Tesla Autopilot Data Incoming / Wayve vs. FSD / Tesla Argentina ⚡️
HTML-код
- Опубликовано: 27 май 2024
- Go to ground.news/electrified to stay fully informed. Subscribe through my link to get 40% off unlimited access. Promo expires soon.
NEW 3D Mats Link: bit.ly/3WsIAQk
🧡 A massive thank you to my Patrons: / electrified
👍🏻 Leaving a like/comment and subscribing are great ways to support the channel for free.
Instagram: dillonloomi...
X (Twitter): / dillonloomis22
💰 Get discounts and/or freebies using these affiliate links:
🔹Athletic Greens (AG1): drinkag1.com/electrified
🔹Surfshark (VPN): surfshark.deals/electrified
🔹Ekster (Smart Wallets & More): shop.ekster.com/electrified
🔹Vessi (100% Waterproof Shoes): vessi.com/electrified
🔹SPAN (Smart Home Panel): www.span.io/?...
🔹 Ground News: ground.news/electrified
🔹EnergySage (free solar quotes, no phone number required) - www.energysage.com/p/electrified
🎥 Executive Producer: Halter Ferguson Financial
Podcast (audio only) version of Electrified:
Spotify: open.spotify.com/show/5yfYEUp...
Apple: podcasts.apple.com/us/podcast...
WARNING: I will never message you in the comment section asking you to message me via WhatsApp - these are scam bots that I'm doing my best to continually block.
===================================
NHTSA Letter: static.nhtsa.gov/odi/inv/2024...
===================================
00:00 - Tesla's FSD
2:19 - Autopilot Data Incoming
4:39 - Don't Be Deceived
6:11 - Tesla Semi
6:55 - Cybertruck
7:11 - Tesla Argentina
8:36 - Tesla China
9:21 - Berlin Protests
9:53 - Piedmont License Official
10:11 - Mr. Beast
10:42 - 5G
11:04 - Optimus Production
11:56 - Great Idea
12:37 - Wayve vs. FSD
16:08 - Apple/Rivian
16:18 - Robin Denholm
16:37 - Rivian Q1 Earnings
18:10 - Tesla Stock & Wisdom
===================================
Some links may be affiliate links - you pay the same price, I get some credit. Great way to support creators you value 👊🏻
*Disclaimer: I was a financial advisor with Edward Jones for a number of years. That said, I am not your financial advisor so please always do your own research and do not base your investment decisions on my videos alone.
#TeslaNews #TeslaStock #TeslaAutopilot 
Авто/Мото
Hope you all have a great few days! Don't forget to check out Ground News if you'd like to stay informed without the bias - you can get 40% off the Vantage plan here: ground.news/electrified - thanks in advance if you decide to support the channel in this way!
I love how NITSA nitpicks 20 issues out of 4 million with FSD but if you want a stop sign or new lane stripes to fix a dangerous road they won’t even consider it until 3 people die there.
There isn't 4 million Tesla's with FSD. You can't use it in most of the world.
That chops your number significantly.
Out of the countries that do allow it, not every Tesla has FSD.
Again chop that number down.
Out if those that have FSD, only a small number use it.
Again chop the number.
Out of those cars using it, those were the stats.
Pretty simple to understand. Definitely no Cherry picking going on
So true - NITSA bullshit is unbelievable. Imagine asking Ford this level of detail on every vehicle.
@@mhsj who says they didn't? Were you there?
NHTSA's not nitpicking about FSD. It's about Autopilot.
@@mervstash3692 really, 9 of the cases they are enquiring about include lane assist, you can’t see the desperate attempt to make anything look like an FSD failure and own the liberal Elon? But just to satisfy your neurotic need for specifics. There are 400000 teslas with FSD in operation for 4 years, day in and day out racking up the miles and your loosing your S#!T over 20 of them when I have to wait until 3 kids get run over walking to school before I can get a speed bump put in. Great priorities Merv
I can turn my cruise control on in my Silverado and drive through a wall. It doesn’t monitor me at all. Why is this so scrutinized?
