NVIDIA CEO: Tesla is FAR AHEAD in Self Driving Cars | Waymo vs Tesla FSD

Подписаться 83 тыс.

Просмотров 11 тыс.

50% 1

Nvidia’s CEO recently said, “Tesla is far ahead in self-driving cars...” however, there are of course experts who disagree with this including Timothy Lee who has a master’s degree in computer science from Princeton and writes for ArsTechnica. In stark contrast, Timothy claims that Waymo is years ahead of Tesla. So which is it?
NVIDIA CEO Interview: finance.yahoo.com/video/nvidi...
Ars Technica Article: arstechnica.com/cars/2024/05/...
AI DRIVR RU-vid Videos: • There’s a HUMAN inside... & • Stress Testing Tesla F...
Support Cleanerwatt:
1. Join the Patreon Community: / cleanerwatt
2. Purchase Cleanerwatt Apparel: cleanerwatt.creator-spring.com/
Cleanerwatt Website: www.cleanerwatt.com/
** All video and pictures are used with permission or in accordance with the copyright owner's stated policies and use allowance, or applicable fair use laws. **
Image & Video Clip Sources:
1. Tesla video clips/images used in the video courtesy of Tesla, Inc. (where noted in the video).
2. Tesla FSD video clips used courtesy of AI DRIVR on RU-vid (where noted in video): • There’s a HUMAN inside... & • Stress Testing Tesla F...
3. NVIDIA video clips/images used in the video courtesy of NVIDIA (where noted in the video).
Data Sources:
finance.yahoo.com/video/nvidi...
arstechnica.com/cars/2024/05/...
/ comments
thegradientpub.substack.com/p...
www.forbes.com/sites/bradtemp...
finance.yahoo.com/video/nvidi...
www.tesla.com/VehicleSafetyRe...
/ pariljain
• Stress Testing Tesla F...
Elon Live Stream: / 1695247110030119054
www.tangramvision.com/blog/se...
NOTE: The content found in this video should NOT be regarded as financial advice. I am not a financial advisor, and this is NOT in any way a recommendation or offer to buy or sell securities. While the information in this video is believed to be accurate at the time of recording, no guarantees are being made about the accuracy of the information presented in the video. As of the recording of this video, I am NOT invested in Tesla stock or securities, nor any other company mentioned in this video.

Авто/Мото

Опубликовано:

27 май 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 128

@flipadavis Месяц назад

Some guy who has a master's degree from Princeton? Really? I have a master's degree in my profession from an Ivy League school. So what. There are tens of thousands of people with masters degrees from Ivy League schools. I'll listen to the CEO of Nvidia.

@tribalypredisposed Месяц назад

Yep. Reading that article, as much as I could force myself to, just kept wondering why, if the author is really a world leading expert on FSD who can lecture Tesla on the error of their ways, why is he working as a writer? Somebody applied to Tesla and had his resume circular filed.

@daviddenley3512 Месяц назад

Huh, you think that's good...I got mine from an Approved School so there.

@reportingsjr Месяц назад

The CEO who has a very strong interest in pushing AI techniques that require massive purchases of GPUs? Hmmmm

@flipadavis Месяц назад

@@reportingsjr Doesn't Waymo need and use Nvidia chips as well?

@ItsBinhRepaired Месяц назад

This just goes to show how your degree doesn't mean you're smart.

@Zripas Месяц назад

Some people will get degree for the sake of having it, how they acquired it is another question, but at the moment someone starts flinging his degrees you will have 95% chance that he knows little to nothing about the topic. Smart people don't say that they are smart.

@kaihuchen5468 Месяц назад

> Timothy wrote, quote: "LLMs (GenAI) hallucinate they fail at simple tasks like counting objects and reading analog clocks." > Timothy has his doubts that an end to-end AI system will ever be accurate enough to power a full self-driving vehicle safely without a human driver Timothy is mistaken here because the right way to use LLMs here is not in specific tasks like counting or reading analog clock, but in using the commonsense embedded in its vision. My own experiments with OpenAI's GPT-4V vision model on road images have shown that GPT-4V makes very good judgements in all kinds of edge cases, which makes it an indispensable component in any level 4/5 autonomous systems. I can't find any in-depth information on whether FSD 12 is using GenAI, but from watching many FSD 12 test drive videos I can see that it is doing a great job emulating very subtle human driving behavior and making sensible decisions, so my guess is that FSD 12 is indeed using GenAI. If this is true, and if Waymo fails to go this way, then I do agree with Jon here that Tesla is going to beat Waymo on autonomous driving. I am not a fan of Elon Musk, so my opinion here is purely technical in nature. For what it is worth, I do have a PhD in Machine Learning from a Top-5 US CS program, if that makes me more believable.

