Waymo and similar companies all use HD Mapping. Ignoring the specifics, it can be thought of as a centimeter-level perfect reconstruction of the environment, including additional metadata such as slopes, exact lane positions, road markings, barriers, traffic signs, and much more.
HD Mapping is great when it's accurate and available. But it requires a ton of data and constant updating, or the car will get "lost," and realistically will never be implemented in general, at best in certain cities.
Reliance on HD Mapping gets you to "robotaxis" quicker and easier, but it doesn't and likely cannot scale.
It remains to be seen if Tesla can generalize FSD enough to reach the same level as HD Mapping everywhere. Still, they have shown that the current limiting factor is not what the car sees or knows but what it does with that information. It is unclear how or why HD mapping would help them at that point.
If you can make the unit economics work for a large quantity of individual cars, mapping is a small fixed cost.
I agree that it's not economical to map every city and road in the US, since you need to generate revenue from every mapped road and city. So you can think of HD maps as amounting to building roads. They will be built in lucrative places. Cruise and Waymo won't make money from putting taxis in nowhere Arkansas, so they don't need to map it.
> the current limiting factor is not what the car sees or knows but what it does with that information. It is unclear how or why HD mapping would help them at that point.
That's simply untrue. All the hard stuff continues to be reliability and sensor gated. Cruise and Waymo have amazing sensors and even they struggle with sensor range, sensor reliability, model performance on tail cases, etc. For example, at night these cars typically do not have IR or Thermal sensing. They are relying on the limited dynamic range of their cameras + active illumination + hoping laser gets enough points / your object is reflective enough. Laser perception also hits limits when lasers shine on small objects (think: skinny railroad arm). Cars also have limits with regard to interpreting written signs, which is a big part of driving.
Occlusions are still public enemy #1. Waymo killed a dog. Cruise crashed into a fire truck coming out of a blind intersection even though their sensors saw the truck within 100ms.
LiDAR and HD mapping together are supremely useful, even if you don't drive with it, for enabling you to simulate accurately. You cannot simulate reliably while guessing at distances and locations. HD maps let you use visual odometry to localize, and distance measurements grounded in physics backstop the realism of your simulation at least in terms of the world's shape.
Tesla lacks the ability to resim counterfactuals with confidence since they don't have HD ground truth. There are believers at the company that maybe you could make "good enough" ground truth from imagery alone but that in and of itself is a huge risk, and it's what skipping steps looks like. Most in the industry agree that barring a major change in strategy they just have no way to regression test their software to the level of reliability required for L4 / no human supervision.
HD mapping, on the other hand, likely needs to be updated frequently and immediately when any construction occurs.
It seems pretty clear that what they are doing today is nowhere near what they need to do. And again, I don't think that is possible.
I dunno if said data would be as high quality as dedicated HD mapping cars, but it's probably at least decent, given the variety of cameras and lidars every Waymo car has.
What does this mean? Electric cars are already required to emit a sound as they drive.
I guess if it has to brake hard for something, honking might be a good idea, but I wouldn't want cars to constantly be beeping at everything in their vicinity if there's no imminent crash.
> I’m not sure why we are trying to teach each car to drive when we could be teaching all the cars and the road to drive.
I'm not sure what you mean. Presumably Waymo's software is the same across its fleet. They're not training one car's model at a time.
Well, if your car brakes hard to avoid a dog, your car should warn me. I’m not sure how to make this concept simpler so I can only repeat it.
> Electric cars are already required to emit a sound as they drive.
I know.
> I guess if it has to brake hard for something, honking might be a good idea, but I wouldn't want cars to constantly be beeping at everything in their vicinity if there's no imminent crash.
If you think in a discussion about robot cars that drive themselves being conducted on a hacker website, I’m suggesting that cars communicate their sensor data to each other by honking their horns, in really not sure what to tell you other than yes, this would be profoundly dim witted.
> I'm not sure what you mean.
I believe it.
> Presumably Waymo's software is the same across its fleet. They're not training one car's model at a time.
I believe it. I also believe you’re deeply missing the point, perhaps intentionally.
