The Tesla ended up in rush hour traffic. But I’ve done carefully calculations and after many years of driving in sf, I’ve determined that anywhere you want to go in San Francisco takes exactly 20 minutes from anywhere in San Francisco.
Waymo seems to have mastered the art of the 4 way stop and never going above 25mph. That works well in the route taken to Bernal Heights though it sure takes a long time! Cool to see the comparison to how beta handles the same drive and how different the route is!
alexi 183: It especially takes a long time for Waymo in comparison (in various direct comparison drives) in longer drives when Tesla uses the highways and Waymo doesn't. For those, with similar traffic for each, Tesla has been CRUSHING Waymo on time to make the drive, by using the highways when that makes sense while Waymo has to fight the city traffic and lights all the way. Obviously high consistency is CRUCIAL for robo-taxis to work, for any provider, and that's where Tesla (overall) still has plenty of work to do between different areas, types of traffic, types of roads, etc.
At 10:44 Waymo goes badly out of turn at a 4 way stop. It would not have passed its driver test doing that. The red truck was clearly first, and although there was a pedestrian, the pedestrian had almost cleared the truck. Indeed the truck started to go, but had to yield to the Waymo. It is not a terrible mistake, but hardly good driving.
OMG that traffic is insane!! Those hills look like 80 degrees up then down. I have never seen anything like that terrain. Definitely doesn't look like that in MA. I am rarely on a 4 lane highway. You have passed more Tesla's in this video than I see after 500 miles/5 days of driving. Unless I drive to Dedham where the service center is (or go to a rich town.) I love seeing the comparison with Waymo.
I have 77K miles on my 2016 model S and find the old autopilot good and useful. I just drove a friend's Model 3 with the FSD option and the latest FSD beta activated. It worked but nowhere near autonomous. On a short 5 mile light traffic trip it made it to the destination of Home Depot; almost. It could not back out of the driveway and when it got into the parking lot it went on a collision course to parked trailers and I had to quickly take over. It turned left at a simple intersection when the light turned green it cut sharply and almost clipped the car sitting at the light for the other direction. On the return trip it started to pull out from a stop sign with a car coming and I had to stop it from doing that. It ignores potholes and it tried to run over a kid playing in a side street, but he got out of the way. I am sure it would run over dogs and cats. The $15K my friend spent for the FSD option is a total rip off profit center for Tesla FSD will probably not be safe until after his car is worn out. The should be paying drivers to test the beta not the other way around.
yet almost all Teslas on the road today will never be able to drive without a driver or allow the driver to be distracted because regulators require redundancy 😅
@@Daniel-jv1zs: Never is a hell of a long time given how much Tesla FSD has improved in about 3 years. Don't be ridiculous. The small area tightly geofenced services like Waymo and Cruise already don't need human drivers, so no redundancy there.
A competitor to Tesla might just be more successful with additional sensors, such as radar and lidar. There is a limit on the useful information Tesla can get from cameras alone. Not all roads are in sunny California.
I'm on FSD Beta 11.4.2. It has gotten worse at taking the exits on the GPS route. It either skips them or it takes some that aren't even on the route and make zero sense. Yesterday it pulled into oncoming traffic. It seemed to think it was a driving lane for me.
Yeah me too I noticed it takes wrong turns way more now than it did before I'm sure they'll patch it in the next update but it's super annoying for now, FSD did get better in some areas to be fair though on 11.4.2.
