Exploring If Nvidia Can Beat Tesla's Self-Driving Cars
HTML-код
- Опубликовано: 10 сен 2024
- Elon Musk is the world’s richest man because Tesla is close to making full self-driving cars a reality. He’s making sure that #Tesla has the best sensors, the most data, and the fastest AI. BUT there’s just one problem. Elon Musk isn’t racing against any one company. He’s racing against them all. #Nvidia ( #NVDA stock ) is the world’s leading AI company, and one of the best stocks to buy now. Thanks to what they JUST announced, the race for full self-driving could be MUCH closer than you think. Want to know who’s REALLY winning? I’ll show you by looking at some of the science behind Tesla stock & Nvidia stock.
📝 Resources & References 📝
GTC Sept 2022 Keynote with NVIDIA (NVDA) CEO Jensen Huang: • GTC Sept 2022 Keynote ...
Tesla AI Day 2022 (TSLA) : • Tesla AI Day 2022
An AV Ride Through Downtown San Jose - Powered by NVIDIA DRIVE (NVDA ): • An AV Ride Through Dow...
GTC 2022 Keynote with NVIDIA (NVDA) CEO Jensen Huang: • GTC 2022 Spring Keynot...
Tesla (TSLA) AI Day 2021: • Tesla AI Day 2021
💬 Join The Conversation & Stay Up to Date 💬
Follow on FinTwit ▶ / tickersymbolyou
5000+ member Discord ▶ / discord
Quick Tips & Videos ▶ / tickersymbolyou
Short-form Insights ▶ / tickersymbolyou
📜 Disclosures 📜
All content on this channel is for informational purposes only and should not be construed as professional financial advice or recommendation to buy or sell any securities. Trading stocks, ETFs, other securities, and/or cryptocurrencies poses a considerable risk of loss. Neither host or guests can be held responsible for any direct or incidental loss incurred by applying any of the information offered. The content of this video is solely the opinions of the speaker who is not a licensed financial advisor or registered investment advisor. Should you need such advice, consult a licensed financial or tax advisor. When you make purchases through links in this video description, the author may earn a commission.
🙏 Thanks for watching!
Is the race for full self-driving closer than you thought? Or do you think one company is firmly in the lead?
Interesting. Thank you Alex (TSY).
No they cant beat tesla but even if nvda can capture a couple % of the self driving tam it would be worth a lot of money the problem is if tesla would decide to liscends their fsd software to other vehicles in 10-20 years +
@@cameronvincent As a Tesla w/ FSD owner. It is both super cool and extremely stressful. Especially when driving in city traffic you have to be ready to take over at any second and it can feel like the car is fighting you for control. I rarely use it now except when driving on the freeways.
Did you really change the title
Ford and VW just shut down Argo AI and pulled out from self driving development. I think Rivian also said that self driving is harder than expected. Wonder if Tesla is so much better or just overhyping what they actually have.
Alex! I don't know what to say. You are doing a unique job that nobody can do in RUclips. You are demystifying AI for ordinary people. I am glad that you are out of academia to share your talent with a much larger audience.
I am grateful and hope you receive the benefit of your effort beyond what you expect.
God bless you for teaching so many people.
I’m hearing a lot of talking heads from ford, nvidia, Chrysler and others talking about the “promise” of their ai products. I SEE Tesla actually doing it. With all the thousands of Tesla cars on the road now generating unfathomable amounts of deep learning training data it’s really hard to imagine anyone out there being close. Even if they are close (I highly doubt they are), Tesla is doing all their data collection with cameras only. Those additional sensors needed by Hyperion will add a lot of cost.
Agreed. i think a big part of self driving is regulations. If FSD requires a minimum of lidar and cameras, this may impact Tesla. Waymo, google's taxis have cameras,, lidar and radar
"I’m hearing a lot of talking heads from ford, nvidia, Chrysler and others talking about the “promise” of their ai products" The vagueness is all so people start talking about terminators destroying the world and other BS, rather than simply observing that the actuality of AI development in most areas is actually really unimpressive.
Finally... a person with a sense of morality and justice
@@archvaldor Google translate is really impressive. I think we'll start seeing more and more good AI implementations in the near future.
artificial data is actualy better.