1. Robotaxi companies might simply stand to make enough money to cover the cost of routine HD mapping. Anywhere the revenue of putting taxi services in a new city outweighs the cost of implementing the necessary updates sufficiently, won't companies do it? We could think of these companies as having similar economics to Uber, but replacing the cost of paying drivers with the cost of routine HD mapping updates.
2. Smaller towns have less frequent construction, so the update costs might be lower as you target less dense areas.
3. I could see a single company that specializes in providing routinely updated maps to a variety of fleet-operating companies. This could potentially be a utility or somehow subsidized by the government. It would also be possible for government to coordinate construction with HD mapping updates. After all, by lowering the rate of accidents and decreasing square footage devoted to cars, governments have a vested interest in seeing robotaxis replace human-owned and driven cars.
IR and thermal sensing are unnecessary if the bar is human level and neither is the lidar. The point is overused, but humans rely on two eyes in the driver seat. I don't see any evidence to suggest the modern model that Tesla has developed for their vision system is their limiting factor in the slightest to reach L4/L5.
Dogs jump into the road in front of cars all the time and get killed, and kids get endangered at school bus crossings. That's a reality of life that robotaxis do not need to solve.
I get annoyed with statements like this because technology changes and advances so quickly, and Tesla has made substantial technical leaps in this field of machine learning. They have the state-of-the-art vision -> voxels/depth models and are only improving.
Waymo, who use additional sensors like lidar, have a driverless taxi service which needs no safety drivers.
The evidence kinda speaks for itself, IMHO.
[1] https://twitter.com/elonmusk/status/1740641473849352450?s=20
> Tesla, who use cameras only, have not demonstrated full self driving
There are entire youtube channels with hours of continuous video showing Teslas driving around SF, but also other parts of California, with no human intervention.
Waymo is geofenced, but within its geofence it requires zero human intervention. Tesla on the other hand is famous for mistaking the moon for a traffic light. Saying "Tesla has so many miles on YouTube" is hilarious because first of all there are channels with lots of Cruise & Waymo footage too, and more importantly it's not the # of miles that matters, but the # of non-trivial scenarios you can handle.
Btw Waymo is nowhere near achieving unit economics. Their cars cost like 5 times what Teslas cost, and the sensors require a lot of upkeep and maintenance.
It's probably not accurate to say that the remote operator is "steering" though. It's more like pathfinding.
<photo of construction zone> Can I drive through here? [Yes] [No]
When this is happening, the car is stopped and lets the passenger know that it’s reaching out for remote help to figure out what to do. For me this has happened two times across my 125 Waymo rides (571 miles) so far, and was resolved in under 20 seconds. Though I must say, 20 seconds feels like ages when you’re in the car and blocking traffic!
Eyes which are orders of magnitude capable than the best cameras, and Teslas come with mediocre cameras, not the best. Eyes which are connected to a brain, and ML is a looooong ways from rivaling that.
> That's a reality of life that robotaxis do not need to solve.
Robotaxis do not need to account for things jumping out unexpectedly in front of them?
My brain tries to hides all this from me, and makes me think that I see the world in glorious 3D technicolor all the time, but that's a lie as revealed by the many amusing optical illusions that have been discovered over the years.
Meanwhile, today I used ML that knows more than me, can think and type faster than me, which is a much better artist than me and which can read and react far faster than me to visual stimuli. Oh, it can also easily look in every direction simultaneously without pausing or ever getting distracted or bored.
Somehow it doesn't feel like I have a big advantage over computers when it comes to driving.
The maximum dynamic range of the eye is ~130dB. It's very difficult to push an imaging system to work well at the dark end of what the eye will do with any decent frame rate.
It's not as different as it used to be, but even so: the Mk. I eyeball does pretty damn well compared to quite fancy cameras.
Depth sensing is again, estimated or using time of flight sensors which is pretty much short-range lidar. HDR is used already in AV perception, but still loses to your eyeballs in dynamic range and processing time.
Eyeballs have high dynamic range but with high mode switching times. Walk from a bright area to a dark area and it'll take seconds for your eyes to adjust. Cameras are so cheap you can just have a regular day camera and a dedicated night vision camera together, switching between feeds can be done in milliseconds.