I think the name is flawed, that's why we nitpick Tesla. But they are stuck with it. We call it FSD beta or supervised, whatever. When you hear full self driving it's naturally concerning. To be clear I drove with FSD on and it has Amazing capabilities!
Cause Elon bad legacy good
Your cruise control doesn't lull you into inattentiveness
@@frednicholson Not true. Plenty of people have had accidents thinking cruise control drives the car for them, or at least keeps them from having accidents. Stupid people are everywhere.
If you don’t advertise you can’t expect a sympathetic media.
He who pays the piper calls the tune.
When you release negative information about the Bidens expect every government agency to come after you…
NHTSA demanding information that no other auto maker can provide should be illegal. That or GM and Ford are legally required to be able to submit the same data.
No matter how big a hit you think Cybertruck will be, you're underestimating how big a hit Cybertruck will be.
Nice play of words but quite empty.
Anybody could say that about anything.
You must provide some evidence to back your statement up.
Absolutely bro.
The biggest fails of all time always find a way to become amazing 😂😂😂
@@Kukaboora Dillon does that for him. Do you watch the videos?
How? Max production planned is 375k or so?
@Kukaboora - the evidence is Model Y being the best selling cat. Cybertruck will be the best selling Truck globally and will be 2X second place once they build a slightly (10-15%) smaller truck for South East Asia and Europe
Mark my words
Wayve started a year after Tesla, now about four years behind. Videos show it mostly driving two lane roads, no lane changes, I have seen one unprotected turn but it is a comically easy one, not remotely close to Chucks turn, and no roundabouts. Wayve doesn’t have the data or the compute, but at least they are not using LIDAR or geofencing.
Additionally, the chat is a real gimmick as, if it’s really a neural net, it should not be able to easily determine the reason why it stopped or slowed down. And it’s only useful when it behaves not as expected
@@lucaslouch Good point. If their program is going to be making safety critical errors, the driver needs to be intently watching the road and not the chat. Even if the chat has useful information, it is a hazard. If they achieve level four or five, then there will be nothing to read in the chat of interest anyway. I suppose it could be a little useful if it was used when the car is confused and wants advice among several options?
@@tribalypredisposed hopefully not! Imagine if the car is in the middle of an intersection and you have to type or dictate what you expect the car to do 😂
Their most recent videos show plenty of unprotected turns and lots of difficult long tail situations that people would rave about if it was FSD. I'm all in on Tesla stock, and a big part of the reason is FSD, but seeing how competent the Wayve system is given they only have a handful of cars on the road makes me question whether the data moat is as important as people make out. How is it even possible that the Wayve cars can drive for apparently as long as 40 minutes through London with no interventions despite having literally less than a billionth of Tesla's training data?
@@JWRB6 My search found a video from February as the most recent, did I miss some? The February video has one unprotected turn, and sure it squeezes through some physically tight situations but I did not see too much that impressed me. Roads in the older areas of cities in Europe are generally two tight lanes, one each way, and that is pretty easy and low speed driving.
I love the incredibly detailed list from NHTSA.
It's kinda dumb to ask for that kind of detail from a company with the best AI on the planet. If I were Elon, I'd send them a 2 million page report, a complete dump on every Tesla vehicle. Every day for a week!
Seems like some insider somewhere is using NHTSA to get to peek behind the curtain of FSD development.
Tesla's near term strategy seems to be to maintain and optimize their existing products while massively increasing power (e.g., 4680, Megapack, and Powerwall-3) and compute (e.g., Project Dojo, Nvidia GPUs, Hardware 5.0) capacity. It appears they are attempting to deliver greater value and reap greater profits through existing product lines and service channels.
Wayve is cute. Like a little puppy barking at the big dog. There will be so many cash grabs in the AI sector before it becomes clear who actually has something worthwhile. And just because you say you have focused on an end-to-end AI approach since 2017 doesn't mean anything if you don't have the data and can't actually produce anything.