Nowhere is there a source that accurately compares Tesla to Nvidia ! This is the BEST analysis of the state of art on self driving. Alex you knocked it out of the park. The important point is the other car companies have a chance but only with Nvidia! The BEST investment is in ticker symbol YOU!
Something everyone seems to miss is that just because Tesla is using cameras only doesn't mean regulatory entities will allow camera-only selfdriving, especially in the EU. So it might end up being that Tesla is shooting themselves in the foot by focusing so much on their camera based AI.
I want to give this example from someone who lives in the north of Sweden; most time of the year here is winter, and winter removes all road markings, makes everything the same color, the color white with reflective properties, signs gets covers in snow, wild life enters the roads because the snow is hard to move through and they prefer the plowed roads, falling snow can sometimes be so thick that you can't see more than a few feet and the roads often get narrower and narrower during the season since it's hard to for the plowing vehicles to know how broad the road actually is after a couple of down falls. Under these conditions I don't want a camera based only system, I want radars that can see through the snowfall and lidar that can help out the AI determine depth and distance to objects. It will be way more about each vehicle understanding it's surroundings than multiple vehicles making a collective data pool since the circumstances can change from not even just hour by hour. I really think Tesla is trying to brute force this to save cost rather than thinking it is actually the best solution.
Wish I could give multiple thumbs up for some videos like this. Well done!
Glad you enjoyed it!
Hi Alex I am big fan of yours. Some questions to spark discussion: How much will be the revenue recognition of FSD of Tesla by 2024? How better will be the prediction of pure vision? I think the NVIDIA has a more convoluted stack and does not account on how do you parse conflicting signals. Should you believe radar or vision if the prediction is not clear? :)
Everything comes back to the real world data.
Alex, great content as usual.
As far as Tesla versus Nvidia, I think Musk was right to try and cutdown the complexity of their system. Notice Mercedes is where Nvidia is starting with their systems. I doubt they could make their system cost make sense in a under fifty thousand dollar vehicle. Maybe they will find solutions to bring sensor costs down, but it also the cost of installing and maintaining the system. The more complex, the more problems. I am saying odds of Nvidia truly having their system in volume production vehicle use before 2025 very unlikely. Mercedes will likely announce a delay.
I am not trying to bash Nvidia, we need these self driving systems to work. It will save millions of lives. I am just being realistic about the Nvidia approach being more complex, thus more costs and likely more issues.
"The best part is no part." - Elon Musk. The inherent problem with multiple types of sensors is not just added cost, it is the decision process when one type of sensor disagrees with another type, and how to deal with that in real time traveling at rapid speeds in a heavy metal object. Musk observed that their program worked better without radar than with. And the more types of sensors, the more failure points. With LIDAR, any kind of bad weather like rain or snow and it is not going to work well, and how does your program recognize that?
@@blakejohnson3864 Waymo, so Alphabet, is geo fenced limited, not what is truly promised for FSD, which is get in the car in San Francisco and it will drive you to Los Angele or Las Vegas. Waymo cannot do that. Tesla FSD seems to have that ability, but is still considered beta. Some of the Chinese builders running driverless taxi services, but geo fenced as well.
To my knowledge no one has true self driving that is considered full developed. I would bet on Tesla to be the first, Their ever expanding number of beta drivers has them collecting real world data at levels no one is matching and the Dojo learnjing system is reviewing and labeling data at a pace never seen before.
If cameras are enough, then nVidias approach will end up costing much more per vehicle it's installed in. So licensing Tesla's system might be the road to go, at least for lower priced cars.
I just think it is super unlikely that the other automakers will agree to license FSD from Tesla, given what a threat Tesla is to their existence. For years now they could have ponied up and joined the Supercharger network, allowing their customers to use the great charging network Tesla built. Not one automaker has.