Robots aren't humans. You need accurate depth perception to maneuver a robot precisely, and you need ground truth depth measurements to train learned depth perceivers as well as to understand their overall performance. Humans learn it by combining their other senses and integrating over very long time using very powerful compute hardware (brain). To date, robots learn it best when you just get the raw supervision signal directly using LiDAR.
> Walk from a bright area to a dark area and it'll take seconds for your eyes to adjust
You do realize cameras have the same issue, and that HDR isn't free / is very computationally intensive?
Your brain is _really really_ good at surmounting challenges including many that you did not mention. We don't know how to get close to this in terms of reliability when using cameras and ML alone. Cameras and ML alone can go very far, but every roboticist understands the problem of compounding errors and catastrophic failure. Every ML person understands how slow our learning loops are.
Consider that ML models used in the field have to get by with a fixed amount of power and ram. If you want to process time context of say 5 seconds, and with temporal context 10Hz and with resolution 1080p, how much data bandwidth are you looking at? Comparing what you see with your eyes with a series of 1080p photos, which is better? Up it to 4k: how long does it take to even run object detection and tracking with a limited temporal context?
Your brain is working with more temporal context, more world context, and has a much more robust active learning loop than the artificial systems we're composing today. It's really impressive what we can achieve, but to those who've worked on the problem it feels laughable to say you can solve it with just cameras and compute.
There are plenty of well respected researchers who think only data and active learning loops are the bottlenecks. In my experience they're focused on treating the self driving task as a canned research problem and not a robotics problem. There are as many if not more respected researchers who've worked on the self-driving problem and see deeper seated issues -- ones that cannot be surmounted without technologies like high fidelity sensors grounded in physics and HD maps.
Even if breadth of data is the problem and Tesla's approach is supposedly yielding more data -- there is also the question of the fidelity of said data (e.g. the distances and velocities from camera-only systems are estimated and have noiser gaussians than ones generated with LiDAR). If you make what you measure, and your measurements are noisy, how can you convince yourself or your loss function for that matter that it's doing a good job of learning?
It's relatively straightforward to build toy systems where subsystems have something on the order of 95% reliability. But robotics requires you to cut the tail much further. https://wheretheroadmapends.com/game-of-9s.html
Secondarily, there is no guarantee that the amount of processing is enough, because the extant human systems use much more.
“Cheating” by using more sensors to simplify out complexities and to cover for the shortcomings of other sensors in the suite seems wise.
Also, "cheating" is just a necessary step to build baseline metrics. You need ground truth.
It may very well be the case that cheating is needed to generate the training data necessary to stop cheating...someday.
IMO the bar is well above human-level if you actually want to attain the level of trust necessary to remove the steering wheel from a car.
For one, frame rate and processing rate on human eyes is way higher than cameras. Dynamic range is another. Also, Cruise and Waymo are some of the only companies that have hard internal data / ability to simulate how well their safety drivers do, and in the very same scenario what their software driver will do. Without LiDAR you can't build that simulation, and once you have that data if you continue to use HD Maps and LiDAR there's probably a good reason.
> Dogs jump into the road in front of cars all the time and get killed, and kids get endangered at school bus crossings. That's a reality of life that robotaxis do not need to solve.
Robotaxis need to avoid any accident that a human would be able to avoid.
> IR and thermal sensing are unnecessary if the bar is human level
See, you could say this if you had some data that showed that incidents per X miles (when the vehicle is driving at night) is sufficiently low, + if the software passes some contrived scenarios to gut-check its ability to see in the dark with the necessary reliability. But you don't have that data, do you? Someone has it though :) and I'd argue regulators should have it too.
I don't think it's exciting to say that you must have theoretical parity with something to use it for this use case. Tesla's solution monitors ~6? cameras at once with accurate depth in each. That's 6x more views than a human can see. I wish people would stop comparing apples to oranges.
> Robotaxis need to avoid any accident that a human would be able to avoid.
I never said anything to the contrary. Animals get hit all the time, not just because a human wasn't paying attention.
No, the depth is estimated. It's not accurate, at least not in the way you need for L4.
> I never said anything to the contrary. Animals get hit all the time, not just because a human wasn't paying attention.