Here's where you're going to see Waymo win, and it could be a game-changer happening soon... there are rumors here in Arizona (from a friend who just went to a focus group, he does it for fun) that Waymo will offering unlimited monthly rides on a subscription model, and they'll be available to provide them for as low as $100/mo. The discussion in this focus group was very data driven and ended up getting pretty 'mathy'... the group asked how much the vehicles cost to acquire - what that monthly debt service would be, and the cost to operate, what average ride lengths/distances were, etc., etc. Another factor was revenue generation from riders who would likely be using their phones instead of not using their phones while driving their own vehicle. It was determined that at $100/mo, Waymo would maintain healthy profitability. They wouldn't even have to go this low, as many people thought $250/mo would get them to ditch owning a vehicle permanently. They were also shown an image of a concept vehicle that was more of an ultra-luxury vehicle... a total concept with a bathroom with Japanese toilet, a vending machine with hot/cold drinks and snacks, and two full body massage chairs and a workstation with monitor, mouse, keyboard that you could plug your phone/laptop in, and the ability to schedule your rides... and asked "if you were a person of means such the typical Uber Black rider or someone who has an exclusive or shared driver or someone who pays over $2000/mo for their vehicle, what would you pay for a monthly service like this where transportation becomes and ultra-luxury experience? Would you pay $500/mo?" and everyone was really excited about this. Thinking about this, I'd choose the more expensive option myself if I needed to go places often... but these days I don't. I'm on my 3rd Model X, and I only drive it about 2 days a week... I love it but almost don't need a car for anything but road trips. Also thinking about this, the additional cost to create an ultra-luxury 'transportation experience' is negligible as the main expense is with the vehicle - adding $20K in gimmicks totally makes sense. To close out this long ramble, it seems Waymo will be able to provide tiered transportation with a monthly subscription model, and since there is no labor to pay for and since electric vehicles have low maintenance costs and since the energy costs are very low, I would argue that we're headed into a world with near-zero levels of vehicle ownership, which could also be really bad news for vehicle manufacturers. This is already what's happened for so many other sectors in the economy... it's all about sharing things and getting more while spending less, and not having the hassle of maintaining, repairing, or suffering from depreciation. Would love to have you explore this... if there are just 25-50 subscribers per month per vehicle, it's profitable no matter how you look at it. I'm excited about the future, for sure... I see a time, soon, when nobody will need to own or take care of a vehicle.
I'll add to this that my friend lives in the Phoenix metro area.... I have a home there but also have a home in Flagstaff Arizona where I spend 95% of my time. Flagstaff is a smaller mountain town of around 65,000 people that operates a very small bus system that still costs an outrageous amount of money to offer very limited service with infrequent buses that only cover a small area... the city council is currently discussing scrapping the bus system as soon as they can attract Waymo or a similar service here... but offer all rides at around $2 and just subsidize the rest. The rides would be available 24/7, offer door-to-door transportation, and be far less than the farebox subsidies they pay for their current bus system. I remember once that their farebox subsidies were $13 per ride, but I'm not sure what they are now - and this doesn't include the cost of building bus stops, installing bus pullouts on roads (exorbitantly expensive), using expensive downtown land for a transit center, etc., etc. This also seems to be a pretty interesting application for Waymo style services... towns that are too small for a robust transportation solution but still need some sort of transportation solution, where the cities can save money while also providing residents with 24/7 door-to-door transportation for less than the cost of a bus ride. Seems like a win-win to me. Again, I'm excited for this and would consider ditching my Model X if my town offered this service.
Super impressive demonstration. Tesla FSD beta performed well except the long delay behind the parked truck. That can be fixed. Thanks for doing this. In my view Tesla FSD team for the win!
Wow! Thanks for filming this. Self driving cars have come a long way and both cars were impressive. The juggernaut in my opinion will be regulatory approval for full self driving cars. Too many lawmakers are old school in thinking and in age IMHO. I think China will be first to allow it though.
Omar, first of all, thanks for all you do. This was an excellent presentation of the current state-of-the-art, even though the rush hour traffic for Tesla FSD was much more challenging than the earlier Waymo drive. Second, I think map metadata is key to Tesla FSD's ability, both today and in the future. You even discussed this with me (FSDnow) in a Twitter Space about why FSD is better for you than many of us elsewhere. In this video (48:30), you mention Waymo probably choose its route (and lanes) for easier traveling. FSD has not been optimized in this way, but could be in the future. From this video, there is right turn (59:56) where you mention that in past versions, it would make the right turn into the service area (side street). To me, this is a sign that Tesla has improved the map metadata and the use of that data rather than an improvement in the FSD computer software. Which is contrary to your commentary, "… a robot using just cameras." From your commentary, for this drive, I would expect Waymo to have a near 100% success rate on this drive and FSD to be less than 50%. This single example was far from an equal comparison (traffic/route/etc) and again, the system's are currently optimized differently. Perhaps Tesla could perform approximately similar to Waymo in 1 in 10 drives. But, flip this around, Waymo cannot even do 1 in a million of Tesla's drives. At 1:03:28, you go into this discussion of "solving self driving." This is a simplistic view and does a disservice towards progress and understanding of achieving self driving cars. Tesla consumer cars are SAE L2 now and should have some L3 in the near future. Tesla L4 cars will initially be company owned. To allow L4, Tesla will have the same challenges (logistics, support, remote driver, service, etc) that Waymo and Cruise have. Consumer owned L4 cars are a distant future. You come back to this "solving fsd" at 1:09:05 and 1:11:28; this is meaningless and unhelpful. Tesla FSD has had successful no intervention, no disengagement drives years ago. I agree it is absolutely amazing what Tesla has done, but we are no where near to L4 consumer owned Tesla vehicles. Key challenges for Tesla FSD in this video 43:36 "Piece of sheet. Go around, go around, go around" 47:39 Arm gesture intervention to get around stopped truck. 53:07 "We might miss the turn. We will see." 54:01 "Ah, come on, what are you doing?" 56:22 "What a [traffic] disaster"
I Disagree. Tesla is L3 and does not intend to pass through L4 (as that allows geofencing as a crutch). They intend to transition from L3 (human override still required but aware of environment) directly to L5. From what I have seen, in this video it is not in some distant future. There is no need for Tesla to have the infrastructure that L4 requires. The truth is that Tesla was very nearly as good as Waymo (with no aversion to the highway) but it can do this in Any city or highway. Not just the few slect cities that have been mapped by Waymo. It is a Production car already sold in the millions and does not require very expensive sensors that will prove a big obstacle to full adoption. For LiDAR to receive mass adoption, Solid State LiDAR will be required (sensors in the hundreds of dollars vs $7,500-$75,000 dollars for Waymo current and near-term sensors. I haven't been able to find a timeline for its introduction.... so. Not holding my breath just yet. By the time they approach production, maybe Tesla will include them in their neuralnet.