That's right. End-to-end isn't new concept that Wayve invent. It has been everybody goal of AI training field since the beginning but it'll required tremendous amount of compute power and incredible amount of data (Elon said it's in the millions video clip). That's why back then people even thought it might be impossible. This is also why we're only see end-to-end in an extremely large tech company like Google or Meta and now Tesla. It'll required +$10B and more on servers (and +$100M operational cost/year) to even have a chance to think about doing end-to-end. Wayve raising $1B is like a drop in ocean.
@@mukamuka0 Absolutely, and Wayve is only getting $475 million of the billion, the rest is going to buy out the previous owner. All the money Wayze has is less than half what Tesla is spending just on computer hardware just this year. But then two of the investors are Nvidia and Microsoft, so super clearly they know about the need for compute and are going to supply it in some way. As for the paucity of data, I assume they are planning to use simulated driving scenarios for training. I have skeptical feelings about the effectiveness of simulated data, but I am not a computer scientist in the least, so I know nothing.
Forces are allying together to stop Tesla from having basically a monopoly in robo taxis and everyone having to go to them for full self driving tech. But by the time Wayve could possibly progress from where they are to having something to offer and then signing automakers up and having the automakers add FSD hardware to their cars...five years would be a miracle and seven would be amazing. Everyone started with the view that Tesla would fail on FSD, and now that Tesla has made them believe it really is too late.
Wow, Rivian is way more profitable than Ford on their ev's.
ha ha!
one has to ask how profitable are GMs EVs? they hide that info under a pile of ICE sales. anyone betting that GM sales are the same as fords?
@@scottmcshannon6821 I can guarantee you that Silverado is a cash burning machine!
I would love to have Waze's combine with google maps with real time traffic and weather.
Isn’t Waze owned by Google?
@@pokemangoduo431 Yes. I believe they do share some data. For example when Google says route delayed by 30 mins due to accident.
Great production Mr Dillon ty for ur time. In all ur infinite wisdom do u think grok is actually Tesla 1 or no ???
@ 19 : 48 CHECKMATE CHECKMATE
Always trust the guy who says they have infinite resources. Genius level quote
Wayne might be looking at a situation where they equip vehicles at cost or free with their tech just to get the data.
They should use Grok to generate a volume of information for SEC request that would take decades to analize. Come on, Dojo and Grok could bury the SEC!
interventions rare? I'd intervene 10x times at least per drive. Potholes, the AI wanting to move onto an on/off ramp when going straightforward, not preparing for an upcoming turn soon enough, braking super hard for stoplights IE not slowing down gently soon enough. Using the FSD without a destination, it doesn't recognize requests to change lanes or make turns. It's also a bit indecisive when choosing turn lanes, hugging the middle of two turn lanes not sure which to use. Also it'd choose either too slow a speed for conditions or too fast.
1:22 not quite. I believe in this instance Elon is referring to validating that the interventions logged are actually unique desirable edge cases for training the system.
Wasn’t Bill “Genital Herpes” Gates focused on malaria?
👍
Why no hits for "Electrified" when doing a search on YT?
Dillon, that 238% inflation rate was for last year. Milei's policies have cut that rate down dramatically, one of the things that I'm sure impressed Elon.
Milei is failing!
You are wrong.
287,9% is from MARCH 2024! 😂😂😂 He is failing! 😂😂
Hey I don't know if you read your comments, but it would be nice if you can add a black border at the bottom of your videos to make the chapters in the video timeline readable. When the video is white, it's very difficult to read.
Viva la libertad carajo!
3:28 item "p" seems like irrelevant fishing.
