Waymo has been avoiding the highways because, unlike Tesla, they have to ensure safety and comfort with nobody in the driver's seat. They also were largely avoiding crossing railroad tracks due to this (hard to GUARANTEE that the driver never gets stuck on the track), but they're crossing them a lot now. Freeways are coming.
Tesla FSD system is so impressive. No where near the same level of hardware needed as the Waymo system and in theory it can drive on any roadway in the world.
Well, it can provide a driver assist. In terms of actually driving itself, it can't go anywhere, because it's orders of magnitude less reliable than Waymo.
Based on the video, I believe the Waymo actually had an intervention! They have remote drivers who can take over automatically when the software has an issue. Notice that it did not actually know where to go at 12:11, and that is why you experienced the delay and it stopped. It was asking a remote human what to do. It ended up following a path in that area which did not match the path on the screen, indicating a remote human changed it. I know someone who I believe worked on the system, so they do actually have a way to allow for the cars to ask for help.
They do not have remote "drivers" who "take over". The remote monitors respond to (rare) prompts from the car and confirm the course of action the car has chosen, or in a few cases select from multiple actions the car is considering.
LIDAR, and big cameras on the outside of the Waymo vehicle vs compact streamlined (non-noticeable) cameras on the Tesla and yet the Tesla did better overall. Goes to show Tesla is years ahead. Excellent video.
It's insane that Tesla's system is years ahead but Waymo actually has a 7 YEAR headstart when it was still named the Google Self-Driving Car Project back in 2009. Tesla started from scratch in 2016 with Hardware 2.0 by designing and engineering their own full self-driving computer in-house after Mobileye ended the partnership with them with Hardware 1.
RU-vid Search S: It's not at all clear that either was "clearly better". They BOTH had issues and BOTH apparently needed an intervention. With nothing weird happening, light traffic overall, and almost PERFECT weather, good lighting, etc. Cheer leading doesn't change the overall reality. They're all making progress, slowly. Given Tesla's approach of not geo-fencing, it has a long way to go, but will be more economic if it gets there (widespread robo-taxi level 4 capable behavior, on a practical fare-earning basis). Things like handling parking lots, backing up, turning around, reading basic traffic signs, better construction handling re cones, etc. are obvious things Tesla needs, plus MUCH better consistency re not needing interventions. And I say that as a long term Tesla shareholder.
I’ve got the same version of FSD on my Tesla, and I gotta say… version 11 is a big upgrade. I’ve done some trips from work to home, including today, driving on city streets and the freeway, and it has only been smooth as silk. Maybe they are getting close.
Glenn Millam: When it can do hard routes in hard / deep snow and hard rain without ANY intervention, MANY times in a row with complex traffic, construction, etc, THEN "they are getting close". There are lots of good reasons that real world, progress is incremental for ALL the competitors.
@@rogergeyer9851 True that. Or just operating in rain or direct sunlight reliably, something they are going to need to address before this can be considered truly autonomous. I can only say that the other day, my wife told me for the first time that FSD actually made her feel relaxed. So, there is definitely progress.