Well I think Tesla will win Self-driving because they are already way ahead of their closest competition, but that does not mean Nvidia will fail. They'll be second place because there is no one else other than these two companies that are meaningfully advancing in self-driving technology. Also, since Nvidia is more of platform company, rather than a car company, it doesn't really matter that much whether Nvidia's self-driving tech is as successful as Tesla's. Even if car companies (Mercedes, Jaguar, Nio etc etc) just pay for Nvidia's Hyperion platform for semi-autonomous self-driving, that still means $$$ for Nvidia's bottom line. Car companies can pick and choose whatever they want from Nvidia's Hyperion platform, whereas in the case of Tesla, they HAVE to solve FULL self-driving by hook or by crook because they are charging $12,000 + for the "promise" of FULL self-driving, which Elon has been promising for half a decade now. If Tesla fails FSD, that will be a gargantuan problem. For Nvidia, not so much.
This a nice discussion of the issue from the aspect of data labelling. However, that is not the only or perhaps main issue.
In particular you characterize Nvidia's support of Lidar only as a potential source of disadvantages..cost multiple sources of data etc...and thereby completely discount the Tesla disadvantage of only being able to rely on imaging. Lidar gives you more and better (3d) data over a larger area and does so even in very low lighting conditions.
Look it can't possibly be that all this money spent on Lidar development is only to make it as close as possible to imaging based solutions..why would Nvidia even consider that?
Also the first announced cars that have at least autonomous highway driving features (the Volvo EX90, the Polestar 3 SUV, and in China SAIC vehicles) are all Lidar based and coming out in 2023-24. Luminar is the company providing the solution for the EX90. Luminar has a partnership with Nvidia which is also involved on some of these projects such as the Mercedes one you mentioned. In addition to Luminar there are dozens of other Lidar companies working on their own partial of complete solutions to either full or partial (highway only) autonomous driving. There are many implementations, some already much cheaper than Luminar's and all of these systems (including Luminar's) are quickly going down in total cost.
The bottom line is that Tesla's main potential Achilles Heel here is their sole reliance of imaging..something you characterize as only a potential advantage.
The problem with NVIDIA simulated data means it never contains unexpected problems, which is the true key for creating a self-driving car. Simulations never completely work in the real world.
You've been delivering some high quality videos of late Alex. The winner will be the company with the higher rate of innovation and execution.
Thanks! I appreciate that 🙌
No, because they don't have the feedback look of constant data to cover all cases and edge cases. Maybe they can copy Tesla and make a simulator but nothing beats real data.
U know real data can't cover all test cases and simulation you can brute force all cases?
I completely disagree.i have FSD beta version 2022.20.18 and I can unequivocally state that Tesla is horrible. Dead last in autonomy today if I were to guess.. Insane to see GMs cruise and Waymo seamless navigate SFO Chinatown but my Tesla cannot understand a lane split. I have send the same feedback 3-5 times on the same road but to no avail.
Anyway it's been almost a year since I have been on the beta and I see zero improvement. In fact even though my 2020 M3 has radar and ultrasonic sensors my dumb Tesla says proudly it's vision only. Really? Their cameras suck, their FSD chips are dog slow and they can never ever solve or even get to L3 with a vision only system..
@@Zmap6191 how many tesla stocks r u short?😅 anyway I'm away from tesla as I hate narcissists
@@aditya234567 the problem is is the human won't be able to come up with all the artificial cases. That's the limitation of a human
@@pjuliano9000
Are you even a programmer? Brute force or backtracking algorithms are written by computer not human🤦 well no point in talking with you. Learn fundamentals of programming
I am not a sucker, but your videos are awesome in every way. You have a knack producing videos that explains complex subjects. I don´t know if you get they already done by the companies, but if you don´t, you can easily get hired to be Hollywood director. I am dead serious. Congrats men, love the flow, 14 minutes goes by and I just don´t see. I love learning from smart people that gives me essencial, since I am a dumb ass in tech. I am buying NVIDIA as soom as the market goes back to 3500 to 3200 hundreds.
This is a very thoughtful exploration of the competitive dynamic between Tesla and Nvidia. It's also worth noting that Elon said on the recent Tesla earnings call that Dojo is a spot of relative competitive weakness for Tesla, simply because Nvidia is so good. He expressed near-complete confidence in Tesla solving FSD, and it appears Dojo is an important part of how Tesla gets there. Dojo as a service sold to others, however, is a game played in a market against only Nvidia.
