Nvidia Drive vs Tesla Full Self Driving (Watch the reveals)
HTML-код
- Опубликовано: 17 сен 2024
- Nvidia and Tesla are both building AI platforms to make self-driving vehicles a ubiquitous reality. Watch Nvidia CEO Jensen Huang and Tesla CEO Elon Musk talk about their companies respective technologies and how they compare head to head.
#nvidia #tesla
The most important part of building for autonomy is wearing leather jackets
obv
Elon probably used it while talking about the topic to look like Jensen whose company helped Tesla in building the initial versions of their system. 😉
Michael Knight cosplay
Well..
Duh.
Gotta ride in style.
The interior shots from NVIDIA show Jensen in the S-Class on a geo-fenced demo drive....Tesla is in another dimension
Not even comparable - Nvidia is trying to seem competitive with their carefully-edited demo.
It's a terrible demo, but you can watch hours worth of Cruise trips on their youtube channel which are impressive; although that's a lot of proprietary stuff I assume built on these Nvidia processors.
@@updlate4756 You a comedian?
@@wheelofcheese100 Yes, why?
@@updlate4756 lol
Wow! Nvidia has a professional driver testing a single car on an empty road with staged obstacles and pedestrians, amazing!
Tesla has thousands of stupid, randomly selected (like me) drivers going for a ridealong on every random street and road in the world.
Hmm, wonder who is ahead?
One company started recently, one has been at it for a long while now. You tell me
@@Mark-zg4ky agree
Elon is a broadly gifted entrepreneur, business savvy, scientist that listens to thousands of other ideas, theories and experiences then melds them, discards some and comes up with ongoing improvement. Some call this continuous improvement but Elon adds a steroids to this, as he is multi talented AND DRIVEN like almost no other, perhaps since FORD, Einstein, and a few others.
Tesla is obliviously way ahead
@@Max-tj5fv is it just me or do you guys not get my satirical tone?
Nvidia needs to release a video of that drive without any cuts. Cutting to drone, and back in the car, and forward in time leads me to believe there were interventions they don’t want to show. Both Tesla and MobilEye have a video of their vehicle driving through more confusing streets without cuts. That likely took many takes to get right, and a specific route that the car can do well. But if Nvidia can’t even find a route the car can do without interventions even after a few attempts, then I’m very not convinced about their technology.
Nvidia is late to their self driving
Weirdly GM Cruise in San Francisco busy streets is one of the best examples. They do have run into problems and stalls but it shows it.
And the test them in heavy rain, dense fog, snow, etc etc
Judging Tesla from their in-house videos is just as unreliable. Watch uncut videos from third party testers using Autopilot in difficult city environments instead.
Agreed. Well said
Both approaches are correct, but one is scalable at a reasonable price, the other is not.
If autonomous taxis are going to replace driving, then what's the difference between a $50k system and a $70k system? (presuming the difference is even that large, which it likely won't be with scale) Not much if the taxi is in service for 10-15 years with tens of thousands of fares. What matters is which system does it smoother, faster, and without incident. Getting to market faster certainly helps as well.
@@updlate4756 so far, price is almost double. Other than that, lidar needs hi definition maps. Can you imagine how much it would cost to map every road on earth? Tesla approach is more difficult, but when solved, it doesn’t require any maintenance and it’s more scalable.
ONLY, Tesla has the BEST publicly deployed Autopilot.
Tesla has the BEST compute , of the Two competing systems.
and its Seemlesly intergrated into EVERY car they make.
NVIDIA cant intergrate its cameras & sensors into the cars.
Nvidia does not build cars.
the Cars that Nvidia uses , cost $100,000 , inc cameras & sensors..........lol.
@@updlate4756 lidar used to be expensive but now its like 500 dollar for a setup. So yeah have both is better
@@robertjonker8131 True, Lidar is falling in price, but lidar is only used to measure distances, which cameras can do via calculation, albeit at slightly lower accuracy. Lidar isn't bad, in fact, its good, but the price, the bulkiness, and the fact that it will be obsolete in the future is why Tesla has never used it. Again, you aren't wrong, you are just not looking at the long-term picture. Think of it this way: You can walk, play sports, drive a car, and ride a bicycle with only your eyes (camera) and your brain (AI/neural net). Having a lidar sensor inputting distances of objects in your brain wouldn't slow you down, assuming you have the processing power to use the data., but it isn't necessary for your tasks. you can do fine without it, which is the same case of a car.