@@rogergeyer9851 If you don't focus on autonomous driving, and instead look at AI in broader terms, then it seems that AI has been improving at an exponential rate in recent years. Some of the exponentially-improving AI technology that is used in, say, Chat-GPT or the "paint a picture based on a textual description" AI systems are *also* being used in autonomous driving systems (source: comments made by Tesla at one of the AI Day events during the past year or two). So, this suggests that autonomous driving may also be increasing at an exponential (rather than linear) rate. Here is a hypothetical scenario with some made-up numbers to illustrate this... It has taken 10 years for a particular AI technology to reach 50% of its "good enough" goal, but the rate of improvement during that period can be described as "doubling its ability every 18 months". How long will it take for the AI technology to reach 100% of its "good enough" goal? Many people (and a psychology experiment suggests that "many people" means "most people") find it difficult to rationalize about exponential growth, so they will instead think in terms of linear growth, and thus conclude "if it took 10 years to achieve 50% of the goal, then it will take another 10 years to achieve the remaining 50% of the goal". But the minority of people who are able to rationalize about exponential growth will reason that only one more doubling of ability is required to reach the 100% goal, and since the technology is doubling its ability every 18 months, this means the technology will achieve its goal in 18 months' time. I get the impression that people who claim "full autonomous driving is decades away" are assuming a linear growth in the ability of autonomous driving, while people who claim "full autonomous driving is only a few years away" are assuming an exponential growth in the ability of autonomous driving. I know very little about AI, but from watching some RU-vid videos by people who appear to be experts (e.g., James Douma interviews on the "Dave Lee on Investing" channel), I repeatedly hear the claim that access to *massive* data sets for training an AI is significantly more important than the algorithms used by the AI. Tesla has AI hardware in every car it sells and that AI can operate in "shadow mode" to compare its expected behaviour against the actual behaviour of the driver, and then upload data about interesting expected-versus-actual disagreements or edge cases to Tesla's HQ where it will be added to the data set used to train the AI. Because of this, Tesla can gather AI training data from the few million Tesla cars on the road. In contrast, most/all other companies working on autonomous driving systems are training their AI models on a much smaller collection of data, such as data from 100 or 1000 cars. Because of this (it is argued), Tesla is likely to reach the goal of fully autonomous driving significantly sooner than competing companies. That argument makes sense to me but, as I said, I know very little about AI, so it is possible my faith in the argument is misplaced.
@@CiaranMcHale I think people who say it’s far away are referring to the level of automation actually needed to reach level 3. If Tesla could handle 99% of situations perfectly it still wouldn’t be good enough. The fact that he doesn’t use LiDAR, Radar, or HD mapping is what is going to prevent them from going from 98% to 99.999%
@@bruh-tq2pw "The fact that [...]" That is not a fact, but rather an opinion. I don't think there is much to be gained from us arguing "My opinion is correct and your opinion is incorrect", especially since I am not an expert in AI and hence my opinion is more of an "ignorant opinion" rather than a "well informed opinion" (and I suspect the same is true for your opinion). I think a better approach is to keep an open mind, and watch as competing technologies/approaches mature, and possibly enjoy some popcorn as events unfold.
I think both technologies are pretty amazing. I lean toward Tesla though because I may commute from LA to Irvine or Las Vegas or even New Orleans and it's pretty much the only option. That will very likely change in the future, however even now, my 2018 Model 3 does about 80% of my driving and that's a pretty big deal for me. I know this video is about self driving, but I'll also add that if you're on long haul commutes, the charging network is also a really big deal. This video was great. Thanks for making it!
The progress being made by Tesla is amazing to watch. It has far more human like behaviour, and to think this is all being done with cameras, and with no pre mapping. Amazing. Waymo does fine, but is nowhere near competing with Tesla.
@@thewatcher5822 Yeah I think that a Tesla could theoretically drive with no map data as long as it was given the ability to read street signs, so teslas definitely don’t need to rely on map data like waymo which is great and the only real way to solve this.
@@ian_bruh1 waymo uses what’s called HD map data. The purpose of the lidars for waymo is to orient itself in the world against the HD map data. Tesla uses normal map data they get from Google maps. The Google map data contains things like speed limit info and stop signs but it doesn’t rely on this. Instead it uses the cameras to recognize things in the world. Tesla has shown off that they have the ability to create 3D maps of the world if they wanted but they don’t.
Very nicely done! What a treat to ride in two versions of self-driving on the same route! Boring it may have been for you, but very enlightening for the rest of us! Thanks!