Great video. Love FSD knowledge videos. That will change so many things in our live.
I own stock in both companies so I want to see both successful.
But I think Tesla’s dropping of ultrasound and over-reliance on cameras is a bad move.
I think these cars need LiDAR, camera and ultrasound.
By the time any of these companies get on the road, Tesla will have 2 years of fleet data, and the next gen self driving chip that is supposed to be released in the Cybertruck with on chip AI features. 2 years is a long way ahead, right now, there are more Tesla EVs with self driving chips than all other companies combined. They just brought their latest dojo supercomputer online and it almost brought down the California grid. Course, some days a toaster could bring down their grid, but it seemed impressive anyway
This is fascinating- it helps deepen our understanding of these 2 companies-and their different competitive advantages-thank you for bringing this excellent content
Seems to me there is plenty of room for both of these companies….and of money to be made by both in the future…..one thing about NVDA approach is that in addition to self driving- it has many other real world applications- making digital models of an engine for example and changing/testing certain variables to see improvement or failures…so many business applications
I expect multiple companies will find solutions to this problem.
Agreed; competition is good for the consumer!
I think the last hurdle for Tesla is to build on top of the AI stack a NN that uses game theory to enable it to properly move through a human occupies space. This also applies to Teslabot as well as FSD.
Why, because other drivers and pedestrians test each other's aggressiveness and patience with different tactics.
This includes edging into traffic, anticipating what another would do and then moving aggressively to get into a opening. In tight congested traffic if one is too passive, then one will be stuck in place for a very long time without progressing to the destination.
The NN will have to be trained to understand the current situation and adapt the most appropriate game theory tactics to win that slop or squeeze into heavy pedestrian traffic without the AI shutting down because it might hit a pedestrian in a heaving crowd.
SDC is just a movie prop. Sure, some people would want that, but most people don't.
Look at airplanes. They have auto-pilot but you never and never will have self driving airplanes. And that is for a reason. Even if the capability is there, people don't want that.
That clip of Elon Musk in a leather jacket is surely just coincidentally in this Tesla vs NVIDIA video 😀
The race is so close, they have to wear the same thing to compete! 😋
loved your content Alex, you should do full fledge documentary and sell it to big media companies such as Discovery, Paramount or Amazon. I think your talent well deserved in building documentaries.
That would be a dream come true!
Nice job. Thanks for sharing.
Thank you. The video was very informative. My long term bet has been Mercedes-Benz. I am encouraged by this German and Japanese cooperation.
Which company is Japanese?
Info above the rest great, thank you.
Hi, thx for your video. Do you know the number of frame per sec and, the number of pixels on each frame for Tesla and Nvidia?
The biggest difference between the two (if/when) they solve autonomy is that Tesla's will be significantly cheaper not only on sensor cost but most importantly on power draw..
14 cameras, 8 radars, 3 lidar's will consume comically more power and hence requiring cars with larger batteries.. Not to mention the CPU's and GPU's that will be needed to process all that data will consume a lot more energy..
There is absolutely no way that bridge is gapped..
Love your content ❤️
The breakdown is amazing
Keep up the great work💯
Appreciate it! Glad you enjoyed it!
What about Waymo? Waymo's already operating fully autonomous vehicles here with their ride hailing app which suggests they're just as far along as most other companies if not further.
Always a great source of clear information about the company’s I care about
I literally subbed as soon as you said "your time is valuable" and got right in. I appreciate it man
Just speaking the truth. Hope you enjoyed the content!
What about Mobileye?
Hmm, I've been driving a self driving car for two years. Maybe you should look at comma ai. After market self driving that i added to my Toyota
brilliant summary, thank you!
Glad it was helpful!
Great, thanks.
I’m betting on Elon!
Why not Zoidberg?
Maybe Tesla can license its Sim' tech and real world street data to video game production houses and TV/film/music video/advertising production houses on a global scale. No harm as a little money spinner with very high margins. A good way to get additional return on investment.
NVidia is the 2nd best performed stock I own. The best one is Tesla. (15x vs. 20x)
brooooooooooooooo, love those ai SiO2 stuff
Well, if theirs don't work, NVIDIA may as well convert all that beautiful looking environment into a metaverse.