@Cleanerwatt Месяц назад

Thanks for your expert opinion on this!

@michaelpaulholmes9667 Месяц назад

They need "Waymo" sensors. I'll be here all week.

@arbitrary76 Месяц назад

9:42 I have 12.3 on free trial. When I intervene or disengage, the car prompts me to record a short audio clip for Tesla to review, and I have used it most times.

@oldrrocr Месяц назад

My Tesla FSD is remarkable. Can't wait for the next updates! Thanks for the info!

@martincday007 8 дней назад

"Remarkable" and yet still needs an update. If Tesla 12.3.1 was a huge improvement over v11, just how dangerous was version 11?

@LegendaryInfortainment Месяц назад

I do not have a masters degree in anything but can easily avoid comparing apples to oranges unlike the author. I laughed so hard at that article.

@rickrys2729 Месяц назад

Have driven Tesla advanced autopilot for years and more recently with one month trial of FSD version 12. Certainly impressive, but FSD is often very slow and cautious and seemed to get speed limits wrong often. Tesla uses only cameras in visible spectrum and seems to distrust the front radar so it has issues in heavy rain, fog, and snow. In principle highly redundant sensors using Radar, LiDAR, Ultrasonic, Infrared, ultraviolet, and more and better resolution cameras should be able to outperform the existing camera system Tesla is using and provide situational awareness that no human could exceed. However Tesla seems to recognize that all the different situations and conflicting data from so many sensors would make "sensor fusion" based FSD training impossibly complex and time consuming. I don't have Waymo experience but it is great to see competition at play.

@davidpearn5925 Месяц назад

The Con is never going to stop....as long as it keeps working on the susceptible.

@MrGiggity890 Месяц назад

@@davidpearn5925THE COPE IS REAL WITH THE HATERS

@davidpearn5925 Месяц назад

@@MrGiggity890 ........ and that election WAS STOLEN......., right ?.

@EinzigfreierName Месяц назад

I'm very curious about the first real Tesla robo taxi. I wouldn't be surprised if they added more sensors, like radar and lidar and restrict the vision only system to always be "supervised".

@MrGiggity890 Месяц назад

@@EinzigfreierName 🤡

@ChiTsang Месяц назад

how did Lee get in Princeton? He doesn't even have a job.

@EinzigfreierName Месяц назад

Waymo is currently testing on freeways and will for sure integrate them in their normal rides soon. I fully agree with the statement at 6:50. As long as Tesla doesn't operate driverless, we don't know how FSD would handle this. If Tesla is ever going to operate a robo taxi fleet, it will of course be geo restricted also and there will be remote operators sitting somewhere in a control center in case a car needs help. I don't understand why Tesla fans always assume this would NOT be the case. It's is more or less inevidable.

@tribalypredisposed Месяц назад

We assume because we know otherwise. There is a zero percent chance of Tesla EVER having remote operators or using geo-fencing. The entire objective from the start was for FSD to be able to drive anywhere a human can but ten times more safely, with a tenth the crashes a human would have. To then have remote fallible humans in the loop…of course not. Tesla FSD is the most experienced driver to ever exist, and soon it will also be the smartest.

@EinzigfreierName Месяц назад

@@tribalypredisposed Tesla will either have remote operators and geofencing or there will be no Tesla robotaxis. You REALLY think there will be permissions for something like this anytime soon? Countrywide permission for a driverless car to go anywhere and without operators in the background who monitor the fleet and take care of cars that are stuck or have other kinds of issues? If that's really what Tesla is going for, then we'll see several robo taxi service providers in every major city long before Tesla has a single robo taxi tire on the road.