@@RicardoPereira-pk3py that is the `real` comparison .. you consistently get people saying Tesla is last in FSD not even level 2 whilst waymo is way in front already level 4 .. whilst this shows the truth
Marvin, Tesla is not addressing the same challenge. If Tesla _ever_ produces robotaxis, they will geofence. There is no alternative. Think through the many types of situations that can happen. Imagine what happens when someone hits your car. Imagine what happens when an accident occurs on the freeway, blocking all traffic. You must have people on hand to respond to a scene, and it must be in locations that they can reach in a timely fashion. A highway presents some real challenges for that. It introduces extra unknowns, and it makes a lot of sense to not add that risk of complication when there isn't a human driver onboard. Despite what is popularly believed, Tesla is not offering their current owners a driverless future. FSD is designed to be a Level 2 feature, both now in beta and, as Tesla explained, in final release form. They are not working on features for telling drivers about all of the situations where they need to take over. Aside from a few situations where FSD panics and asks the driver to take over immediately, we've seen pretty much no new development in this area. That development would be the clue that they intend to move up to Level 3. Tesla has given us many clues now that the verbal promise of customer-owned vehicles becoming robotaxis is all but vacated. They directly told California regulators that FSD City Streets is always going to be Level 2. They have told all Tesla owners that they will not be getting the new Hardware 4.0 upgrades, including camera modifications, CPU redundancy, and HD radar. It should be clear to us that Elon Musk doesn't add these hardware costs and changes when he doesn't think they're needed for autonomy. And of course, in talking about Level 4, we should keep in mind that the definition of SAE Level 4 includes geofencing. It's an acknowledgement of the fact that there are still things humans need to handle, and thus it's a service that requires human infrastructure to support. FSD is cool and impressive. But it's not a competitor for what Waymo is doing here. It is designed from the ground up to rely on a human driver being ready to intervene when it gets itself in trouble. There is no intent to move past that.
Exceptional video. Very helpful comparison. I like the fact that the Tesla operator was patient, allowing the vehicle to operate safely and based on actual posted speed limits.
@@tv-ld3wv It does when you set the settings accordingly. But it still has sometimes issues with recognizing speed limits. But the progress Tesla is making is incredible and with the single stack now on FSD the progress will be faster than before.
@@tv-ld3wv Tesla is constantly battling NHTSA to let FSD drive like people drive, and NHTSA always shuts them down. Eg FSD used to roll stop signs just like people do, but NHTSA forced them to come to a complete stop at the beginning of 2022, and then they extended the time it must remain stopped in March of this year, as it would only come to complete stop for the barest fraction of a second. It made Tesla leave more of a time buffer for stale yellow lights as well. You also used to be able to set how many mi/km over the speed limit you wanted the car to go, but NHTSA shut that down and now it can only be up to a max of 15% over (eg 35 in a 30, or 75 in a 65). Tesla FSD peoples want it to be indistinguishable from the best human drivers and to not have it drive like a robot, but the robotic bureaucrats NHTSA won't let that happen.
The fresh commentary was surprisingly good (“…we may never know” and “where are you taking me, Waymo?”), but the music track was awesome! I generally find music tracks annoying, at best, but this one is so appealing I want to watch the vid again, or at least more vids if they have similar tracks. As for the meat of the vid, it was spot on! Thanks for doing a great comparison.
Been waiting for your comparison review Tesla Beta FSD versus Waymo FSD Level 5. So cool review and I’m so glad that I’m a Tesla Beta FSD V11 user on my Model 3 too. Tesla FSD is indeed getting closer and closer to FSD level 5 !
Exceptional job. Just watched your Bolt Video too, and you are a great reviewer/presenter with clear presentation and great anticipation of questions. Bravo. Our experience with FSD 11 is pretty good, but it is rare that we can make a trip with zero intervention.
Thanks, would have been great to have a time comparison, how much faster was the Tesla FSD. So roughly 14:40 (FSD) vs 18:20 minutes (Waymo) - the strange maneuver of the Waymo at the end included, as you could not exit before.
Video: "Lets compare these two cars on this route." Me "Yeah lets do it!" Video: "The first car made it in 20 minutes!" Me: "Cool I wonder how long the other one took." Video: "Thank you for watching" Me: "Umm.."
@@Xanthopteryx If we want treat intervention as failure, the Waymo failed too: as other has commented, Waymo has a remote intervention at the end - where it stopped and decided to make a U-turn.
@@joag962 Waymo reached the destination, and the intervention was done still without any driver so it was entirely robotaxi all the way until then, and the intervention probably was just the dispatch point on a map where it should go. The car can reverse like that by itself. Remember that you must ALWAYS babysit Tesla because it will suddenly do stupid things. It can not even follow speed signs correctly.