You do realize most government workers are useless, and thusly they create bullshit work for themselves to justify their continued employment. Nhtsa should only be a 1 time meeting by the states to set nationwide vehicle safety and road standards, it's a completely redundant make work agency like the EPA
NHTSA absolute numpties
Biggest corporate fleets will do their own car designs by 2030 like Bolt and Uber. Bolt will need like 1-5M cars/year . Bolt will do like 10% of all trips driven in Europe that time. The first city that will be 10% is Bolt's home Tallinn Bolt Drive service is allready 5-10x the Bolt as Taxi service and that will grow . Soon they run 20k Bolt drive cars In Tallin area and it is about 5% total cars in city. That may project to 10M additional cars into service by 2030
Zzzzzzzzzz
AI especially in driving will be a mess. It's like reaching a finish line but only cutting the distance in half each iteration until you are infinitesimally close to the finish but yet you can always cut that distance in half infinitely each cut harder and harder to come by thereby always leaving open a lawsuit and liability. Will it be such a superior system as a whole that when a death happens the law will be "tough nuts", indemnify the system and pay from a fund but NOT penalize the system at all because they can prove it's 10,000% safer than humans? 👀
Yes.
If regulators and legislators don't make that so, history will judge them harshly.
Each country can decide.
It only needs to be twice as safe as a human driver. Insurance will cover accidents and liability as it does now. The cost of insurance will drop significantly as FSD gets better.
$300,000,000 for Robyn $56,000,000,000 for Elon = shareholders getting ripped off.
You don't know how the package works at all do you?
@@TheEvilmooseofdoom do you realize just how much 56 billion is? It is 10 grand for every car Tesla has ever sold...it is more than the profit the company has made in its history. Not to mention the value of the shares has been cut in half.
The self driving data request will be massive. Tesla could go down the list as data is compiled sending it in. My make up is it'll be overwhelming.
Nhtsa is requesting information it has no right to. They're fishing to find a reason to get litigation moving
The cost and difficulty of getting the data is 100 percent of the reason they have asked for it NHTSA doesn't have the people , computers or knowledge to do anything meaningful with that data.
Ground news only seems to let you select your preferred bias. Not what I need which is fact checking.
Wayve 🤣🤣🤣🤣🤣🤣🤣🤣 No data, no drive.
Question. Why is Tesla now the biggest customer of Luminar for q1 of their lidar sensors? Elon has said multiple times that lidar is a fool's errand and anybody using it for autonomous driving is doomed, but now they are the largest buyer of them totaling more than 10% of all luminar lidar sensor sales for q1.
Please provide a source.
Tesla uses lidar to calibrate FSD. They will never put lidar in production cars.
Yes I understand that they use lidar for calibrating their camera systems. However that doesn't explain why they bought so many of them in q1 to become the largest customer of luminar totaling more than 10% of all of luminar q1 lidar sales
@@conradseelye5306you can check it out on the luminar q1 earnings call. So luminar is the source. However it's also been reported on Reuters and TechCrunch and many other media outlets
@@jdreynolds29 How many is that? It could be 10% of 1000 systems. Not very much.
Dillon, enjoy your days off! And Mr. Beast, happy birthday!
As someone who drives my Model 3 60K per year, i state without hesitation, that while FSD 12.3.6 is city good neither FSD nor EAP, are acceptable on highways, very unreliable typically 10 fails per 100km trip on sunny day (bat shite craziness on bad weather days).. fake lane changes, not lane changing, not returning to lane, phantom breaking, no reason lane change, non-standard lane and return to lane changes. so unreliable my wife refuses to use. great when it works but flaky beyond belief, good for NHTSA.
I wonder if that could be a region thing. I have a friend with a 6 month old Y with no FSD. The autopilot on the highway is flawless. Hours of smooth driving.
When it does sommin crazy do u send time stamp or report to Tesla or no ???
Hopefully this problem will be gone by reaching level 5 and it will be history.
Many offices in this region (Germany, Austria, Netherlands) have the Thursday holiday and employees take Friday off to make it a loooong weekend
there is no way NHSTA has the ability to actually go thru terabyte's of data, this is just harassment.