@tribalypredisposed Месяц назад

@@EinzigfreierName Yes, in fact I REALLY know that the majority of states already have laws allowing it.

@James-cq9dp Месяц назад

waymo has "waymo sensors" 🤣🤣

@user-ny2bx8ez1c Месяц назад

For a Princeton grad, Mr. Lee is a real bone head!

@tedg1609 Месяц назад

Lee has some interesting detailed concerns, not just some lunkhead reaction.

@martincday007 8 дней назад

Take just one of the points made, that Waymo does not operate on Freeways while Tesla does. Tesla-FSD-Supervised operates with a driver. Tesla do not have a driverless system, so Tesla may operate anywhere when supervised, but nowhere, without a driver. Waymo's remote services are part of the complete system, Tesla doesn't yet need remote services because the driver is always in the vehicle. If it operated without a driver, it would also need remote services. There is no comparing Tesla FSD-Supervised with Waymo, one is driverless, the other isn't. The reason that Waymo is only now starting to be certified to operate on freeways is because Waymo put safety first and don't over-promise. In a built up area, when there is an incident, the autonomous vehicle can more often than not stop safely, and where it is safe to do so, pull over and seek assistance. On a freeway, stopping and pulling over might itself be dangerous. Tesla currently get over the problem by disengaging and having the driver take over and take the necessary action. Waymo took the decision to get millions of experience under their belt, before tackling faster roads where split second decision making is required, and where the emergency action taken has to itself be safe.

@Ed-qd5tj Месяц назад

timothy lee doesn‘t know what the checker, chess is

@joewilder Месяц назад

Was this comparison by Timothy written in 2020?

@tribalypredisposed Месяц назад

Yep, and by multiple authors that year and every year since. But everyone at Waymo is smart enough to know they are doomed so just a matter of time until their plug gets pulled. They chose their path believing the Tesla way was impossible, so now they are shipwrecked on a deserted island and trying to make the most of it.

@poneill65 Месяц назад

Not to doubt Tesla's lead, but taking a statement from a guy that sells squillions of H100's to Tesla as honest is pretty foolish, or wilfully deluded.

@tribalypredisposed Месяц назад

Nvidia has 75% profit margins because they have years of orders on the books waiting for H100s and can charge what they like, so the idea that their CEO is trying to suck up to Tesla for orders is pretty foolish, or willfully deluded.

@tribalypredisposed Месяц назад

@@sunrisejak2709 Are you assuming Waymo is not a very good customer? What else do you think Waymo is buying for their compute needs?

@tribalypredisposed Месяц назад

@@sunrisejak2709 Waymo's parent company is Alphabet, as in Google, who spent $12 BILLION dollars last quarter alone on capital expenditures, almost all of that was spent on compute, so the Nvidia CEO is praising a good customer at the expense of a very very good customer, at which point I think this whole line of attack that the CEO is doing this to curry favor with a good customer falls apart, right?

@poneill65 Месяц назад

@@tribalypredisposed Regardless of their other cusomers, you don't say bad things about a MAJOR customer,.. unless you're an uber genius that thinks it o.k. to tell your advertizers to go eff themselves.

@casperhansen826 Месяц назад

@@poneill65right, you would just shut up and not mention the customer at all

@johnmorris1162 Месяц назад

I think the only good point that Timothy Lee has in his article is that Tesla probably will need to have people remote monitoring driverless taxis and intervening when necessary, at least at first. So Tesla job ads for that purpose would be good to look out for to work out when Robotaxi's are happening for real.

@tribalypredisposed Месяц назад

Absolutely zero chance of that ever happening, if Tesla was going to make a weak move like that they could have robo taxis today very easily, do you see any?

@hoffbd1 Месяц назад

The chess guy should learn to play checkers

@Cleanerwatt Месяц назад

haha...

@YinLawn Месяц назад

What makes Tesla different from others is how fast it can learn. I literally don't watch any FSD videos from last year because they are outdated. Anything you said about Tesla today, will be outdated next month. However, I can't say the same about others, including waymo.

@Zripas Месяц назад

I always only watch videos with the latest version, as there is no real point in watching even 0.0.1 older version

@ahmadalanjary2020 Месяц назад

I am listening to this video in my model 3 using FSD 12.3.7 , And part of my drive was a dirt road. 😂. Keep on keeping on there Chevy. Waymo…. More like way-less .