Ok Wow, I have the 2022 Y pro It's great on drives I can't believe the advancement they've made since I first got FSD beta. I paid 8k Was on order for a year. It amazes me how well it drives. I do a trip with a lot of traffic and turns through a few states very rarely do I need to interact. Each time it upgrade comes out it just gets less and less interaction. This FSD beta the people that work there for Tesla amaze me. Mine blowing steps forward. I truly believe no one will ever catch them They are leaps and bounds ahead. Great video Thanks for sharing!!✌️
Great video! It's good to see FSD Beta vs other systems in real world scenarios. The fact that the Tesla made the entire drive with no interventions is impressive. The fact that it did it, and beat the Waymo by 5 minutes, in the Waymo's very own geo fenced area is mind blowing.
The Tesla did have one intervention where the driver had to press the accelerator for the car to continue at a left turn to get on the freeway. Most of my drives with FSD Beta on city roads require a few interventions and disengagements at least. It can’t handle a lot of scenarios.
Nice comparison. Previous versions of FSD/AP would have failed in that heavy merge onto the highway. Pretty impressive to see a generalized vision-only solution do so well against a geo-fenced vehicle festuned with sensors. Still, there are plenty of areas where FSD struggles a lot, but at least you can use it anywhere if you wish.
The fact that tesla FSD using far less hardware equipment and can technically used everywhere even in unmapped road like a parking lot (Not Geo-locked like Waymo) is very impressive
Alfonso: Tesla isn't rated to handle parking lots. Or turning around on roadways. There are reasons Tesla can't handle robo-taxi duties yet. Advantages AND disadvantages with all the competing AV systems.
@@rogergeyer9851 The bigger problem is it's just not reliably safe yet. But that's to be expected when trying to solve a much bigger problem with far fewer resources. Waymo uses very expensive equipment to drive in very small regions. Tesla is trying to use very cheap equipment to drive everywhere. Takes more time.
First time watching your channel. You do a great job explaining. The comparison video was very enlightening. I've had no experience with Waymo, so it's great to see a head-to-head comtest.
there is only one autonomous car in this comparison. One on a preprogrammed route and one that figures it out in real time. And it doesn't look like a science project. Waymo has been at this for twenty years and the cars STILL look like that. Wow.
dclpgh: For a cab, why in the WORLD do the looks matter? Your bias is clearly showing. And I say this as a long term Tesla shareholder looking forward to practical robo-taxis in my city in the next decade at low prices.
@@rogergeyer9851 More than just looks. Tells you where they are at with their tech. If you still need all those sensors after 20 yrs and the other guy is using just cameras ans software to accomplish the same thing, Whos tech do you think is going to win Autonomy? Tooling up and down the same streets in the same city like a bus line will not solve the 9s and edge cases. And what if your trip is past city limits? I have never seen a Waymo on the interstate. Even with the beta software in the cars, A Tesla will go where you point it. Dont sell those shares bud. Tesla will be the first with general autonomy.
I like to see how the two will perform in busy city center areas, with lots of pedestrians, bicycles and difficult turns... That will be a better test case over the straight planes on highways and suburban areas. Still great video! Thanx 😊
I love seeing these comparison videos. Really shows the different approach and limitations of each system. I am surprised Waymo doesn't take the highway, but I guess it makes sense to avoid it when possible, since they are optimizing for city streets and city travel.
I think Waymo doesn’t drive on highways is that it can’t do so safely and reliably. Otherwise why would an “optimized” system avoid taking the faster route?
SCP Database: Yes, it is interesting. For places with good weather and strict geofencing in moderate sized areas, Waymo is likely converging to an acceptable (but expensive re the support network and hardware) solution fairly fast. I think it's hard to say who "wins" re cost and overall practicality over the longer term, like say out to 2030 or so.
@@rogergeyer9851 I don't think it is hard to say at all. Waymo as you point out is very limited in what it can do and is not cost effective. Tesla assuming they carry on the progress they have been making have a system capable of operating anywhere in any weather and at a cost effective price. At this time there is no one close to what Tesla is doing.