I was just clarifying what the bar is. The bar is that avoidable accidents need to be avoided. Nobody will get mad if a plane crashes due to unavoidable circumstances (freak accident where two engines go out due to bird strikes or something). People will stop flying in the plane when it becomes clear that the airline is not doing everything it can to avoid fatalities.
This one in itself continues to kinda surprise me. High end Audis have full-blown thermal cameras, and will literally flag heat producing objects on the console: https://www.audi-technology-portal.de/en/electrics-electroni...
Tesla does have HD ground truth data for verification generated by their own LIDAR-equipped vehicles. However, according to a recent tweet by Elon Musk [1], they don't need LIDAR for that anymore.
[1]: https://nitter.poast.org/elonmusk/status/1788064109633875999
I seem to recall Musk saying in the last couple years that "full self driving will basically require AGI." This appeared to me to be extremely honest and accurate, though I believe that in the moment he was trying to promote the idea that Tesla was an AGI company.
Does anyone happen to remember when he said that?
Waymo have said time and again they don’t rely on maps being 100% accurate to be able to drive. It's one of the key assumptions of the system. They use it as prior knowledge to aid in decision making. If they got "lost" whenever there was a road change, they wouldn't be successfully navigating construction zones in San Francisco as we've seen in many videos.
They can also do constant updates because the cars themselves are able to detect road changes, self update maps and rollout changes to the entire fleet. See https://waymo.com/blog/2020/09/the-waymo-driver-handbook-map...
At this point, the whole “HD maps are not practical” is just a trope perpetuated by the Tesla community for years.
Which leads to mapping failures being unchecked, as the system that generated the data is the one checking the data by driving it. See bullet point 1 in their recent recall for an example.
> Prior to the Waymo ADS receiving the remedy described in this report, a collision could occur if the Waymo ADS encountered a pole or pole-like permanent object and all of the following were true:
> 1) the object was within the the boundaries of the road and the map did not include a hard road edge between the object and the driveable surface;
> 2) the Waymo ADS’s perception system assigned a low damage score to the object;
> 3) the object was located within the Waymo ADS’s intended path (e.g. when executing a pullover near the object); and
> 4) there were no other objects near the pole that the ADS would react to and avoid.
and Tesla would do better how in this case? It also routinely crashes into stationary objects, presumably because the system assumes it wouldn't cause damage.
Are the Teslas in the room with you right now?
Please point out in my comment where I mentioned Tesla. I can wait.
> We’ve automated most of that process to ensure it’s efficient and scalable. Every time our cars detect changes on the road, they automatically upload the data, which gets shared with the rest of the fleet after, in some cases, being additionally checked by our mapping team.
Doesn’t mean it’s foolproof. But the benefits far outweigh the drawbacks.
Not a problem for Google, right?
> constant updating
I guess the cars can and will update the mapping in real time ?
> at best in certain cities
If mapping a city is possible, so it's mapping a highway, even easier.
If cars do update the maps themselves, they require might just a couple of human-driven passes of the standard WWaymo cars on a highway to generate the maps.
Waymo has shown they can make robotaxis work, but the big catch so far is that it takes them a long time to open in a new city. They have several phases before they open fully, from what I've seen it seems to be: safety driver no passenger testing, safety drivers with employee passengers, driverless with employee passengers, limited rollout to paid passengers under NDA, wider rollout but with waitlist, and finally getting rid of the waitlist.
This means that hitting even all the major metro areas in just the US is going to take them a long time, let alone the rest of the world (or at least developed world). That does give Tesla some time to potentially catch up, since they don't seem to be bounded by geography in the same way.
Now, that said, I personally don't think Tesla's strategy is workable except maybe the very long term. Doing this with only vision seems like taking something that was already enormously challenging and making it nearly impossible instead. Their slow progress and inability to get their cars to avoid even basic errors frequently, despite near a decade of development now, I think points to this strategy just being bad.
It's certainly true that they need to do a bunch of extensive mapping for each city, but I don't think we should expect their roll-out speed in later cities to be as slow as the first couple of cities. Most of the stuff they are learning in the initial roll-out will generalize to other location; it's not all city-specific learning.