Tesla should integrate radar data into fsd
If you can drive without radar, so can FSD. No part is the best part.
Why?
@@jmattoxriskpro cameras can't drive in fog. Radar can.
@@billweberx I can't drive in fog. Neither can FSD.
@@DanielASchaefferRadar sees almost nothing. It's good for seeing an airplane in the sky, or the distance to the car in front for adaptive cruise control, which was good 20 years ago. But that's it. It can't read signs, road markings (e.g. lane markings). It does not provide redundancy for camera systems.
Enjoy your trip!
Who is Mr beast?
He is an Youtubist.
Ask Mr Google ?
Multiply by 69 INSTED of 100 duuuh
@ 1 1 : 1 2
Elon has said from the start that the closer they get to full autonomy, the harder it gets. That's proving to be the case. The really big question is how close is good enough? Users won't accept it's safer if it does weird or sketchy things that they would never do. Human drivers have a vast experience of the World around them, they're not limited to a specific training set. This is the real issue, and why I think it's way too early to be thinking about unsupervised FSD and Robotaxis. It's an unnecessary risk to Tesla's reputation to roll this out before it's truly safer in every situation. That could take another 5 years or more. Personally, I think that's preferable to the firestorm coming Tesla's way if they jump the gun. That could set back the cause for FSD far more than just being patient and diligently working to make it as safe as possible.
There is a LOT of variation in the relative skill levels of human drivers. From the 2 week license holder to the 40+ years of driving. Humans do a lot of weird and sketchy things too. FSD only needs to be as safe as the average driver and that is getting pretty close. And unlike humans, the improvements in 'skill' roll out fleet wide, not one person at a time. I think you're being somewhat short sighted in your appraisal.
Wayve CEO gives me Trevor Milton vibes.
You think it's teleoperated? 😂 That would be jokes
Governmental over-regulation.
What happened to this channel? Why no new video?
No post on twitter no community post on RUclips? Hope Dillon is ok but still wondering what happened
LOL Wayve saying "welcome to the club" is like Jeff Bezos saying "welcome to the club" back in 2015 when Falcon 9 had its first successful rocket landing, because Jeff's tiny suborbital toy rocket had also successfully landed one month prior. What have SpaceX and Blue Origin done in the subsequent 9 years? Blue Origin launched a half dozen 10-minute suborbital joyrides to the edge of space for 3-4 minutes of zero gravity. Blue still hasn't launched anything to orbit. SpaceX has launched and LANDED over 300 ORBITAL rockets, after delivering both satellites and astronauts to orbit.
It'll be the same story with Wayve. In a few years, Tesla will dominate autonomous transport and people will say "Wayve who?"
BUY TESLA & NVIDIA STOCK now
TESLA SEMI = NO BRAINER EVEN AT $400K but I doubt Tesla will charge that much in 2024 and 2025.
NHTSA seems to be going just "a little" overboard. I'd like to see Tesla challenge it forcing NHTSA to detail why they need each bit of data requested.
TACC not disengaging when the steering wheel was turned was a very dangerous oversight on Tesla's behalf. It took them far too long to fix this issue. Crazy thing is it was always set to disengage TACC in the X and S, the problem was only in Y and X.
Thanks for another one!
Ha, the next generation won’t be driving. Manual driving will be mostly hobby or sport.
Hey, Ive been at Costco in Santa Cruze. Maybe I drove on the same road as a semi
Merci !
The next generation won't be driving or buying cars anymore.
So how long do you think it will be after robo taxi hits the road that some bad guy decides to put a package in the back seat that has a timer or a remote control? Self driving cars are basically slow speed cruise missiles that will provide a means to deliver a package where it will cause damage.
Interesting...