The FSD did very well. As a Tesla owner I'm still not convinced getting rid of radar was the best move. I feel as though FSD would be further along with it than without it. That being said both cars did very well. Waymo gets a check because it at least pulled over to the side instead of stopping in the street once it reached it's destination...although I must admit I thought it was stuck at the end of the street. I'm also very impressed it backed up and turned around to come back closer to your drop off point and put you on the side of the road to allow you to exit. FSD is good but it still has a ways to go.
The old radar was definitely too low resolution and accuracy to be anything but detrimental. I can understand how removing radar entirely was possibly bad, even though they did show their immediate improvements after. Tesla actually is bringing radar back but with a much more advanced resolution and capability that it can be relied on without getting false readings.
I would love to compare driving from my house to the supermarket (10min) here in Queens NY, and from NYC to Florida (20hrs), oh wait no waymo here, Tesla wins 😂 (and Tesla did both short and super long drives autonomous beautifully).
Waymo uses “pre mapped“ aka if the road changed - you’re toasted. Tesla uses AI and real time aka “I am leaning the formulas”. You can throw anything at me and I can figure it out in real time. No comparison.
Phil: When Tesla FSD Beta gets consistent enough for any robo-taxi networks at level 4, be sure and get back to us. Until then, your bias is showing when you claim "no comparison".
Thanks for the video; really like it. At the 2:14 mark of this video the Waymo indicated that it would arrive in 20 minutes at 10:53 am. At the 10:51 mark of this video the Tesla arrived at the destination in 16 minutes. The Waymo indicated that it would take another 4 minutes, which it basically did, as indicated at the 13:23 mark of this video. So the Tesla, was 20% faster getting to the same destination. This is for a small Geo-mapped area of a large city. If you live in the suburbs, Waymo can't pick you up nor drive you home because you are outside of the Geo-mapped area. Another reason Waymo can't pick up a passenger in the suburbs is because it is also limited to 35mph. Most cities have freeways that connect one part of the city to other parts of the city. Thinking of San Antonio TX, where we once caught a cab, after a new years eve party, to take us from downtown to the outskirts of the city, near the airport, where our hotel was. That was an hour trip instead of 20 minutes. Like this video, at the 3:16 mark, where Waymo takes you through the suburbs while Tesla keeps to the main roads, the flat rate cab driver didn't like taking the freeways and threaded her way through the dark suburb streets. Being late at night, we were terrified that this was a setup where the driver was going to be 'forcefully' stopped and we were going to get mugged and robbed. I'll not be taking a 'taxi' that can't go over 35 and has to stick to the shady parts of town.
Clearly shows that Waymo should close down operations now. They have a geo-fenced limited service area robo taxi that cost them at least $150,000 to get on the road that requires them to have support staff nearby and to periodically remap the area, which gets you where you are going more slowly, and will be competing against Tesla robo taxis that cost Tesla $20,000 to make, or less, that can go anywhere with no support staff and no mapping and will be far more numerous and will let you play video games or watch movies on the way at a far lower cost per mile.
A geofenced approach makes some sense because most robo taxi traffic will happen near the world's major cities anyway. This means a lot of the revenue can be captured by this capability on its own. If it can be price competitive, that is...
Both of them seem to be behind the Mobileye autonomous Taxi I've ridden in. Regardless, its great to see American companies pioneer autonomous driving, making the cars once again an American industry.
I could never tell by looking at that waymo that it’s a self driving car! It looks so natural. No extra or bulky equipment all around and on top of the car at all. Very nice waymo, you’re on the right track
Honestly, as a person driving another car, running, or biking I like to know if the car near me is using autonomous driving. There are a lot of little things I can assume about human drivers that I can't about autonomous drivers. For instance: I can make eye contact with other drivers, assume pretty well that they will notice and respect me edging in, I can trust they understand the biking hand signals. I can trust their visibility of me at night is decent so long as I wear reflectors and have a few lights.
Clear difference in behavior: 1) Waymo is much more careful, can drive longer routes with less traffic just to be sure 2) Tesla is more aggressive but can give up and requires constant supervision. Pretty much reflects priorities of respective companies.