1. Robotaxi is a better target than general self-driving because the human baseline is much lower for robotaxis (most people dislike their experience with uber, while most people think that they are a better-than-average driver)
2. Google took the high road on safety. The move-slow-and-dont-break-things DNA of Google (that hurts them in so many domains) is a golden asset in self-driving.
https://www.businessinsider.com/anthony-levandowski-google-s...
Waymo would not exist as it is today if that crash had any fair scrutiny.
Waymos do occasionally screw up, but if they did it as much as Tesla's FSD, it'd be chaos in the streets in SF, so it seems like it must be fairly infrequent.
So Waymo has 2.7 times more incidents with injuries then Teslas using autopilot have incidents, with or without injuries.
Maybe if I checked more sites they'd give different numbers, but from those initial numbers it seems your perception of reality of Waymo "screwing up" less is not accurate.
[0] https://waymo.com/blog/2023/12/waymo-significantly-outperfor...)
[1] https://insideevs.com/news/720730/tesla-autopilot-crash-data....
Tesla cars with FSD have a driver behind the wheel who can instantly take over if the car is about to crash into a stationary object. Any time a Tesla would've crashed into something an object but its human driver saved it, that doesn't count in stats like these. Many Tesla owners have reported that they have to regularly disengage FSD because it's trying to do something dangerous or looks like it's headed for a crash.
In contrast, Waymo cars do not have a human who can take the wheel if they try to run into a wall. The closest equivalent is that if Waymo cars get confused and don't know how to proceed, they can stop, then phone home and ask a human navigator to give them 'advice' or a general path; these people don't directly control the car, they're more comparable to a human navigator in the front passenger seat. It's still human assistance obviously, but it's not gonna save the car from running into an object that it didn't think was there.
With Tesla the responsibility is on the person in the driver's seat, so there is a (rightfully!) a bias for overreaction on the part of the driver. We will never know many of these disengagements were necessary.
The only way to get a true comparison of data is to compare robotaxis with robotaxis.
> The only way to get a true comparison of data is to compare robotaxis with robotaxis.
100% agreed. And so far, Tesla hasn't taken the step of actually letting the cars be driverless.
People saw the steering wheel turn and perceived it to be the system going haywire, or thought that "the car was about to crash" as you put it, and intervened.
The newer NN based FSD acts more like what a human would do.
And there's definitely cases where the Tesla in question just tries to run into parked cars or similar for no apparent reason, but Tesla fans always have some excuse about why that's irrelevant, especially if it's not on whatever the absolute latest version is.
Then they accuse the people horrified at Teslas making basic errors and trying to crash of being "anti-Musk".
The reality is that their progress has been slowly incremental and often mixed: https://insideevs.com/news/724479/tesla-fsd-12-4-1-real-worl...
There were some private driveway situations where the uploader intervened to back out to go to a new destination (but Waymo drops you off half a block away and makes you walk instead of entering your driveway, so it's not possible to compare). And there were some situations where a human driver honked - this has happened to me in Waymo as well. There was one situation where the Tesla didn't seem sure if it could proceed, but Waymo in that scenario would ask a remote operator (this has happened to me in both Waymo and Cruise) and presumably Tesla robotaxi can also have remote operators.
The only case where he actually disengaged was at a stop sign with a slip lane, and the car turned right at the stop sign instead of turning right using the slip lane. He went there again at the end of the video and the car used the slip lane. I don't see this as an unfixable problem, because clearly the car can use slip lanes to turn, it just needs to be taught to always prefer slip lanes when turning.
So, your own video disproves what you're saying. It isn't failing at basic driving tasks.
I think the mistake you're making is assuming that they will never be good enough. A lot of people said the same thing about Google/Waymo until they actually rode in one.
It's worth remembering that Waymo required their users to sign NDA's during beta, while the Tesla FSD beta was open to everyone with no NDA. So there was a lot more Tesla content being posted and going viral.
Meanwhile Tesla hasn't required NDA's except for employees running internal builds.
So essentially we're comparing footage from Waymo when it was at the end stages of its development to footage from Tesla at the early stages of its development.