This is exactly what I'm talking about. The average joe, @@avgjoe5969, is coming to believe Tesla FSD will "walk on water" in the not too distant future. Our FSD enabled cars will give Uber rides, find non-existent addresses, disallow minors and babies without baby seats, slap riders hands when they grab for the steering wheel, return left behind cell phones, clean up puke out of the backseat, fix flat tires, deal with officers informing them of a closed road, fight false damage claims in accidents, go to court when a rider sues for "emotional damage," ...
You can't know anything to justify your "distant future" statement, or anything about rate of change. Progress in this is entirely impossible to predict. Forecasts of the inability of AI to do x in time y are all baseless. We just don't know. We know AIs are proving extremely capable in unpredictable ways. We cannot know what is 3 months away, 1 year away, or 5 years away. Retrospectively we can say that Waymo engineer's past statements about the impossibility of achieving self-driving on the Tesla path were mistaken. What is looking to be impossible to scale is the Waymo path, because of its capital intensity.
@@avgjoe5969 I totally agree with Avg Joe and Omar. Eric isnt getting it. Tesla may not have the diploma in hand, but its already L3. What we are seeing in the recent progression of Tesla FSD beta, is that realistically, Erics list of "challenges" aren't really challenges at all. The basic FOUNDATION of Tesla's self-driving (with only vision) is proving out in 11.x. The issues arent truely "challenges" anymore - its just training time. As an analogy, I've never summed the first 100 odd numbers before; but I as I have learned the FOUNDATION of arithmetic; its just setting aside time to do it. Likewise, I'm not seeing anything from Omar's drive that is overwhelming processing power of the Tesla, requiring more instrumentation, or in anyway pushing the limits of the Tesla software.
This is amazing and a bit baffling. I have a 2022 Model S with FSD beta and I've never seen performance anywhere near this level. Maybe it has better handling built in for SF somehow. I've gotten to the point where I don't even try FSD any more I've had so many close calls and bad experiences with it.
I'm at 27:01, and it looks like there is not much traffic in San Francisco, and what traffic there is seems to lack the frenetic tension of traffic in New York or other cities along the East Coast. I'l like to see how Waymo does in NYC or Washington, DC for example.
Would like to see how they handle swedish roundabouts. Would it just wait there forever (like many swedes do causing queues) or would it try to time it and scoot out in between like one is supposed to do.
Great video. Most important thing this demonstrates is how much safer an autonomous vehicle is compared to a human. The human is worried about speed and getting to the green light before it turns red. The car cares about human life. I can't wait until its illegal for a human to operate a motor vehicle.
another few years and they will have mastered self driving cars. Pretty cool, sit back relax and enjoy the ride, I'd love it on my 4am commute to work every day. Take a nap
Thanks for the unbiased delivery…… of a technology competence comparison. ( albeit with ‘their’ baby versus ‘your’ baby) So…..FSD Cars(Taxis) ARE a REAL thing Right NOW. Waymo works very well ! At a functional product/service level Tesla is entering an occupied market, but late. At a Commercial level, cost per Car & cost per Ride, there may be NO evident competition to Tesla. ( Waymo cant scale with an Ipace) I’m still trying to figure out how to think about this as an All IN obsessive Tesla Investor who DOES have Robotaxi in his spreadsheet. Initially my thoughts are ……somewhat confused.😮 But educated - Thanks Omar.