Oh my God. Somebody call missy. NHTSA must intervene TODAY. The person in Nvidia sedan did not have hands on the wheel and the car did allow it!
@De A Ford is hands free until the road turns a bit too much. I'm laughing thinking of the Sandy Monro video where he test drove it with the Ford rep.
@De A please tell me why "hands free" systems could not do turns without your hands.
Ford, Honda, Mercedes etc. have level 3 approved systems that legally allow you to take your hands of the wheel and eyes off the road while their systems are activated. If you do the same thing in a Tesla and cause an accident, you are criminally and personally liable.
Since Tesla confirmed to the NHTSA that their FSD feature will always be level 2. That means no FSD on any Tesla currently on the road will ever… ever allow you to relax and take your attention away from the road. Which, again, is what some Ford, Honda and Mercedes already allow you to do.
Leather jacket is part of the MVP of FSD, obv.
FOR REAL
STANDARD
That's what you take from this video? Ehehuuheeee!
@@pietskiet8763 NVIDA co developt FSD with Tesla at the begining of tesla
@@AndrianHendrawan I'm talking about the c@#&* commenting on leather jackets!
ultmately all vehicles will be autonomous and Tesla will be the most cost effective and available to far more people far faster. I have the FSD Beta and can attest that it os doing some amazing things. I think it will be ready for general release in a year... a year later to run at a level to operate robotaxis. Even if I am being overly optimistic I see no way for these heavy mapping lidar systems to compete.
How is your over all experience for Tesla ?
The problem with FSD beta is other drivers on the road ..it makes a decisions based upon safety which is the exact opposite of human drivers
@@bencrilly209
When I read their annual report they said
We will improve it with our data analysis? As much as data Tesla FSD gets it will improve.
But I think Tesla is far ahead of other carmakers in terms of technology & the prices of auto-driving. What do you think?
Why would you think Tesla would be the most cost effective in the long run? They may end up with the largest market share, that is to be seen. They will incur a great deal of expenses to recoup in doing so. It's very likely that someone who comes late to the market will be the most cost effective. That's typically how it works. Tesla will most likely be like Apple is today producing very expensive premium products that will be among the least cost effective products in the market place serving more a status symbol of one wealth.
I have access to FSD BETA as well...there is no way FSD will be ready in 1 year. IMO the problem with FSD is that its hit or miss sometimes, i feel like the cameras are not enough to detect its surroundings every single time reliably, i wish they added a lidar system. Sure sometimes it works flawlessly but sometimes it shocks me how bad it handles certain situations, its the inconsistency that makes me doubt that actual FSD will be achieved within the next 5 years. Either way, even if cameras turn out to actually be enough, a certain level of redundancy in sensors (f.e. lidar+camera) is what eventually will be mandatory for governments to approve of FSD. Its like in aerospace where you always have a backup sensor/system for vital components in case the primary one fails.
Tesla has more data, more beta trainers and over a million vehicles on the road with the required hardware installed in every vehicle. The cost per vehicle is much less and already worked into the price of the vehicle! They are already making profits on purchases of FSD and their subscription model! Once they perfect FSD game over IMO !
@@user-pz6mc2oj3f Wow good point! Any examples?
Tesla will not be the only company with FSD, but they will be the company making the cars. At the end of the day Tesla can give away FSD to drive vehicle sale and drive everyone else out.
@@AustinNorman *repeats louder* nvidia better tech ugh
Since Tesla only uses camera there is a limit to how far they can go. More sensors (aka data inputs) is essential to make a general purpose driving system. A combination of cameras, sonar and LiDAR will be superior compared to a subset of them. Not to mention redundancy.
@@user-pz6mc2oj3f hahaha... Tech.. Are you serious ? What maters for AI training is data. Compare billions, to millions of miles.... ordres of magnatude of difference
Could you imagine if how well your car drives was based on how good of a graphics card you could afford? I don't want anything less than a 3080 in my car.