Don't do it
Synthetic data is important say the car has a sticky situation in real life. You can recreate that specific scenario and run it hundreds of times at all different speeds for both cars. Making it no longer an issue.
i don't think "factuality" is word
Wayne is the next Nikola! Pure BS
Tesla is their grandpapi
Excellent commentary once again
When has any TSLA director BOUGHT shares? They always sell. When has Warren Buffet sold shares in brk? He gifts shares but never sells.
The nhtsa data is interesting. Of the 20 accidents only 9 are relatable to driver attentiveness, i.e. the frontal collisions. The traction loss accidents will still happen, i doubt autopilot or FSD to be any better at detecting slippery surfaces than a person, likely worse. And Tacc remaining active after a manual takeover, indicates the the driver did take over.
1:10 FSD doesn't need to be engaged if the user has accepted the terms. They can run the model in the background and see if it deviates from what the user does. It isn't as likely they'd get stuff, but it still can be used sometimes.
That's the ghost mode they talked about years ago.
@@TheEvilmooseofdoom basically... Although I thought ghost mode compared new versions to currently running versions. It might have done both though.
I ordered the 3M mats two days and tried to use your code. Unfortunately, it didn't work; so, I glad that you got a new code.
We need to boost Dillon’s channel, friends. He deserves to be over 300k subs by now. Thank you, Dillon, for the info and your level take on the data.
NHTSA is bullying Tesla again
Explain
@@mervstash3692 I think it is pretty evident. Has NHTSA forcefully demanded the same amount of detailed information to other vehicle manufacturers regarding the performance of their ADAS systems? That's what they should do if they are committed to an objective and fair playing field. Don't you agree? What about the numerous highly publicized "recalls" for mostly minor stuff, that were or that could be fixed with a quick over the air software update?
@photogol again, how do you know they didn't? Speculation doesn't = fact.
Re recalls, every recall is serious. It's irrelevant if it can be done remotely or not. And every company that faces recalls is equally scrutinised. You just don't like it when it's about the thing you like.
@@mervstash3692 Have you seen the extent of the information request that they sent to Tesla? The level of details that they ask for and the response time that they require for the information to be submitted? do you think that’s reasonable? Don’t you see the amount of pressure that the mainstream media, the short sellers and other interested actors, including people in government, exert in the regulatory bodies that have anything to do with Tesla to harm the company. I don’t know if you fit unto one of those groups, but I can see that you are not an objective observer, so give me a break…
Search for the 9's... We haven't gotten to 90% yet, but no harm in being optimistic even if the 99% is many years out.
Based on what measure?
I absolutely hate NITSA’s tampering of FSD. I have gotten 5 strikes on the FSD for me looking at the navigation screen. Some of them for looking at the screen for state of change information. They’re ruining FSD by making supervision as eyes forward during the drive. I don’t even do that for normal cruise control in ICE cars.
And that's the point, but the nerff was not big enough, as people are still buying it.
The public is to blame. I seem to be the only person on the planet who objected to the inward facing cameras. Hands on steering wheel is annoying, but an inward facing camera spying on your every move is creepy. Now every Tesla and F150 lightning has them and it’s the standard.
This is why I’m never replacing my 2017 MS.
FSD training is running into the law of diminishing returns.
Yeah, pretty much. FSD will never be 100% perfect. But it will asymptotically approach 100% perfect.
I believe the more vehicles using AI driven FSD the easier it will be. Imagine the AI that knows what all the other cars will do because it's the one doing it. It will know and recognize human drivers around it and be able to react like a flock of birds reaction presumably stopping accidents at some pojnt.
Attention ALL humans …… The artificial intelligence regulatory authority has jus issued a warning for everyone buying a Optimus sub prime robot . Learn to manipulate charger robot uses………….. anyone laid off from Tesla can build Optimus robots . Don’t put up a mailbox cause when there’s enough made to work at Tesla th robots will build robots
My hands on wheel warnings has gone through the roof even with one hand always on the wheel since the recall.
What "RECALL"?