@philipandregjone8873 Месяц назад

Why cant we have a crew of 10 drivers driving 10 hours each after every new version (ex 12.4 next) and then compare the versions in miles pr interventions (ex 12.3 vs 12.4) !? Driving the same route each time with different situations , from highway to city centers, and with video of all hours as documentation. Only then we can identify progression in a objective way.

@Zripas Месяц назад

There are youtube channels doing basically this, driving each version over same routes and checking how it handles those. Tho lately it became really boring to watch as nothing special happens, car just does its thing as it should do it.

@gerald1964 Месяц назад

mmmm, Is there any plausable reason why Jasen Huang would make such a statement? Is there any benefit for Nvidia? In any case, I would certainly not expect any real-time vehicle control input between any Tesla cloud and a Tesla vehicle given the latencies that are involved, i.e., perhaps something on the order of a few tens of mSec.

@billwoznicki1012 Месяц назад

I don’t understand the doubt about Tesla’s FSD sounds like a lot of prejudice. Now Prejudice is making a decision based on feeling not fact. I think someone once said “ Using prejudice is a fools way of reason.” I taught my daughter how to drive a few years ago she drives very well now. If you teach a computer how to drive using AI it will become substantially better day in and day out due to uploads. The amount of experience will compound. I was a commercial driver who has driven over a million miles. There is no substitution for experience. The more miles on the road the more safe you will drive. Maybe the doubters don’t have the drive time. Knowledge always trumps intellect. Is it better to be smart or wise? Wise up people.

@Meatball2022 Месяц назад

Waymo has at least two dozen miles of data. Clearly they’re ahead….😂

@EinzigfreierName Месяц назад

Waymo has done more than 1 millions miles of driverless operation. Tesla so far has 0. Yes, Tesla has a lot of data. Most of it is redundant. And for the area where they operate, Waymo probably has much more and much more detailed data.

@Meatball2022 Месяц назад

@@EinzigfreierName sure they do. Gathered by their massive fleet. Ok 👌🏼

@tribalypredisposed Месяц назад

@@EinzigfreierName LOL! Every mile that a driver doesn’t intervene is the same as “driverless operation.” But what Tesla LEARNS FROM and improves by analyzing, is the interventions, the interventions Waymo doesn’t have…because no drivers. Waymo is trying to generate data artificially with computers, but no humans will intervene in computer land either. Waymo has been doing great work, but their basic approach was wrong from the start and is a path that never leads to FSD level 4 or 5 that is scalable and economically competitive. Tesla is working to perfect iPhone 30 and Waymo is working hard on their Blackberry…

@GET2222 Месяц назад

@@EinzigfreierNameand WAYMO is locked in their garden with limited access to anything that isn’t GEO FENCED. No way China ever lets them operate there. No way in h-ll. WAYMO costs north of $150k per car and then they have a fleet of people driving them when they are stuck. It’s not scalable. That is all that matters. It’s not scalable and it never will be. The ALL IN PODCAST had the Uber investors on and they broke in down in great detail on their podcast for about an hour. Tesla has half of a trillion dollars of robotaxis on the road right now and every year it grows by more than $100B in product that Tesla doesn’t have to buy. It takes years for WAYMO to make their money back on one robotaxi, YEARS!!! The hardware and infrastructure to support it is way too expensive. Starting in another city and like starting all over again. Tesla is working right now in other countries and including China. Mind numbing. WAYMO will have to pivot to AI and neuronets and then they don’t make their own cars. They have no way to scale their product and will never be able to meet teslas robotaxi rate be cause their nearly is no infrastructure needed because people own the product and service the cars. They went on to talk about how brilliant it was. Go watch it. Nothing of what you mentioned matters. It can’t scale. There is no route for it the beat Tesla on cost. Why do you think China wants robotaxi? Because it’s scalable and 5x cheaper than their cheapest solution that is government subsided. Imagine that.

@EinzigfreierName Месяц назад

@@tribalypredisposed Waymos approach works for a robo taxi. Teslas approach has not proven this yet. I don't really believe in "vision only" for example. And no other company beside Tesla does.