That's the problem. That makes it way less scalable.
Unfortunately, all that’ll be available is cpu integrated graphics:)
3090 SLI FFS!!!!!
🤦♀️
Extremely exciting to see these two companies working on a problem which will improve road safety
Maybe, maybe not, but the incentive behind autonomous driving has nothing to do with safety. Autonomous driving will track on monitor people on a daily basis. The end result will take away car ownership.
@@danieldeblasio9368 there are multiple incentives at play. Consumer incentives are safety, reduced insurance rates, and convenience. Major auto and tech players are primarily focused on launching an AV taxi service. Some ad tech players are focused an add'l consumer data, but the opportunity is generally considered much smaller than an AV taxi service. Much of the new data is not additive (e.g.,, Google Maps already knows where most people are going)
@@The_Stanman No, a consumer has no incentive. The rich want to control your movement and take away your driving privilege.
@@danieldeblasio9368 that's fine if you feel that way for you, but it's a fact is that many consumers (including myself) are choosing to pay for ADAS safety systems. Car fatalities are ~20% lower with even basic systems. 1 million people die a year and some people want to lower that, even if you personally don't value it.
@@The_Stanman "Any society that would give up freedom for security, will lose both and deserve neither."-Ben Franklin
The polish of the NVIDIA clips compared to those of Tesla conveyed some interesting messages: (1) NVIDIA was addressing the public and potential customers, while Tesla was addressing potential job seekers to make development more rapid. (2) The NVIDIA trips were equally polished, either reflecting how perfect their system works with the help of expensive sensors and HD maps, or how reluctant they are to be transparent about their actual rate of progress. So, who will you trust with your life? Those striving to improve as rapidly as possible, then "chase the nines" for continuous improvement, or those convincing you they're already there?
Nvidia is facing designers, manufacturers and engineers. They're selling the dev tools and environment. They won't sell cars.
Tesla is facing the consumers. They're selling cars. Tesla's initial versions were built on a previous Nvidia platform. They designed their own chips only later on.
Exactly. As a share holder I am holding both or merely forever.
Don’t bet against Elon!
Just appreciate that there are many people solving this problem and making a better future.
Nvidia has a very strong marketing team and the buzzwords alone help the mercedez stay in lane. Which is basically the only thing they have actually shown, a car keeping the same lane. Everything else was drone shots. I really wish they went into more detail about what their tech can see, because even beyond hype, nvidia's ai team is world class!
Tesla does no marketing, but it’s millions of customers and fans do. 😅
@@stebarg All large companies have strong marketing teams. Who do you think created the video promoting Nvidia?
im 100% sure the AI capabalities of nvidia far exceeds tesla tho. Nvidia are masters of this field, even AMD and intel are far behind.
@@Hehehe-hf7rq Tesla once used NVidia chip, but they decided to invent a new one from scratch, because of lacking performance or/and high power consumption. The real world data is the most important thing when training a NN.
@@Hehehe-hf7rq not all ai is the same, everyone is kinda 5 years behind Tesla on the self driving, but we’re still 10years away from it actually being a thing so anything can happen
With respect to debating which system will prevail, for me personally, I trust Elon / Tesla to do the “right thing,” His mission goes far beyond corporate profit.
If his mission goes beyond profit, why does he charge extra for the chance to beta test FSD with the implication being he'll charge future customers even more?
I'm looking forward to more from Nvidia. It looks really promising.
That being said... Nvidia's demo lacks credibility.
At first blush it performs similar to Tesla's FSD beta, but the Tesla footage is from public beta testers who are putting the cars into myriad unfamiliar situations. We know for a fact that it's not staged, and we know where the system works well and where it falls down. Nvidia, meanwhile, is presenting carefully curated footage. We have no idea why they chose that particular route or how many takes it took to get it right. It's impossible to judge from the demo how well Nvidia's system performs in real life.
Tesla's system is correctly predicated on the car seeing a road for the first time, every time .
It can take account of roadworks or wide loads that weren't there earlier, and doesn't rely on being told
" what " to think, so much as " how " to think .
And therein lies the secret .
Well, Mercedes has reached level 3 FSD first. Tesla is still at level 2.5, so....