My FSD "test loops" are both in adjacent small towns, a combination of winding residential streets, 2 and 4 lane commercial area streets, T intersections, 4 way stops and a complex angular diagonal intersection, merges, and "only" lanes. I experience speeding, braking for shadows, breaking for a stop ahead sign, and breaking then speeding up as it approaches some stop signs, crossing the double yellow to give a cyclist or walker more than the required three feet, and pouring on the coal to stay ahead of cross traffic, also moving into the parking lane before a right turn and turning off the turn signal before reaching the corner.
If the only way to get Tesla is to see some these problems is an intervention, then there would be several every trip of even 5 miles or less.
That kind of request gets a gfys response, the secratary cleans it up for me.
Several simple querys of their database will provide the info.
I think everyone is getting too hyped by FSD. The harsh reality is that GPT 4, a large language model based on transformer AI architecture has a 3% hallucination rate. That means it puts stuff together to generate wrong results. FSD is also based on transformer architecture, so one might expect a similar hallucination rate, basically doing the wrong thing. The failures are going to occur in edge cases and in cases, for example, light and shadows confuse the AI.
The first publication on Transformer architecture was published in 2017. Is this software architecture the right one for AI, or is there another advance in AI research required? I suspect further advances will be required and just throwing data at the FSD problem will generate diminishing returns.
So the current issue as alluded to is how do you verify that the training has generated a safe FSD? However, beyond that do you need a system that learns from mistakes? Do you need an in car AI system that has the ability to do some adaption to its local environment? Ultimately do you need a system that has enough real world understanding to reason its way out of trouble?
The Elon hype machine creates expectations which maybe difficult to meet.
11:39
Test ride 12.3.6 and tell me if it is hype.
@@snookmeister55 Well, I would be pleasantly surprised if it works as well as you imply. However, I think it will fail in difficult conditions like, rain at night. I also suspect it will make the odd unexpected wrong choice. I just don’t think the AI algorithms are good enough yet. I think there are a few needed research break throughs yet to come.
We will see anyway.
Let's see those millions of GPT 4 miles for comparison ?
@@fredturk6447 Your expectations are your business. Sure, Tesla will invest $10 billion in AI this year alone in but it won't work in the rain. Perhaps someone at Tesla will see this, remember that it rains and simply give up. Then you'll be right.
@@rogerstarkey5390 GPT 4 training set 13 Trillion tokens, which is likely well over 13 Tb depending on how many bytes are allocated to a token. Now Tesla processes millions of Terabytes, however, most of it is not that useful because uneventful driving along a freeway does not improve your training outcome. The other critical point is that regardless how big the dataset is the information contained has to be compressed down into effectively vectors in multidimensional arrays. So I think GPT 4 requires 16GB of VRAM and a couple of hundred GB of disk storage to run. I do not know what the Tesla FSD computer specs are but I would think it’s going to be that sort of size.
So the available memory/storage in the car will limit how well the FSD AI system can deal with edge cases and also avoid hallucinations, although that’s probably a more complex issue.. At least that’s how best I understand it. I could easily be wrong. Anyway it’s certainly interesting to see the progress taking place with AI.
Ford has a driver aid, GM has a driver aid, Stellantis has a driver aid. All are being "investigated" however no information about the investigation "requirements is provided. In the other companies the system was at fault, in Tesla case the humans are not following the rules for using the system. There is a huge difference on system caused accidents and accidents caused by humans not following system warnings.
0:45 Based on my month free trial, find it hard to believe interventions are that rare at all outside California, Austin, and maybe other areas with a lot more data than around here.
Dillon, I’m a bit worried about sales in Europe. There’s so much fud it’s unbelievable. And a powerful lobby against EV. What to do about it?
What kind of beast is Jimmy Donaldson? A GOAT
@lucalla
35 seconds ago
NHTSA has had a hard on for Tesla ever since Elon locked horns with what's her face... She keeps coming off second best and she can't stand it... so she uses her department to go after Tesla... A real professional
Trump had it right. For every new government regulation, you must remove 2 government regulations.