I am thanking you for the educTION YOU ARE PROVIDING…THAT ALLOWS ME TO UNDERSTAND THE GREATER CONTEXT SO IF THERE IS A NEGATIVE EVENT IT WILL BE SEEN AS THE FLEA NOT THE DOG
there are more then one way to solve the problem
most people (user) dont care
how technology work as long as it works
The BEST and ONLY comparison. But almost two years old - in AI terms thats two light year afar. Please do an update !!
Will do!
@@TickerSymbolYOU right after 8 aug announcements will be perfect timing ... looking forward !
Even if you disregard the difference in approach, there's the question of DATA, no one comes even close to the amount of data Tesla has/had/will have.
The deeper you get at understanding what is going on the more you realize that all the other companies are still ten years behind tesla even the experts like Nvidia they are really Good at somethings tho still at the start of the race on the rest
They will have a easier time making progress because they can just copy what they can from tesla tho they are allways going to be way behind tesla even if they are in second place
Tesla is so far ahead the videos already the car driving itself , Data on AI vision is on another level and yep real world data of each street on the world where Tesla already drive this atleast 3 to 5 years apart
I have one word for you to remind you of your predictive abilities: WEJO
Well, WEJO was a SPAC that went public during one of the worst downturns in history. Has absolutely nothing to do with the business itself. I still hold all my shares.
Hi I like your videos but I would like for you to explain more why you think NVIDIA is the world's leading AI company. Is it their AI software or their accelerators? Because I doubt their self driving AI is as good as Tesla's for example. Or their Search AI is as good as Google's for example.
Their hardware is what powers a lot of the software you're talking about. It's their accelerators.
@@TickerSymbolYOU okay gotcha. Thanks for the info
Great video. Thanks!
Glad you liked it!
Hi Alex. If you have time, could you make an update on SE stock please?
Thank you and I appreciate all the work you put in for the community!
They all just need to drop some AKIDA IP into their solutions..... and BINGO!
The limiting factor is not hardware.
How will Nvidia integrate all those licensees using their solution into an integrated system to collect data and to distribute OTA updates to?
Using Nvidia will imply an interface between Nvidia sensor and electronics with the proprietary electronics of each licensee. To maintain that will be up to the licencee often a legacy car maker with underdeveloped software engineering.
Very informative! thanks!
I anticipate saving lots of time when my self driving car can wheel me around my work from home job safely negotiating room by room by room without a hitch.
Awesome video Alex....many thanks
Glad you liked it!
That LIDAR tower on top of my $100,000 Mercedes S class sure looks great!
😭 can you imagine
Do you think lidar sensors will be able to communicate on the road? Would be great for safety if all cars spoke to each other while driving.
One of your best episodes!
🙌
Hi there, Have you made a video of Tesla's Dojo?
I think it makes sense to own both the stocks. Tesla is becoming like Apple of auto industry and NVIDIA is trying to do what google has done for mobile manufacturers. But at the current valuations NVIDIA seems cheaper
Well microchip makers have the edge they have the hardware.
Do you use split testing software to change and monitor the results of different titles and thumbnails? or do you do it manually?
All manual right now. Any advice on software I can use besides Tube Buddy, which will only change them every 24 hours?
@@TickerSymbolYOU unfortunately not. It's surprising there's not other solutions out there for youtube split testing (that I know of, at least). It's such a massive platform and split testing can make a big difference in some cases
One problem with the valuation of FSD is that the rest of the economy is also changing. The need to drive anywhere is falling fast. Also, the number of people is also falling! Even Elon Musk has acknowledged the importance of this. FSD is great for the 2000 people in the world that need it.
We live in interesting times
Investing in both!
Also AMD, Intel..
Well Tesla is self implementation where they're marketing it themselves while Nvidia don't make car so they're betting on ecology would support them so it's really whether ecology supports Nvidia
By 2024, Tesla will have mostly solved general self driving. There will still be plenty of edge cases and it won't be perfect, but Tesla has the driver data and nothing can trump that. Tesla has incredible hardware and they are upgrading their cards and cameras, plus they are a vision-only system, if they can function on that alone, they are way ahead of the competition. nVidia is an incredibly capable company, but they can't beat the amount of user data Tesla has, the stuff they're proposing are just promotional materials.