@@silimarina. Waymo has (I guess) level 4, with an operating robotaxi fleet.
That doesn't actually mean they're more capable. Rather took a different strategy and cut down the scope of their service until it was limited enough that they could deliver their product.
Same with Mercedes. Their "Level 3" is limited to certain situations.
Nothing wrong with that, but it means you can't use these ratings to compare technical capabilities.
Making self-driving car isn't Nvidia's core business though. Doesn't mean they can't do it but they have a lot of other stuffs to do, GPU for example, and they'll have to support hundred of cars model, unlike Tesla who put self-driving as their core business. It's unimaginable that Nvidia will be able to match the resources and effort that Tesla has throwing at the self-driving.
You know Nvidia provides Tesla with the GPUs for their Self-driving system?
@@TenshiR no anymore, but nvidia main goals is to sell chip, not software or AI programming
@@TenshiR Not in the car, not in the server. Tesla is transition out of Nvidia chip and technology.
@@TenshiR not true. Was true years ago. Tesla developed its own hardware and software
I’ve been waiting to by a graphics card for my computer for like 2 years…maybe they could concentrate on that 🤣
Waymo has a system that works in a controlled environment and will allow driverless in a set area. Nvidia will have a high end system that will probably cost $25,000 and be useful for high end cars. The Tesla system has a long way to go, and lots of anti-Tesla regulators to navigate through, but is the farthest along overall. The US government won't let Tesla drive autonomously without at least one other competitor who will be in the same space at the same time. I think the more companies that are working on a level 5 system the better the chances that Tesla will be able to actually go driverless. I also believe there is room for multiple platforms in the same space and that fundamentally computers will be far safer than humans ever were at driving.
Tesla is at lvl 2 now though
Tesla's advantage is its data collection. To solve generalized autonomy, you need to use vision neural nets, because all other sensors just aren't enough to cover all normal scenarios. And neural nets need good data to train on, for as wide a palette of scenarios as possible. That is why millions of miles recorded is a relevant metric to track, because all brands can do simulations, but real world quirkiness can only come from data. And Tesla's fleet is by far the leader here. In spite of this, their system's functionality is only level 2 currently and still not really reliable to the average consumer off the highway. The geofenced approach may win out on the city-specific taxi business though. Once Tesla solves it, many will follow, I think.
Driving will be like gaming now. If you drop frames you might lose.
Nvidia got years to get to where Tesla is at now and Tesla will so fair ahead.
And Tesla is doing it with hardware on a car you can buy today.
@@nonconsensualopinion Dojo is a hole other animal and when it's up and going everyone else is going to get left in the dust even more.
@@BrianPaine82 FYI dojo uses Nvidia chips. And in case you weren't aware, Tesla cars used to use Nvidia chips, then switched to AMD (prob more of a price issue). Also Tesla trained their self driving software on Nvidia's virtual universe environment. If you actually looked up Nvidia involvement in self driving, you'd see they been doing this for almost a decade, they're just not a consumer facing brand in this space. However, they'e partnered with a lot of companies working on self driving.
@@rahhjur Just because the dojo contains Nvidia chips, doesn't mean that Nvidia has any advantage, in case that was what you were arguing. Tesla designed the dojo and will use it in pursuit of FSD training, and it will be one of the best supercomputers in the world (6th). Sure, it uses Nvidia GPUs, but what significance does that have to Tesla's advantage. Moreover, with a quick search, I couldn't find any source indicating the Teslas FSD's neural net or simulation is driven or in any way connected to Nvidias. If I'm wrong, please do not hesitate to correct me.
@@rahhjur didn’t you watch Tesla AI day? Dojo will use Tesla own chips developed in house. Current Tesla super computer does use Nvidia chips.
Potential external factors for FSD improvements:
- New road lines
- New road signs including QR codes
- Cars communicating with one another to improve forecasting
- Internet communications for live new road events
🤔
most of these would help, maybe so.. but designing a system that requires any of these things to work correctly just won't work because it's impossible to have any one of these things everywhere, let alone all of them 🤷♀️
FSD doesn't work on prefed data so it's all irrelevant to it.
It's vision is realtime no need of communication for knowing new events when car can literally see them
Nice content, please how can I invest or buy tesla stock.