So much bullcrap about humanoid robots. As if Tesla can afford to waste efforts on sci-fi. Musk is so lost.
Since NHTSA forced Tesla into the last recall( OTA) for Autopilot it is clearly not as good as it was before. It sounds like the data might be showing that. I have put well more than 6,000 miles on Autopilot in my first year. It was impeccable before the forced update.
I was thinking validation meant curating the videos for good training data
Man I send them so many interventions every single day from city driving in my area I can't believe Elon on this.
No you just don't actually understand what's going on or they why and how.
If the AI is intuitive and deductive, you do not need a lot of data. AI will recognize the context, the circumstances to act appropriately.
I can go to Timbuktu for the first time and I can drive on any road there without prior knowledge (data) of any road. I can determine which is a drivable or not drivable surface.
I think the role of data is way overhyped.
According to Statistica, in the e US only 3% of the population need taxis 2-5 times a week, 58% need 2-5 times a month and the rest go anywhere with their cars.
Uber operates worldwide and last year their revenue is around 50 billions. I think the 10 trillions number is exaggerated.
Tesla will face many and fierce completion in both area of autonomy and robotaxis.
To name a few, Boston dynamics, Agility industrial robots, XPeng FSD, Waymo, China robots…
Tesla is not alone.
In the mean time Tesla could face an operational loss in Q2 24.
NITSA is sharing Tesla data with Detroit
Hi Dillon Loomis 22 from Trisha 🙂💕
SEC still allows Elmo to manipulate the stock market. Amazing how Elmo fanboys believe everything Elmo says.
No they just understand what manipulation actually is unlike you people who get told what to think. But you haters are not the smartest bunch and certainly have never yet been honest. You do wear your failure on your sleeve though, everyone knows you for what you are.
@@TheEvilmooseofdoom LMAO, Full Self Driving (FSD) still doesn't work. How long have you been following Nick Fuentes?
NHTSA losers
A broader picture of Tesla's strengths:
Tesla is the most efficient car manufacturer on Earth. Tesla is building its AI in-house, both hardware and software. Tesla's vehicles have a $1,500 autonomous hardware kit that has been installed in every Tesla vehicle since Feb 2019. Tesla AI is based on LVM (Large Vision Model) which is trained on real-world video.
Everybody else is using LLMs (Large Language Models) which are trained on text and photos. They don't have in-car hardware unless you consider Waymo's $250,000 solution viable (PS if you do, you're a dummy).
Wayve are buying themselves a computer setup and hiring some programmers. They're not going to start from basics, which Tesla di over 9 years ago, nor are they going to find a Pete Bannon or Andrei Karpathy. Wayve's smartass comments about Tesla catching up are hilarious. Wayve are so far behind they don't even begin to realize how far ahead Tesla is. Tesla isn't catching up, Tesla's lapping them!
I hope someone tells the governments in the EU about this "all electric company car" push. They are going to need to bring a lot of nuclear power plants online, and pronto. They can barely handle current power demand as it is, thanks to eliminating working "legacy" power generation for promised future "sustainable" power production.
The additional EV power requirements are not that high. Most EVs are charged at night when demand is low and capacity is high.
not interested about Cyber truck or robots or FSD anymore.
invested in Tesla since 2013.
don't want a robotaxi that some has taken a shit it in.
I don't use Uber .
I drive my own car . because I don't want to share my vehicle with a puking drunk.
so no longer interested with Tesla.
will divest another $100 million .
I am over the drama.
firing the super charger team was the last straw.
Tesla is run by raving lunatic.
Blind as a Kiwi.
after 11 years owning the stock and owned all the Tesla vehicles.
half with FSD.
no longer want the drama .
will vote no to Elon pay package .
I want a new CEO.
that is not a lunatic.
one man did not build Tesla.
LOL. Mkay.
@@kiwimajic1 ruclips.net/video/zl4VjvZu5os/видео.html