I don't think we'll actually know until Tesla has an actual robotaxi product. When Cruise had one, most people who had tried both Cruise and Waymo said Waymo was better. That was my opinion as well.
Well, except for the fact that one is doing it completely driverless. And it has to do that every single time without having the luxury of a driver to prevent accidents.
Big difference in reliability, which makes them wildly different.
We can't make an apples to apples comparison until Tesla also has a robotaxi product, but even then there will be questions around the role of remote operators.
Well, no. A Tesla doesn't operate without a driver's supervision, so it can't be driverless. It did that particular drive without intervention, that's it. The stats [1] clearly show it's nowhere near capable of doing it without a driver in the seat. Community tracker puts them at 30 miles per disengagement.
It's not even at the point where the challenge is handling weird edge cases with construction or strange intersections, it's still struggling with not running into parked cars and walls. How many years do you think it should take a self-driving system to be able to handle those basic tasks in the general case? Because Tesla has been working on self-driving for almost a decade at this point, and they still seem to be barely past the starting line.
> it's still struggling with not running into parked cars and walls
Based on the phrasing of your sentences, I must ask which FSD version you are running, and if you can share footage.
> Because Tesla has been working on self-driving for almost a decade at this point, and they still seem to be barely past the starting line.
Waymo/Google has been working on this since 2009, which was itself based on a Stanford project (whose team was hired from Stanford to Google) that started in 2004. So that's either 15 or 20 years depending on how you count, and it doesn't include the much harder tasks of mass producing vehicles and making electric vehicles commercially viable.
I would imagine that while Waymo's mapping efforts have been very human effort-intensive so far, they will be looking at developing this automatic map-making capability as a high priority for rolling out new cities. Scaling the rate of expansion is then mostly a matter of throwing hardware and compute at the problem.
The very low take-up rate of FSD during the trial period indicates that most people are buying Tesla's because they are arguably the best EV with the best charging network.
Not because of FSD.
Whether that happens remains to be seen.
The rate of innovation at Tesla > Waymo
The cost of building Tesla FSD = 1/100 * cost of building Waymo FSD
The cost of delivering Tesla FSD = 1/10 * cost of delivering Waymo FSD
Tesla has economies of scale. Waymo has all the details figured out. Waymo can never get to the scale of Tesla (it can never buy 5 Million FSD cars, while Tesla is delivering them every 2 years)
Mathematically, Tesla has an upper hand over Waymo and it'll play out as that.
Larry, Sergei are extremely poor capital allocators. Musk is brilliant (despite him being a narcisstic a*hole).
Larry/Sergei left Waymo at a limbo state because they don't think in terms of economics, just coolness.
Waymo is successful enough to not kill it, but also not a cash-flow positive to scale it up
Edit : Tch, Tch expected HN anti-Musk hate showing up in downvotes.
They have driving assist that people still measure in 'number of times I had to grab the wheel'.
Larry/Sergei didn't create the Cybertruck
The kind of founders who knows shipping > waiting-for-perfection?
move fast and break things may be ok for social networks, but never ok for endangering the safety of others
to launch a 6th model of a marque, so well established, theoretically knowing a lot about cars, a model that dies when it gets wet, it's just embarrassing
If Tesla does manag to jump straight from level 3 to level 5, they have a chance to compete, but that seems unlikely. They also might move to level 4 and be able to expand level 4 coverage faster, but that still remains to be seen.
Waymo has years of experience with the other hard part of self driving taxis: actually picking up and dropping off people without a human driver.
Anti-musk partisanship frustrates me, but I suspect it is your fan-boy talking points that drive the downvotes of your comment
Technology progress is non-linear. Yes, Tesla is at L3 and Waymo at L4.
But, you completely missed the point of rate-of-innovation, which was why I made that as first point.
My GOOG holdings are 4x than TSLA. But in terms of who will deliver FSD at better margins, it's hands down TSLA. It's simple Math and Elon's obsession about cutting material costs & process to it's barest minimum. Waymo has no such discipline or culture.
FWIW - You definitely come off as as irrationally in favor of tesla, regardless of reason.
Nobody missed any of your points, they just don't agree with most them or believe the rest follow from your points.