Regardless if Nvidia had problems or not when compared to other competitors. I'm glad to see more companies in the space which will drive this technology forward.
Even if they both could drive my car just as safe.I would pick a Tesla just for the aesthetic look.Who wants to see all the training wheels on a otherwise beautiful car.
I pick cheaper one
Big difference. One is technical the other is marketing BS just like the press wants
I appreciated the information transparency of Tesla techs
Geofenced + desparately low on data collection analysis and processing
for nvidia. nice empty streets, real impressive. why not show inside during the roundabout or pedestrians? maybe cause driver took over?
Settle down fanboy. Nobody hurt elon.
@@onetwothreefour-s1n that the best you got, eh. attacking the orange man.
we need to buy expensive nvidia graphic cards to render npc cars!
there are tons of Videos with Nvidia autonomous cars running in the traffic and roundabout etc.
Tesla is way ahead, because it does not need street mapping like other autonomous drive programs. Also it does not use error prone sensors like lidar. It only relies on cameras and a radar, so there is no big problem in bad conditions.
Musk and Huang share many things in common they even share the same leather jacket. They take it in turns wearing it for presentations.
Nvidia's presentation was bunk, 3090 TI is DOA, nvidia's solution for self driving and autonomous driving is DOA, TESLA has already beat them to it and is a way ahead of the competition. Nvidia "GAME OVER".
One is reading from a cue card, the other recites from memory. You don't even need to see their tech, too know who is putting in the work.
🤣 so because musk has a photographic memory his engineers are better? Do you understand that musk is not the one building the tech? 🤦🤡
@@jeffsteyn7174 not 100%? lol
Clips were shown of the engineers as well, in case you didn't notice.
And yes, his engineers are better.
Thanks for a comparison video. Would have been nice if you showed them in an apple to apples point of view comparison. Thanks for the effort.
Nvidia has no apples to show.
It's a lot easier to drive when there's no traffic on the road, absent of pedestrians, trucks and cars cutting into your lane or slightly obstructing.
All cars will be using Tesla's FSD technology in the future. They are way ahead of the competition.
It is funny whenever it came to a curve or complex path the video switched to outside view 🧐
Tesla and Nivida use different technology for autonomous driving. I see other automakers are also testing different kind technology. Time will tell who is better. But the competition is good for the consumers.
Nope. They use absolutely the same technology. The only difference is that Tesla can't process important information from lidars and radars, willingly handicapping their system.
The Nvidia "vector space" looks really empty while the FSD render shows just about every object, static or dynamic, that you want to worry about.
How can it be that with a slew of sensors, a pre-mapped data set this is not beta-ready and FSD clearly is? It is embarrassing to be honest and I can't see how this system could ever compete against FSD as long as it needs that pre-generated visual space data. That stuff needs to be updated often for an AV with "hyperion" to be safe. FSD just needs a destination and figures the world out as it is.
Worse still the car has to read a 4D map while it anticipates the road ahead....imagine multiple cars doing that and the servers crashing at some point
@@jabulaniharvey Yikes!
Damn that NVIDIA car demo had a lot of cut shots. It's definetly not ready for real world.
nvidia has to do this. If they dont they arent seen or aren’t the leaders of their fields anymore.
I remember when Tesla released fsd when i was in college. I remember how they made their own chip processor and wanted to not rely on third party. At first I thought they were crazy for that. That choice really age well.
After doing a full look back it just shows how fast someone can take your position without you knowing it.
If you dont innovate, you will lose
I see tesla ultimately getting rid of every sensor but the cameras. A.I and regular maps should be able to drive the car. We do it as humans..
Why would they get rid of the remaining sensors of the car? They have already f-cked up by removing lidar and radar.
The more sensors the better. Tesla only need to hire engineers that know how to deal with semi redundant data from multiple sensors.
Wouldn't you be pissed if someone took away your super power of being able to see hundreds of meters in the dark?
Why make the robot us sucky as humans?
@@mackan7086 imagine if you had eyes in the back and side of your head. Sensors only tell you distance and velocity. You can do that with A.I and cameras.
Both presentations are good. Although one is a real engineering talk, the other one is full of corporate jargons.
