Tesla vs Waymo SELF DRIVING CAR Tech Compared + LiDAR vs Tesla Vision

XPeng

Can a Tesla with FSD drive across the border into Mexico or Canada and keep driving for you? No? Seems like a geofenced system to me. 😉

If you dig deeper you see that Tesla is no where near Waymo. You actually do not even have to dig deeper. It is obvious from far ahead.
Tesla is really NOT that capable at all. First of all, remember that it is ADAS. Nothing more. Not only technically - reality speaking.
And FSD V12 will not change much at all. Still mostly a marketing gimmick. Teslas system is faulty from the start and will not reach any real self driving with todays fleet.
Not only that, Tesla is overtaken by more and more companies. They are really down in the bottom somewhere. And they will stay there.

Human vision may allow us to drive ,but why make it only as good a a human, make it better. Radar for fog and extended dark. What about when the sun blinds the camera etc?

If there's a self driving system from Tesla in the next five years without a driver I would be very surprised. They're cutting corners. It's never going to work.

Tesla is nowhere near Waymo. I don't think it's ever going to work without Radar.

Ole Foggywatt here is a bit biased …

These comparisons are not helpful. We should only compare apples to apples. You can argue one or another, both have their limitations, errors and so on. We need to wait until the systems are really released and more important ready to sell.

Does Tesla FSD takes the relevant corrective actions when various warnings are ignored. E.g. 1) Turn on hazard lights, 2) Slow down vehicle, 3) Find a safe place / position to stopped while the hazard light is on, 4) Inform relevant authority after 2 more minutes of inactivity.

These corrective actions would be useful when 1) Someone is drunk, 2) Someone fell asleep, 3) Someone hai lost consciousness, etc.

Tesla should also identify various zones where 1) Tesla FSD has been working well with little interventions, 2) Tesla FSD has moderate level of intervention, or location only have moderate level of driving data for training FSD, 3) Tesla FSD has high level of interventions or location has low level of training data for training FSD. This way the driver will know which zone or location requires more, less or moderate level of attention (e.g. in red, orange and green zones or locations).

Tesla FSD should also try to differentiate moving, non-movable objects and objects that are movable, but not moving, around the vehicle. This allows FSD to focus more computing resources to track moving and movable objects.

Agree on camera approach in good visibility conditions, what about fog and snow conditions?
Are the Tesla camera IR capability good enough to see through fog?
Snowing condition beyond reducing the visibility have two additional disadvantages like cover camera lens and hide completely the ground signs. First could be solved but what about self driving without full snow covered ground?

WHAT??? why would anyone go with a 40k-50k solution when they go can go with something cheaper? Tesla is 40k-50k itself. you lost all credibility with this video. thank goodness i am not a subscriber. you think a company like good will invest in the wrong technology then double down on it. oh. they might not target the same segment of the population... thought about that?

Critics believe that Cruise and Waymo are leading in autonomous vehicle development because of their deployment of driverless taxi services in several cities in the US. I consider these deployments as “Faux Autonomy” and not what I call “Ubiquitous Autonomy”. When we learn how to drive a car; we have true ubiquitous autonomy. We can be dropped almost anywhere in the US and perhaps the world and we will be able to drive a car even if we have never driven on the roads or experienced the unique traffic conditions or situations. Our neural networks in our brains can figure it out based on previous experiences and training. The current implementation of “Faux Autonomy” is restricted to specific areas in the city, specific hours of operation, specific speed limits, etc. This is not full or ubiquitous autonomy. It is not even Level 4 autonomy, especially when Cruise has 1.5 operators per car deployed that have to intervene every 2.5 to 5 miles! This is not Level 4 autonomy. This is more like remote operated drones than autonomous vehicles. If regulators consider this Level 4 autonomy, then Tesla has been at this level for years operating across the entire country on all roads and highways. The only difference between Tesla and Cruise is that Tesla doesn’t use remote operators and Tesla’s intervention rate is much better.

I can teach my Siberia Huskies 10 to 15 English commands or phrases such as “Sit”, “Stay”, “Come Here”, “Get Down”, “Go outside”, “Fetch”, “Wait”, “Rollover”, etc. After they have learned these commands and phrases, we might say that my dogs understand English, but understanding English is not the same thing as being fluent in the English language. Faux autonomy is comparable to my dogs knowing a few English commands and phrases. This level of Level 4 autonomy is Faux Autonomy!

