Waymo if they restricted the routes that Waymo currently drives. In other words, FSD is as good or better then Waymo as I write this. So, Tesla robo taxies will EASY work as well as Waymo, or even better. You leaving out the HUGE LARGE elephant in the room. The simple matter is Tesla is approaching FSD as a general driving solution, Waymo is NOT EVEN CLOSE to doing this.
If you use a level playing field and restrict FSD to the same limited routes as Waymo? Then they can restrict the data to that GIVEN geo-fenced area that Waymo drives? Then you of course eliminate the noise you speak of.
Comparing Waymo to Tesla FSD is beyond brain dead here.
They are not the SAME type of system. The INSTANT you restrict Tesla FSD to a geo-fenced waymot route, then FSD will easy compare to Waymo.
Again:
Leaving out that Waymo is restricted to a given route? The INSTANT you do this, then you are NOT using data from everyone – and Tesla FSD will easy compare to Waymo, if you use the SAME COMPARE and SAME restricted geo fenced defined route.
You should find another job bro – from what I could barely understand, you sound like you don't know what you're talking about.
Good critique but incredibly repetitious.
I'm a Tesla bull, but always eager to hear bear arguments, so I was excited about this, but this was not good.
Started well with concerns with savings in drivers being offset by other costs (fair), but then went south when talking about LiDAR and poor training data.
The way you are wrong is criminal. I hope you have your money where your mouth is because you gonna lose it all. I am not even argue with you because I don't have lifetime for that.
Based on past performance, it's not a good idea to bet against Tesla.