I used to think Nvidia thought ahead of others but their showcase was extremely unimpressive, while I do think more sensors do have a place (when we get them working well (we have tried far too long)) geo-fencing does not, mapping although it might be useful in a city as a taxi driver (you do indeed remember the roads you've driven on prior, you don't treat the roundabout outside your home as a new challenge every day, I'm unsure if Teslas do) if Tesla vision without mapping does work flawlessly, there is no place for the rest of the technology.
There's no place place for the rest of technology??!! There is certainly no place for Fanboys in real technology!
the dreaded leather jacket man strikes again, this time with automobiles
Nividia should demonstrate how it works in the heavy rain
NVIDIA can run at 500TOPs for AI. But Tesla has capability of only 140TOPs. So Nvidia hardware is much more capable of L5 or FULL SELF DRIVING. It has a open approach on sensors unlike Tesla which is quite practical.
That NVIDIA 2024 chip makes more TOPS than Tesla 2019 chip is not what matters, what matters is how many TOPS you get per W of power.
@@philippeferreiradesousa8673
They are making it under 100W. Twice the tops per Watt than Tesla.
It will be in cars in 2022.
In 2025, they are going to make an SOC with 1000TOPs.
@@mahendrakumarangamuthugane2690 That sounds awesome. I hope most of those TOPS are optimised for low precision neural network inference. Tesla hardware 4 should be out next year too, in time for the Cybertruck. So we’ll see what they’ve been making for 3 years.
@@philippeferreiradesousa8673 Yeah
Like hydrogen FC LiDAR is also fools errand and the limitation is not tech its the physics. They may be viable in other planer where the rules of physics are different but not on earth.
@@Steph1 for Hydrogen FC..the limitation is energy loss in every conversion...you produce electricity then convert to hydrogen , then compress it then transport it then again in fuel cell convert back to electricity...even if we optimize it to insane level it can never compete with electricity stored in battery and we directly use it ..
@@Steph1 the limitation with lidar is, no matter what the roads are designed for eyes, so stop signs, bill boards so many others you will need camera to interpret ...and it's just a matter of time at some points of time lidar and camera will contradict ...what will you do .. so if you look you have to solve vision no matter what...so when you have to solve vision then why have crutches at the beginning as lidar..?
@@Steph1 Also to add on, what other companies are doing is utilizing lidar, radar, and camera vision. Tesla has never used lidar, only using vision and radar in the past. Now all Teslas with radar have had them disabled via software, fully relying on their ai, algorithms, and vision from its 8 cameras. The point that Teslas AI team is making on stage, is that just like you and I can navigate and potentially drive a car using just our eyes, so should a car be able to. They on stage were comparing their work to creating an animal's brain, just with ai, neural nets, and software. Lastly, another reason Tesla is ahead of competitors like other automakers, Nvidia and google (waymo), is because the full self-driving software isn't restricted to a zone that is mapped by the cars in HD, combined with a lidar map. This means that for the competitor's cars to drive, they need it to be regularly mapped by their own cars, on top of all the sensor output. Tesla on the other hand is predicting turns, junctions, intersections as well as the rest with its AI and neural net. You often see it trusting its own predictions more than the satellite map (like Google Maps) input it receives for its navigation. On top of this, Tesla is getting a constant flow of data from over a million Teslas on the road, and an ever-expanding number as well. Any person familiar with AI and neural net training will say that raw data is what matters most.
@@sebastianorye2702 isnt the whole point of lidar to have more a qualitative data source. Tesla could drive a billion km but if its data is of low quality, they can never make FSD work
@@sudeeptaghosh Actually, most systems being used today have a combination of radar, lidar, and cameras. What you do when they contradict is you train the system to search for those cases, in the exact same way Tesla is doing it. In Tesla's case, if the cameras contradict with the human, the human reports the issue, and back home, Tesla adds a special case into their logic. The issue should be clear in that simple statement. Tesla is still relying on humans to double check everything their system does, and then is relying on other humans and systems to build logic to search for that specific case in the future. The issue is that there are always going to be new fringe cases. The point of multiple visual systems is to try and weed out the fringe cases without needing to involve the human at all.