Awesome, thanks for the super informative video! Tesla is ahead of the game, but Waymo definitely a close 2nd. Waymo is working on camera-only and generalization to drop the geofencing. I think this could take them only a bit longer than it takes Tesla to drop the “hands-on” requirement. Great to have options.
a) The cost of riding Waymo is much higher than $15 when you consider the extra equipment, backend costs plus the 5-7 min of time ("time is money!") covering the same distance. b) if the Tesla ride was a robotaxi it would have out-bid Waymo. c) again, if the Tesla ride was a robotaxi it could pick-up another passenger where it left-off the last one and go to a non-geo-fenced destination. On the other hand, Waymo, if it lucky, will find another geofenced destination, if not, go back empty, raising the cost of the original fare. It is crystal clear, Waymo is far behind (not just by your measure)! If I were the one to decide, I would stop the Waymo bleeding immediately. There is NO competition!
Doesn't Tesla charge $200 a month for their driver assistance software? But also, less control of environmental variables means more potential liability and accidents even if they are not the systems fault. I don't think Tesla would pay for the insurance required or pay the team of people in their contact centers who are on hold in case of accidents. Of course with the increased to number of repairs needed due to wear and tear from running a taxi service and the build quality issues that plague well used Tesla cars. People already have issues with their current customer service and I can't imagine them actually paying to appropriately build out the required support services for an auto taxi business.
puffns: If Tesla were a robo-taxi it would be STUCK at the green light. (You can't expect paying taxi riders to babysit the cab software). Stop with all the bias.
@@LC-sc3en They do charge $200, for unlimited use. But thats just because they think its worth that much, not because the expenses are that high. Waymo charges $15 for a 20 minute ride, and they still lose billions
Looks like Waymo trying to kidnap the girl...stop 12:22 - no go - "please buckle your seatbelt"... taking you back to the main road. While Tesla had been parked for 4 minutes already. No competition really!
I still cannot believe that the Tesla FSD system is performing equally to the Waymo system but without all those bulky hardware. Edit: After thoroughly watching this video again, I can pretty much say there is no comparison. Waymo is still lacking behind because it is geofenced and has a preprogrammed map, which if the road changes it will no longer work, whereas Tesla can use sensors and navigate itself.
@@Xanthopteryx okay reading thru these comments, you keep repeating the same thing over and over lol But yes telsa got stuck at a green light, it’s true.. but waymo got stuck at a u-turn and needed help .. so why not comment about that too? Do you work for waymo?
@@taylorb2783 Waymo drove all the way without anyone supervising. And with Tesla you have to babysit, ALL THE TIME, because it will do stupid things. And it can not even follow basic speed signs, and it will phantom brake, it will stop in intersections with oncoming cars and so on. No, i do not work for Waymo. And that does not matter anyway because Tesla is no where near self driving.
@@Xanthopteryx I’m not trying to argue their progress, just that both systems aren’t there yet.. pretty sure Waymo is supervised, as you could see it had to call for help, so idk what point you’re trying to make 🤷🏼♂️ Waymo has downsides too, in addition to the wrecks it has sometimes, to the fact it can’t drive on the highway, to the limited area and weather it can drive in.. there both “not even close” in your words.. but it’s still cool to see how far they’ve come :)
@@taylorb2783 Yes, Waymo can radio for help, but no one is driving it. The car will ask like "I want to do this, is that good?" "Or, here are two suggestions i have" and so on. So when it reversed, they probably just said, turn around - the car do the reversing itself (Tesla is incapable of that too). And Yes, Waymo is working exactly as advertised: It is a Level 4 system. It drives by itself. No one ready to take over at any second. Tesla, does not drive by itself. They will not be anything close to self driving with the current cars out there. It will drive into clearly visible things. It will do really dangerous stuff. It will stop in the middle of intersections with cars coming at it and so on. Waymo can drive on the highway, and can drive in the rain, but they must be hundred percent sure that it is safe, and that takes time, a lot of time, and testing, and testing.