Not holding my breath for true FSD. Many promises and missed deadlines has made me a bit cynical re its future.

I am not impressed with FSD. The recent update is so much better. But, it's still dangerous. I have to rescue FSD from slamming into my neighbor's driveway posts, stop it from going around a curb too quickly, and it will brake for no apparent reason. Don't get me started when the weather is bad!! I think there needs to be some level of learning at the car level for it to function well in my environment. I truly want FSD to succeed as an owner and investor. But they have a VERY LONG way to go.

Waymo geo-fencing is not "self driving". Robot-driving at best.

The in-town Waymo product seems to far outperform Tesla’s FSD.

Tesla has a long way to go to get to robotaxis.

All this talk by Tesla fanboys about the millions of potential robotaxis is comical in light of the current performance of Tesla’s Smart Summons — the things can hardly make it across a parking lot to collect its driver, much less independently drive to a rider pick-up point across town …

Every point you make has a counter point for that favors a Lidar system. My suggestion is that you need to try to go deeper into why a smart company like Google that started with just cameras, moved to other censors. BTW.. no way is tesla a better software company than Google.

How does it drive?

Wamo and Cruise will never work at scale. What would it cost to buy a Waymo or Cruise equipped car. 125k 150k vs 35k tesla. And fyi I drive as does every single person in the world using a vision based system. So yes a vision based system works, problem is human drivers can get distracted and or simply suck at driving.

Hey Jon : excellent collation and elaboration of the facts surrounding the $TSLA Call Option as it has been referenced recently . I would add
Waymo have no control over the evolution or modification of the Ipace as a platform - OR indeed it’s longevity .
I am especially keen to understand how Elon will fulfil his commitment to RoboTaxi Income for Tesla Owners : THE ultimate purchase incentive . the idea that owners may fulfill the role of Supervisor in the early phase is interesting as I expect consumer resistance to the idea of driverless - initially. (CookieUK) 😊

IMO the arguments about Tesla's superiority over Waymo are solely theoretical. Waymo has actual Robotaxis working right now. Despite being very cool FSD doesn't actually work autonomously and there is no prospect as it stands of it doing so. So the argument is really "If FSD actually worked then it would be superior to Waymo". In the real, unmapped, uncontrolled world that FSD wants to conquer there will always be edge cases that require genuine intelligence to solve. Countless online footage demonstrates this. Waymo needs a much more restricted and carefully prepared area to work in but it actually works from what I can see (although the slow pace of expansion must tell us something too).
(Edit - just watched a couple of Waymo videos and was reminded that actually they sometimes require remote human driver intervention too!)

Relying solely on cameras to drive because humans primarily use eyes is stupid. The ability to sense things that a human could not is a good thing! The goal is to be better than human drivers. Better sensors are obviously part of the solution. This is especially true when it comes to sensing obscured vehicles to avoid collisions that a human driver could not avoid.

I honestly think this is a little biased waymo has all those sensors because any crash that happens waymo is reliable. With tesla there is more room for reckless testing because even if 100k crashes happens tomorrow Tesla will just say should have been paying attention. While waymo takes full responsibility for any crash that the car makes. It makes sense why they are so careful and don't go on highways yet

The scientific community speaks highly of mobileye, perhaps a comparison in a next video? From what I know they offer different solutions with camera or lidar

Who cares, l want to actually drive my car, not let software do it,
And who's the blame when a self-driving car runs into me 😮😮

Waymo is NOT level 4, not with the words; "limited conditions," "local," and "will not operate unless ALL required conditions are met." Those words are far from a true L4 condition.

The truth is Waymo self driving taxi is already in service. For normal people, we only live and go to places within a city. We don't need a self driving car for a trip from SF to Austin. I will fly from SF to Austin and take Waymo in another city.

We've all seen emergency road work. What happens to Waymo if its hypermap suddenly becomes inaccurate?

This is too long. Waymo car costs $250k and has a range of 135miles. Tesla Model S costs $90k and has a range of 450 miles. Done.

Bosch stopped lidar production for automotive applications....