No Nvidia autopilot ever got anyone killed unlike others in this field…
One guy is reading from a screen and the other is talking from knowledge and experience
Yup. One is theory. One is real world. Tesla is 10 years ahead. Let’s hope competition leads to more innovations
@@brandonlong2064 Tesla's self driving runs on Nvidia tech
I also noticed the same thing.
@@brandonlong2064 look at the autonomous driving Chinese companies using NVIDIA tech their doing great
@@TenshiR Tesla used to use NVidia boards, but not for a couple years now.. HW2.5 works fine for standard AutoPilot.. wasn't powerful or tuned enough for demands of full self driving, especially the neural net systems, not just raw or GPU computing power.
So, Tesla designed their OWN Neural Net chips and FSD3 computer, it is in over a half-million cars RIGHT NOW, working and driving on streets (standard AutoPilot, not FSD)... collecting data at scale and feeding back into the training systems... and next, their own AI/NN training supercomputer chips and supercluster w/ Dojo... and a newer even more powerful FSD computer is coming soon.
If you want to test a self driving car, come test it in Nigeria. If it succeeds in Nigeria it'll succeed anywhere in the entire galaxy and beyond
The Nvidia demo is HIGHLY edited and cut up from different sections of the drive. The insecurity and sketch from Nvidia speaks for its self. Oops
The first technology which is always failing in snow storm at first is radar. The cameras are still working, even on our old 2017 Opel Astra K.
You get what you pay for. Not all the radars fail in snow conditions.
Opel has that level of reliability, that a 3 year old car is considered old??
Tesla needs a competitor. It could be NVIDA.
If Tesla FSD shows that self-driving doesn't require radar, lidar and maps then using them shows Nvidia need crutches and aren't yet on the true path.
i like how alot of the comments are "hurr durr leather jacket". excited for self driving vehicles, i would love to just go hiking, then when i arrive at another location, have the car drive to me so i can drive home. future is lit, now just give me universal basic income and this is basically paradise.
ok you lazy democrat
It’s almost like comparing PPT and reality
PEOPLE DONT GET IT , LIVES are on the line .... Compatition only makes it better for the customer ...LOVE BOTH
Tesla more than 2 mil cars on road by end of '22 and 3.5mil by end '23. In Tesla we trust.
I'm here to point out that the Nvidia Drive PX2 was built into Teslas in 2016.
Anyone who uses LIDAR is doomed
Tesla used to use NVIDIA tech. There was a lot of reasons they moved off that platform.
In this game, data is life!
... anyone else pissed off that YT have removed our ability to dislike a video? I used to rely upon it all the time as a gauge to what people thought, rather than accepting what YTs political masters want the MSM to portray.
The dislike button was a valuable tool for communities.
Tesla's looks more organic than the Nvidia's in both the presentation and the demo drive.
Imagine if you will, a world where half the people think the only way to prevent the world from dying is electrified vehicles, windmills and solar panels. There's no other way. But instead of looking for better solutions they decide to just invest in automated driving technology because it's profitable and capturing carbon or creating new energy storage mediums is too much work.
Huang said the two magic words: open, and LIDAR. Team green FTW. ;)
Least we can almost guarantee Tesla will have a mustang car park exit peel out mode (minus the pedestrian injuries)
Does it have rtx on?
yea, it uses more power than the vehicle to drive
at 5:32 Nvidia told us exactly why they wont be able to scale to the entire world
two coffee cups on top of car. good place to put them.
FSD is for Tesla only. NVIDIA is for all carmakers except Tesla. It's a no-brainer who will win.
Tesla?
Well one is a self driving video and the other is a camera footage from a car driving around with the interrior of the car and the driver animated, witch Pixar could have made to look more realistic
Lol I hear them say "building a synthetic animal" and I can't help but pictures cats running into sliding doors.
Yep, I see why Tesla FSD ain't perfect now.
Who is the artist that made the soundtrack at the end of the video?
I love it. 👍🏿👍🏿👍🏿
I think with neuralink and Tesla, maybe Elon wants us to be Pixar's Cars Avatars in real life.
Tesla's is based on real life use. The other is based on preset routes like GMs version. Try again.