@vermontsownboy6957 Месяц назад

The ARS Technica article is likely incorrect about Tesla's robo-taxi road-proving ability: If it proves necessary after Tesla rolls-out the robo-taxi model, nothing prevents Tesla from remote monitoring and oversight using the Tesla's on-car video camera feed the same way Waymo does today. Chess versus checkers? We'll see how that goes.

@jalexand007 Месяц назад

It is not always about a degree but experience. I would take a super smart guy with hands on experience vs some book smart guy.

@steve2007 Месяц назад

Tesla is attempting to solve general ai vision to understand the real world, that can then be implemented across multiple platforms. Not just in its cars. For that reason alone, the vision only approach seems to be the logical way forward. Tesla seem best placed to achieve this given the amount of data they have access to. On a personal note I find it strange how quick people are to throw shade on this approach. Rome wasn’t built in a day. We should really be marvelling at the speed of progress and get behind the potential this tech has to save lives, ease congestion etc etc If it’s come this far already then it’s going to happen.

@cleanergyjr9701 Месяц назад

Wonder if Mr. LEE would be qualified to work in Tesla or Nvidia.

@tribalypredisposed Месяц назад

I almost guarantee he applied and didn’t get an interview, they can smell a weak mind for miles at Tesla.

@johnpob2486 Месяц назад

Have we seen any Nvidia cars driving around, gathering data?

@daveinpublic Месяц назад

You’re supposed that Waymo doesn’t run on highways? You may be even more surprised to learn that Waymo has started running on highways in limited quantities.

@joewilder Месяц назад

I thought Neural networks don't need to label.

@tribalypredisposed Месяц назад

They don’t need human labeling, and they don’t label in a way humans can recognize or describe even, but they learn what to do at stop signs without saying , look, that is a “stop sign”

@ernestkumeh3800 Месяц назад

Is there possibly a bias by the Nvidia CEO because he is selling his GPUs to Tesla?

@Zripas Месяц назад

What would have change if CEO said that Tesla is behind? What would Tesla do? Buy AMD GPUs? They are buying Nvidia because its the best option to train AI. On top of that, Tesla does use AMD gpu's in the cars itself.

@MrGiggity890 Месяц назад

Timothy Lee is not an expert in AI

@xMercuryx56 Месяц назад

nope but he has a masters degree in computer science from princeton so he must be more credible than the ceo of nvidia!!!

@tribalypredisposed Месяц назад

I mean, obviously Tesla is way ahead, Waymo is making good progress off climbing their mountain on island LiDAR. Meanwhile on the mainland where solving FSD is possible, Tesla is the only one climbing FSD mountain.

@doolittlegeorge Месяц назад

I would think cell phone device providers are the "furthest ahead" as they have the least demanded of them whilst driving yet most demanded as part of their feature set namely AI enabled cell phones. Still waiting on Microsoft to come out with a mobility device as has been promised! Microsoft hasn't even revisited their .net platform to clean the mess up that is Internet 2.0 still currently. 5g still is total junk by way of specific example. As for the automotive market "Tesla Vision" remains unique and is an apparent item anyways with only Rivian competing with that. What that means as far as an entire Automobile being a "mobility device" I have no idea but laptop prices sure are starting to go off a cliff/down going into the Dread Back to School Season.

@mymodely7755 Месяц назад

My Tesla almost drove into cones on the highway when I was next to a semi. I’m on .3

@tribalypredisposed Месяц назад

But the highway is still lagging, apparently 12.4 will finally get highway driving a big advance to neural net programming.

@mikus4242 Месяц назад

@@tribalypredisposedIt’s just one year away……still.

@JayEs31 Месяц назад

Obviously the arstechnica guy is comparing Wayne today to Tesla today and not the versions in beta and that will be released shortly. So it’s not correct analysis.