Tesla has promised us robotaxis. Waymo is operating them already. Not much more to say. If Teslas FSD is so superior, why are there no Tesla robotaxis on the road anywhere in the world? Yes I know .. they are "just around the corner" - for four years now.

Why does no one take about Mobile Eye? They appear to be heading down the Tesla path with a large fleet of OEM vehicles having their system installed and therefore garhering the sorts of data volume required for that NN approach.

Musk is betting that the gap between waymo and Tesla is within 2-3 years. It doesn't really matter even if waymo is leading for 3 years. Once Tesla's FSD is successful (let's say 2026), there will be about 10 million vehicles on the road with self driving capabilities in an instant with OTA. Any competitor, including Waymo, that wants to mass-produce and roll out 100,000 LiDAR-based driverless cars on the road within the 3-4 years of the lead is a fantasy. Musk changes the requirement from "when can we make self driving cars at the earliest" to "when can we mass-produce self driving cars including existing produced cars at the earliest," After redefining the problem, it is obvious that the pure vision solution is the only answer.

BTW, Tesla has 5M+ cars on the road to collect data. Even a Tesla user didn't buy FSD, Tesla can still run FSD as shadow mode, and collect road data and user controls.

Promising performance of Version 12 is useless. We will not know until it exists publicly. Musk is NOT a reliable source. Never has been.

This report was disappointing. First, it reported extensively on Tesla's opinion about what is required to render self-driving both safe and efficient. Waymo's approach wasn't given nearly the same degree of attention and quite a bit of that attention was directed towards dismissing Waymo's approach based on opinions formed under the influence of Tesla's widely publicised outlook on self-driving. Jon, that is a serious deviation from impartiality and an improper way to report this matter.

Also, did you check your understanding of what Waymo is doing with the data it collects? No, you didn't. You just pretended that all of that data must be 'confusing'.

Guess what, you have to study the data in order to work out what is useful and what isn't. That is a major flaw with Tesla's live beta testing and the impetuous bets its CEO makes without evident push back from Tesla engineers.

Companies like Waymo don't love Lidar irrationally. Lidar is a sensing device that serves a purpose. First, it provides a redundant source of data for the self driving computer and redundancy in data sources is connected to safe autonomous driving irrespective of whether Tesla fans like to accept that or not. Second, Lidar data doesn't just pointlessly feed into the maintenance of overrated HD maps. On the contrary, it forms an informational base for rapid vehicle response in dangerous situations. The self-driving computer can use Lidar data to build that 3D vector space that Tesla crows about more straightforwardly than Tesla does without dependence on analysis of 2D images.

Tesla is a vision fixated company. Insisting that a self-driving vehicle computer should 'be like a human driver', viz. should start from visual cues alone, just appears to be an arbitrary and constraining imposition on autonomous driving computer development. Would an autonomous driving computer having Lidar, Radar and other data sources at its disposal make a better driver. I can't be sure that it would. And, having declined to collect the data, Tesla can't be sure that it wouldn't.

Waymo is really more competition for Uber and Lyft.

I've been taking Waymo in SF without a safety driver for a few months now. I took cruise as well although they're now suspended.

Looking forward to V12 as I think that is likely the path to level four.

The latest Huawei lidar self driving is looking pretty good in China, fair do’s to them👍

My concern with cameras is mostly related to poor visibility scenarios. I wonder if FSD robotaxis will be limited to good visibility situations?

What is happening with George Hotz FSD? Elon tried to hire him

FSD is very, very good. I've been using it for 3 years on every drive. It's true that FSD doesn't rely on HD maps, but it does rely on sparse street maps. These maps contain errors and are never completely up to date. This dependency on map data causes FSD to make consistent mistakes where the map has bad or missing data.

You are completely biased towards tesla... Your research has so many flaws in it.

All of the lidar-based autonomous rivals require expensive modifications, and they aim only for robotaxi operations.

None offer autonomy for personal vehicles.

That is a stark differentiator. Only Tesla intends to release self-driving for both personal vehicles and robotaxies.

This differentiator isn't as significant as the others you discussed. But it's real.

Excited to hear youre thoughts

Seriously! There is no comparison between Tesla FSD and any / every other self-driving technology. I mean they are two different creatures!
When it comes to which is better, it boils down to economics! It is impossible to wireframe the entire globe, but Tesla's artificial general intelligence robot can go anywhere!