The music for Nvidia/tesla made this driving demo so much better thumps up to CNET for making this way cooler and the music sounds futuristic and mysterious
Those who can walk alone have the strongest direction those who can fly alone have the strongest wings 💪 💪 ✊
If you ever see tesla FSD beta testing video, you know this edited nvidia selfdriving is not even in the same level.
Lol pardon me.. but anyone realised during the Nvidia video portion of the autonomous driving, lots of switching scenes between exterior and interior of the car.. in fact, the view time of Nvidia self driving at junctions is suspiciously little comparing to Tesla’s FSD video. What you guys think?
Seems like tech is ready for autonomous
Those yellow turns lines so static as geofenced even if they are not painted on the road
i was looking for nvda stock while watching the video. Just by watching their system "in action" at the end of the video... i realized how ahead FSD and dojo is! may be in 1 year, i will check if they improved anything
I own NVDA and am thinking about selling after watching this.
@@pwells10 in theory its an amazing stock. The only TSLA competitor, and with amazing clients (NIO, OEMs and others). If everyone buys FSD to nvda just not to give more money to tesla, the stocks will go way up... but they seem to be really really really behind in their tech. I want FSD in the world ASAP!
@@pwells10 one market is definitely only validation for Nvidia stock
You realize that Dojo does not exist yet and all the current "AI" training is done on a nVidia supercomputer, right?
Nvidia have an ambitious plan that is building a virtual world wich they are calling Terra 2.0, that uses complex physics. I can barelly imagine the impact of using a world simulator in real time. If it works as it's supposed to do that will be like... the start of a new era. They can just simulate the traffic as they want or put real people to drive in there. I presume that everything in simulated worlds will be processed by an AI.
But if the question is auto driving cars tesla is obviously beyond, for now.
why can’t they show drivers’ perspective for the entire drive for Nvidia Drive like the FSD? How do we know the system is actually driving fully autonomously? This is a fake video promoting Nvidia Drive disguised as a comparison!
Is a RTX 3090 fully installed?
Will anyone point out that the nvidia guy has the same leather jacket as Elon Musk? 🤣🤣🤣
You can see from this video that Tesla adopted the Nvidia technology for mapping at 05:55 in this years AI day
So basically Telsa is developing for their own while Nvidia is expanding it for the broad. Good job Nvidia, giving the tech to our future bright developers.
"Expanding" is a strong word here. NVidia uses generic lidar system and there are quite few companies already doing that and getting decent results, but none of existing company are even close to what Tesla can do it today by using fraction of technology.
Main issue with Nvidia approach are those bulky extremely expensive lidars which cant really even be used on regular cars without introducing extreme amounts of drag. Its just... Not future proofed technology and sooner or later will need to ditch lidars and replace those with something else.
엔비디아는 그래픽 성능을 보여주는겁니까?? 실사진 같이 잘 나왔군요^^ 그래픽이 좋습니다
자~ 자율주행을 보여주시죠?
One day we will use our cars as graphics cards
I believe, the whole secret is how to make a neuron no to learn but to 'intuitive improvise/act' as to fulfill/ accomplish the task of somehow imitating the same capabilities of a human... Indeed, even humans need training as to undertake a certain task... but the way humans learn is more open oriented than that of machine/ deep learning models. Much respect to this work, though!
This is like comparing google maps to the Nokia Here Maps. Both are great, but one is used officially by businesses in the transportation sector whereas the other is used predominantly by consumers
yeah, this system help to solve the sleep-deprived problem on highway. This problem tends to occur much in EVs.
Don’t want a robot driving me around.
Nvidia I know your gathering road data BUT have you thought about gathering data on what a automatic car wash machine will do to that unicorn horn on the roof.... POST a video on that destruction plz lolz
Don't sleep on Intel. Note, that intel purchased mobile eye for a reason. Do your dd
Well marked streets, no traffic . . . People will fall for it. But I think it likely that Tesla has enough of a lead that their performance and success will be good for the win.
0:42 just a reminder for me
This race of DoJo v Nvidia is cool. We will have both. Tesla will own autonomous driving to 70% and Nvidia will provide chips to the rest. Make sure you own both stocks in a ratio 3:1 for Tesla and you will double your assets in 5 years no problem
The Mercedes drive seems highly curated