As a Tesla owner, I'm actually super impressed by Waymo here. It's pretty daring to offer 100% driverless rides, and it actually worked with zero interventions. All in all, I can clearly see that everyone's car is going to be able to pull out of their garage and pick them up from a night out, and that, friends, is insane.
Sami had a bit of a problem with where the destination was. Looked like it got remote driver assist. FSD was very impressive navigating lane changes and traffic. Knowing that FSD can do this anywhere and not just in geofenced area makes it the clear winner in my book
That was a very well done comparison! Nice job. Everyone is so biased these days it is nice to see such objectivity. One small point. I could not see the clocks but it appears to me that the Tesla finished 33% to 50% ish ?? faster than the waymo? Hard to tell with the video editing. It would be interesting if you had the Tesla speed % over set as you mentioned in the beginning... Great work.
Always love these comparison videos. Once Tesla offers its own robotaxi, there will be a price war so your same trip should be less than $14.40 just like how Tesla is cutting the price of its cars now to fend off competitors for EV ownership. It will be a tough price war and unless Uber and Apple get in now, they won't stand a chance. With v11, our X is getting better and better. It's only a matter of time. You bring up a good point that these comparisons test not only the driving tech but also the navigation which is crucial for any delivery service to succeed. The one thing I do wish Tesla would do is offer users the ability to custom select routes rather than rely on waypoints. I'm a firm believe not all roads are engineered equally, and frankly, there are some streets I avoid simply becuase of the higher risk for crashes. I think it is the one weakness that Tesla has. It assumes if the road is paved, its equal in safety with all other nearby roads. Also, Callas is a good fsd driver :)
There will not be a price war in robo taxis, Tesla will announce their price and no one else will be remotely able to match it. This competition here, Waymo paid probably $200,000 for the vehicle ($69,000 i-Pace) and all the LIDARS and cameras and computer hardware plus installing it. The Model 3 cost Tesla about $25,000 to manufacture. Then Waymo has some facility in the area with staff monitoring the robo taxis and able to physically respond to help one in person if needed, and the expense of periodically remapping the area to maintain accuracy. Per mile, Tesla will cost maybe 25% of what Waymo costs, and Tesla will be able to go everywhere and will have far far more robo taxis available. This war is already over, and Waymo should stop burning their money.
It's a long way from Tesla taxis. Because that would require Tesla to have set up an emergency notification to operations teams in case of accidents or system failure and to take all the liability costs of at fault accidents. I don't think they will do that. They are charging over 15K a year to their customers just to have the driver assistance it currently has. I don't think they would front the additional costs for everything needed to make a taxi system to charge less than $15 a ride.
@@LC-sc3en 6 mths ago, 99.9% of the population had never used chatgpt. Now, Alphabet and Meta are rushing to play catch up. Disruptive technology always happens a lot faster than most people expect or anticipate. We've had fsd beta for a couple of years now and v11 is pretty decent as an asad. I'm disengaging less and less to the point of zero per trip. If Tesla wanted to, it could replicate the same service that Waymo is currently offering in Arizona (no highway, geofenced, etc) and it would probably do a pretty good job of it. Yes, it would have to create the human support and technical infrastructure but it wouldn't as difficult as say, ramping up M3 production. In fact, when Tesla does start rolling out AVs, they will likely follow Waymo's example to a certain extent to build experience. As for price competition, once a good or service is commodized, there's always a price war. EVs are about 15% of the Cali market (one of the largest vehicle markets in the world) and more OEMs are bringing out their own models so EVs are becoming more common. It's a prudent move to cut Tesla prices at this point.
denero66: And Tesla isn't consistent enough to have robo-taxi networks (much less large ones) ANYWHERE yet. And I'm not convinced it will without good HD context sensitive maps that get updated, perhaps via crowdsourcing like Mobileye's brilliant solution for that. Progress is incremental. No one knows what "winning" solutions will look like yet.
It happens somewhat regullarly for me on the East Coast. Not sure what triggers it. I don't think it is only the length of the red as Happy Camper, INc. postulates.
Av8: It's not that unusual either, and though it's slowly getting better, it's an example of why Tesla FSD Beta is WAY too inconsistent to be a practical robo-taxi network at all, much less a huge one.