I'm shocked how much FSD has improved in the city. I've been trying out the Cruise, but I'm still on the wait list for Waymo. I guess in July it will open up for everyone, and now city officials are panicking and trying to stop the opening up of the system. I don't know if you remember vision 2000 to try to stop so many pedestrians fatalities, but apparently we no longer care about Uber and taxi drivers running people over. On Monday I had an Uber to SFO and so many close calls -a lot of people need to stop driving.
You can’t look at one high profile case of a semi-autonomous vehicle hitting something. You have to compare it with the rate for human drivers. Plus take into account the improvements in all self-driving vehicles in the past four years since the Uber accident.
Hi, it sounds like was not your first time trying this. I was wondering did you feel safe and was the car using it's blinkers when turning left and right?
At 10:46 you complimented the Waymo for slowing down for the red truck, but actually it made a mistake. The red truck arrived at the four way stop first so it had the right of way. The car should have waited until the red truck finished crossing the intersection. On the whole, though, both Waymo and Tesla are very impressive. It seems that Waymo is better at selecting a route it knows it can negotiate easily. Tesla took the route that might have been optimal for a human driver but required the car to negotiate several intersections it didn't handle very well.
How would the cars react, if a small dog or a small child (e.g. fallen from the bike) would lay on the ground if a police officer would wave with his hands to say "don't drive into this way, you have to divert" if you are in a situation where you have to reverse?
It seems like they have modified the steering wheel because of the shaking on rough roads, something you would not want the driver to feel. Perhaps for more precise steering.
the fact that the Tesla can't effectively merge over for a stopped truck in its lane is a major short coming. In downtown Atlanta where I live, you literally could be stuck there for hours until the truck itself moved. the traffic is unrelenting during certain points of the day. another point is you'd be blocking all the parked cars to the right if they wanted to leave. its cool to watch the screen as the computer tries to figure out solutions.
Indeed, that is what I think too. You really can't say it was without intervention, as communicating directly with others is an integral part of driving, of course.
At around 43:00 FSD encounters a lot more traffic than Waymo. The stopped truck and FSD's inability to get around the truck shows that FSD is not yet ready for autonomous service in typical urban traffic.
@@benjaminquinones1890 In real life, "a little bit" is way too long. Furthermore, FSD missed the left turn to get onto the highway in its first attempt. Try "jumping the line" as FSD tried to do (a failed attempt, by the way) like that in NYC and experience the reaction you get from other drivers. I'm a big fan of FSD. I have it myself on my model 3, but I know its severe limitations in dense, urban traffic so a "fan boy" I'm not.
I don't think anyone, other than Elon, is claiming it is ready for autonomous service. But all these problems are fixable. Unfortunately it seems Waymo is only "ready" in very limited geo fenced areas with hard programming designed to avoid congestion. The Tesla trying go left out of the straight lane was hilarious (and illegal). But it isn't like that hasn't happened to me. They just need to dial up the aggression on slow speed / no speed lane changes.
@@andrewf67 Well, Whole Mars came very close when he concluded, at the end of this video, that FSD was ready to scale...by which I think he meant it is ready for the general public. FSD is not ready for dense traffic in a dense urban environment in real life...drivers of FSD are in a hurry, as is everybody else in traffic. I don't think the drivers waiting impatiently in line in that left-turn lane would have been amused if FSD had rolled serenely to the head of the queue and then turned left out of the straight-ahead lane. The reality is that FSD has to be better than Waymo in the city. The argument that Waymo cannot "scale" is not very convincing...it just has to be as good or better than a human-driven taxi in, say, Los Angeles, Chicago, Washington DC, Philadelphia, New York, and Boston. The video shows that it already works in San Francisco. I presume that Waymo is aiming toward an autonomous taxi fleet in parts of the world where most of the traffic is -- big cities; I don't think they have the same ambitions as FSD.
Impressive for both, but the real question is how they will reacts to a more complex scenarios. In my city, for example, it's a total hell to drive with all the electric scooters that obey no laws.
I’ve always liked those chill Disney rides bc u can just relax n enjoy the experience around u while the vehicle is driving you. W/ fsd I sorta get that feeling with the rest of the WORLD.