@Noaixs Месяц назад

Tesla has a solid auto labeling system. Why? Because Elon said they were compute constrained, but they are not compute constrained anymore. He never mentioned they were data constrained, so that is a logical conclusion

@MrQuay03 Месяц назад

No more labels, it's video in, driving out now. In fact, the AI industry moved away from labeling

@Noaixs Месяц назад

@@MrQuay03 Yeah Self supervised learning SSL is the future

@mukamuka0 Месяц назад

Waymo car need a constantly monitor by operator. They will never be able to operate in place with low connection

@EinzigfreierName Месяц назад

That's the case for all robotaxi services. Tesla will need something like this for sure if they ever start robo taxis. Or do you think a system like this will ever work 100% perfectly? There always needs to be a team of humans in the background to tell the cars what to do if they are stuck.

@pennytrui1149 Месяц назад

Um waymo doesn't need connection to drive it has all the data already stored in it's database and the computer is inside the car. Even if the whole world didn't have wifi waymo would still be able to drive lol. That's the beauty of premapping this is why waymo knows where stop signs are even before it sees a stop because it knows oh this area is already mapped etc

@mukamuka0 Месяц назад

@@pennytrui1149 Dude...Waymo just showcase about it on their own website....Their operator uses camera and sensors to monitor and instruct the car. It's a requirement to have a connection

@agildehaus Месяц назад

@@mukamuka0 It's a requirement because Calforinia autonomous vehicle code mandates it, and rightly so. Plenty of situations where these cars, including Teslas, get stuck and, as a robotaxi with no human in the seat, need assistance to get unstuck. The long tail is long. Every robotaxi also will have mechanical issues, or software issues, or will be hit by another driver. This requires appropriate remote monitoring. In some ideal world where no problems ever occur, sure, you can not have it. But that world doesn't exist.

@EinzigfreierName Месяц назад

@@pennytrui1149 They need connection to their operating center in case the car or the customer needs help. But that's the case for every robo taxi service.

@jrsands Месяц назад

“Waymo is playing chess while Tesla plays chess” What an ignorant statement.

@MrQuay03 Месяц назад

What the f does that mean?

@chungloankenny Месяц назад

Tesla is playing ""checker ""not ""chess""

@MyUniversalUniversity Месяц назад

Take a waymo car, send it to France and see if it will work. Take a waymo and drop it in PA in the US, it won’t work. It want work anywhere that is not mapped and geofenced. A degree does not make you an expert😂😂😂😂😂😂😂😂😂😂😂😂😂😂!! Experience and basic logic can!’

@pennytrui1149 Месяц назад

Um it can work lol but for safety the premap everything bruh waymo barley makes any accidents but if they make one everyone is like waymo is unsafe lol 💀. I'm pretty sure waymo can drive without mapping anything but would it be saver ?

@Zripas Месяц назад

@@pennytrui1149 You have no actual evidence or examples of Waymo working anywhere else besides heavily premapped areas, so please, explain how you got to conclusion that it could work?

@mtwn Месяц назад

Tesla is the best in terms of cost/scale

@Agent77X Месяц назад

Lidar is needed said FSD experts!

@Zripas Месяц назад

who are those "experts"?

@glennthomas7435 Месяц назад

How many times has Waymo driven on the streets of Boston??

@davidpearn5925 Месяц назад

In geolocated America........ possibly. Beyond...... irrelevant.

@JohnSmith-ug5ci Месяц назад

Self Driving Cars should be illegal on public roads.

@saff226 Месяц назад

Why. The data shows they are much safer

@JohnSmith-ug5ci Месяц назад

@@saff226 your claim is false.

@Zripas Месяц назад

@@JohnSmith-ug5ci And your claim that this claim is false is based on what evidence exactly?

@cpangws Месяц назад

Waymo Driver operates at Level 4 autonomy, which means that no human driver is needed in Waymo’s defined operational conditions. In other words, Waymo’s self-driving technology is fully independent from a human driver. Unlike some other autonomous systems, Waymo’s vehicles do not rely on remote human teleoperation for control. So, all in all Waymo is way ahead of Tesla FSD Supervise.

@nguyep4 Месяц назад

Waymo are not profitable and cannot easily scale. The system falls apart when conditions are not ideal, ie, poor weather, poorly mark lanes, or none mark roads. Needs much more expensive and multiple hardwares. If that is leading sure.

@Zripas Месяц назад

If its fully independent, then why it has remote driver which will take the car out of stuck position? I mean, you can just search for waymo videos on youtube, there are plenty of those where Waymo gets stuck and passenger needs to call support to get that car moving.