Network will beat mapping any day! I can only imagine how powerful a car would be with both the hardware of Waymo and the software excellence of Tesla FSDB!
I'd be curious to see how the two systems do in snow, rain, fog, etc. I'm skeptical that Tesla would do well with only cameras but perhaps they have a way of keeping them clear/clean. Very cool to see though. Great idea for comparing the two systems at the same time and on the same route! Thanks for the video.
While it's probably no big deal in reality, it does look a little scary how much the Waymo car is shaking its steering wheel all the time. Fascinating test results!
So let me summarize this comparison: Waymo - many more expensive sensors, geo-fenced (can only drive in a few defined areas), avoids highways and main roads...arrives later than the Tesla Tesla - cameras only, uses highways, can drive anywhere in the US (even unmapped dirt roads)...still beat the Waymo Pretty clear who has the lead in AV tech.
To be fair you need a comparison outside a Waymo mapped area. That shows a Tesla driving like in the video and Waymo just sitting there. Tesla wins every time.
Tesla is already so far ahead of everyone else its scary. To think that Waymo still ist relying on a geomapped/premapped/interpreted area data that was manually edited by humans and still not able to drive on highways is really showing that even companies that had much more experience in AI than Tesla doesnt mean a thing if you dont have the necessary strategy and data to go forward. Waymo might look more autonomous at first sight and be even safer at the moment, but you have to have the balls to go public and let real life szenarios take over and teach the AI even if this might be a big disadvantage at first as you can see from the headlines Musk and Tesla gets for this decision. But Simulations and driving on a human made trail just wont cut it. There are so many edgecases developers even cant imagine if they tried. This wont work with a small team of drivers in some local areas and telling the AI how to interprete every mm of a city and street. In addition the expensive sensors which render the cars not be able to be used by everyone. Sooner or later you have to fly the plane and not simulate it. If Waymo wont make this step sooner or later they are out of the game and only good for e.g. bus routes, fair transportation and so on, but they wont be the one used around the world. The best AI training data will win the game, not programming or simulating.
Tesla chose the much harder route and already surpasses Waymo in performance. I am quite confident that Tesla can iron out the kinks and have increasingly less interventions. I suppose DOJO is already in use and that combined with the single stack architecture will lead to faster progress than before. Very exiting to follow. Once Tesla can prove that FSD is saver than an average human driver, legislation will have a hard time to be against it, as it save lives and then would be irresponsible to delay it.
@@nilsfrederking62 There is no reason to prove FSD is safer than a human driver beyond what has already been done, it is safer than a human, but Elon wants it to be 100 times safer.
@@tedmoss Not in the video I saw ! Without human supervision it would have caused a serious accident ! It's not ready by far yet and somehow watching very different experiences of the same version, I sometimes do have the feeling Tesla also geofences !
@@Borg-vb2lk Tesla has no interest in geofencing, it only would slow down progress, as they need to collect he cases where it fails. But they target not all markets at once, US is the first and afterwards Europe and others.
Of course self driving vehicles are the future, but it takes longer than I would have expected, and the safety and responsibilty issues are still a hot topic in many countries. I also think that, if I look at those videos of a "city" that to me more looks like a quite uniform area of uniform flat houses and wide steets, self driving here in Europe with its very inhomogenous, narrow streets and dense cities will pose some additonal challenges as well, as will the whole issue of aquiring route data for learining (many here will not be ready to share their route data for privacy reasons).
What a nice video I never thought I wanted to see! thanks to RU-vid AI. Nice human assisted reverse. 😉 I wish you concluded with time difference and maybe one day you could force tesla to not use highway.
Thanks for the video but really, no comparison. Waymo with all those restrictions will cost way more to get to full automation. Tesla's approach (Software + vision) is the way to go.
GK: If only the world ran on intuition / assumptions. It's not flat, and it doesn't run that way. We won't know "the way (or ways) to go" until solution(s) actually get there that are robust enough to consistently work well without significant human intervention / backup. Given the definition of AV L4, and that L5 is clearly decades away (or never attainable) -- it's a long wait.