Waymo seemed to predict/react more swiftly without the abundance of caution of Tesla FSD. Also Waymo seems to not be stopped at intersections for quite as long (again more cautious Tesla FSD). I would like to see how both negotiate more complicated construction zones with lots of cones giving direction to different lanes etc.
$chris: Waymo has a big advantage in its own small geofenced area with the HD maps. The problem re scaling that for large areas AND keeping it affordable is a big one though. As Tesla improves, erring on the cautious side and not doing things like hitting pedestrians (where it would be at fault) is obviously important.
The car is in motion. those tiny adjustments would be negligible it’s not like the car is in a loaded corner where the weight is shifting or grip is lost. If those minor adjustments add significant tire wear than an alignment issue would be more likely.
When driving straight, the Tesla wheel is not shaky. The Waymo wheel is very "nervous", and would make me as the passenger a little nervous. And why doesn't the Waymo vehicle keep its mirrors folded?
Looks like that in the Jag, the passenger is in the back, filming. In the Tesla, the passenger is behind the wheel. That would be way to weird. I'd want to be away from the controls.
The big difference: in the Tesla you have to stare out of the window for half an hour while in the Waymo you could check your e-mails, surf the web or read a book. Waymo is what Tesla always wanted to be - a REAL robotaxi.
Always look on the road and dont use your phone, if it notice you don't pay attention the nag will increase. However he might be also and early tester so he got maybe another profile in the background.
I hope you have shared this video with Tesla. The way it sits there and waits when the lane is blocked even when other cars are slowing down to let it in and other cars are passing it is ridiculous.
jg-bd3hr: It's an edge case where Tesla clearly needs work. There are a LOT of them. It's still FAR better to have this sort of behavior than risk accidents, which Tesla FSD used to do a LOT more than it does now. It's still cumulatively getting MUCH better, but it's incremental. The fact that Tesla is as good as it is without geofencing in the US still has it WAY ahead of Waymo and Cruise, given their geofencing requirements.
on 58:31, Tesla chose left line when turn right and it should have space for right line. more dramatically, it switch to right lines in seconds after finish the turning. so what is the logic here? but impressive for both software anyway
well,, I guess it think the right line is faster, and it is his original plan, but when you turn right, the right line should always your premier choice. so..... Oh, try to stare the monitor all the time, not like at the windows, you will understand how the autopilot works, it like a video game, it is all about how to similarize the reality a better way in real time
they do have operators in the background who can take over at any time. it's possible they might be able to oversee multiple cars, which would make it an advantage, but Waymos still have many problems as well.
@@LunnarisLP The real question is, how many times were remote operators actually driving the Waymo??? I doubt they display that for the world to see... imagine having millions of cars needing remote assist? Oops.
Disagree on the solving phase comment near the end. First, Tesla would need to get VERY consistent, re completing drives with NO human intervention in ALL sorts of traffic and locations for LARGE areas, to be the practical robo-taxi service they're aiming for. It doesn't have to be perfect, but it needs to be solid level 4 where it only phones home for help occasionally. Today it's hit or miss. Second, Waymo has a viable taxi service, but on an EXTREMELY limited scale / geographic area. If Waymo is going to be cost competitive on a broad scale, they need to scale WAY up and in MANY cities. Also, if it can't do highways, it can't compete with Tesla re time of Journey, re various comparison drives, already. The race is interesting for several firms. But none of them are yet close, re real world robo-taxis on a large consistent network with low rates and profitable.
This is the crux of it right here; both Waymo and Tesla FSD programs are doing an excellent job now, Tesla rapidly improving and Waymo will finally take a person anywhere in SF they want to go...but each Waymo taxi costs them a huge amount of money compares to a Model 3, they can only be found in three cities, and we do not know if they can ever be used in cities with rain and snow because LIDAR does not do well in snow especially. What is the elevator pitch for Waymo now; we will charge more, make you wait longer, get you there slower, and be available in a lot fewer cities than Tesla? Respect to the Waymo team for their good effort, but it seems their business plan was to do all this and for Tesla to ALSO fail completely, because Waymo legit believed Tesla was trying the impossible. Now that Tesla is not failing...I do not see a case for Waymo, or Cruise of course.
@@tv-ld3wv they claim, as does MobilEye, to have a way of quickly and at low cost generating the maps they need. It seems possible. But still, with their cost to make the taxis and the electricity to run the computer and LIDAR and then having support staff ready to respond 24/7, they end up at Uber prices with the benefit of no driver, if that is a benefit for you. Tesla will cost a lot less and there will be far more Tesla robo taxis and they will be in every city and able to go on highways and between cities and so on.