That's all great but if the other legacy automakers cannot sell cars there will be no vehicles to install the self driving
No more stupid people driving in the streets, that is my dream come true.....no more, car accident, expencer insurance, car part, bla bla bla....where may I SING?.🙋♀️🙏🙏🙏Thanks Alex...🥰
Brilliant video! Thanks!
Glad you liked it!
The co with the most data AND the best engineers AND spend the most to get the best results at the short period of time will be the winner.
Thank you.
One is using a dynamic solution and the other a more linear one and wind blowing and a kid on a skateboard with a falling tree is dynamic so only one will win.
Tesla is using the linear way and is why there cars run into jersey walls.
Also Nvidia is in alot more cars and trucks then you think because the company's do not say what they are using they make it sound more like it is something they made but it is easy to look up lol.
Elon has the car and the system growing together, the other is a working relationship with another company
Self driving may not have the consumer uptake that many anticipate because many drivers actually enjoy driving !
Try enjoying in traffic jam 2hour+
Without a large volume geographically dispersed vehicle test bed to reap real use data like Tesla has, there can be no beating Tesla. NVIDIA chips are a data processing tool. NVIDIA can create massive value by sticking to its knitting.
Teslas lead is unassailable I hope.
Good one
Hey Alex, fab breakdown. Your analysis is superb and fully engaging. thanks 👍
Based on this vid, would love your view and analysis on Lex Fs recent interview with Andrej K 😃
I just saw that on my feed! I'll watch soon!
1 Problem. " Elron "
No mention of Waymo and others??
No. They're nowhere near first place.
I disagree. It's not close.
In AI, whoever has the most real world data wins. Nvidia's data so far is just words and promises; Tesla I can actually see it.
Can't argue with that; real data always trumps simulated data
How about companies such as Aurora?
I'll check it out!
Thank you!
I like your channel and hope that you don't drink the Tesla kool-aid. I own a Tesla and paid for FSD and it is still very much a work in progress that can't be trusted to do day-to-day driving. I use it only road trips. Plus by Tesla's own admission with the state of California, its system will never achieve level-5 autonomy. Originally, Elon said that Tesla would have RoboTaxis in 2020, but that hasn't happen yet. You should also take a look at the system under-development from Mobileye.
Tesla is really saying it will never accept full liability for the FSD accidents by saying it will never achieve level-5 autonomy.
Sure his prediction times are off, they always are, but he gets there.
I really see no reason to complain about bete software not being final release software.
Nvidia could provide us with a VR GTA that is a 1:1 copy of reality :0
Great video!
Glad you enjoyed it!
Simple answer: Does Nvidia have more driving data than Tesla? Remember, in AI, Data is King! (Tesla has >3,000,000 vehicles on the road, feeding it data all the time).
Hahahahahaha this cracked me up. Can you imagine nvidia doing this effectively efficiently. They couldn't even get 3090 cards out fast enough and same with 4090 nvm mass production with economies of scale on cars.
Stick to graphics cards and at least try and get them out without a 6 month wait time
Isn’t Palantir the worlds leading ai company
perhaps NVIDIA needs to acquire Wejo. i'm sure the price would be nice right about now.
Yeah, Or the price might be right once it hits 1 cent. No one should forget that he was recommending and promoting Wejo all the way down.
The best commute is no commute
Step one to being successful: get leather jacket.
Seems legit!
As a tesla owner (vehicle and $TSLA) there are something that I would change. None of this is with FSD. FSD is more of a $TSLA issue due to the bottom line. I would love ventilated seats on 3 and Y, able to add APPS on the screen and change the screen like an iphone. Get my Cybertruck ASAP. VW, Jag nad MB are not catching TSLA. On two blocks around me people are moving to Tesla S3XY models and dropping luxury brands. In just 4 homes next to me there are 5 Tesla vehicle.
Soooooooooo ❤what ya saying is that both when hot bottom will be the Best Buy’s at those prices great 😊 news thanks