Any driver that makes this kind of move is typically in the top 5%ile of driving skills. Yes, it's an asshole and slightly illegal move, but the level of intelligence that needs to be applied for this move is all you need to know about Tesla's advantages.
I have been extremely critical of Musk's reliance on only cameras, but I'm impressed with the progress they have made because of the new architecture in 12.x series. Considering it'll be trained with 100x more compute, I'm willing to bet that Tesla will overtake Waymo's capabilities.
Yes, it's a speculation and you can disagree, but unless proven you can't tell it's wrong
It was luck that the left lane was moving and that there was space. I don't want to see this kind of move from a driverless car.
A skilled driver always makes the smooth transition into those gaps. I know most HNers are goody-rule-followers and can never appreciate the skill required to make that maneuver consistently. But, this is a clear example of separation of intelligence / skill and I'm happy with the bet that these are signals of intelligence
What you call first-principle-thinkers, the rest of the world calls dicks. Everybody sees that manoeuvre, it doesn't require a genius. Most of the people don't do it because they are afraid of repercussions (if they get caught) or are civilised enough to realise that their time is not more important then the time of others. Yes, that manoeuvre doesn't "save you 5 minutes", it steals 1-2 seconds of everybody else's time.
Just read this https://fooledbyrandomness.com/ConvexityScience.pdf
From fig (1)
Waymo/ HNers : Knowledge Edge
Musk : Convexity Bias
That's why Musk's approach will win. Waymo will no doubt provide a safe, rule-following driving experience, but Tesla will have frontier breakthroughs and provide more human-like, adapting and sufficiently aggressive driving experience
So Tesla will necessarily converge to the performance of your average driver, because that's where the data leads it.
Waymo's is developing a new type of understanding and modelling of the world. It perceives and tracks items above the capability of humans to track and understand. Therefore its limits are outside of the bounds of the most capable human drivers.
Your own argument highlights exactly why Tesla's approach is a dead end (exactly in the same manners that LLMs will lead to the dead internet), while Waymo's approach will likely generate super-intelligence.
Hard disagree.
It's pretty clear the car just didn't understand what the line of cars was doing.
> level of intelligence that needs to be applied for this move
If you were a human making this move, I would judge your intelligence as below average.
Every idiot can drive like an impatient asshole. Choosing not to do is is the true sign of intelligence. The car is literally driving itself at this point. What does getting there three minutes faster so you can doom scroll Instagram from the lobby instead of doing that while your car is driving itself do for you? Are you more important than everyone else and thus deserve not to wait your turn?
If it helps, Waymo's will run red lights, so it's not like Tesla's got a monopoly on being a bad driver.
How is the "rate of innovation" higher but yet they've innovated less and have a less functional product? It's made up metric and was ignored because it adds no value to the conversation unless you have some actual data.
from Fig 1: https://fooledbyrandomness.com/ConvexityScience.pdf
Waymo (and most HNers) : Knowledge Edge
Musk / (Billionaires hated by HNers) : Convexity Bias
Tesla isn't even comparable to Waymo at this point because Tesla has zero level 4 capability. Teslas don't even have the redundant sensors needed for level 4+. All Tesla has is empty promises and an u unclear path for ever getting past level 3.
The cost of delivering Tesla FSD = 1/10 * cost of delivering Waymo FSD"
This seems like one of the key assumptions, but is not proven out at all because Tesla does not even have a level 4 vehicle. So the cost of delivering one comparable to waymo is infinite right now!
Your other key assumption is "Waymo can never get to the scale of Tesla (it can never buy 5 Million FSD cars, while Tesla is delivering them every 2 years)".
Both the assertion in the first part and the second part seem like super-strange assumptions, and not obviously true at all, yet are also critical to your analysis.
Waymo could get to the scale of tesla. It may or may not be too expensive to do. It could in fact, buy 5 million FSD cars. It may or may not be too expensive to do.
"Mathematically, Tesla has an upper hand over Waymo and it'll play out as that."
Or you know, if needed, Waymo could change?
It's funny to watch someone say "this one company will be able to adapt in every possible way to it's advantage, and nobody else can or will"
That almost never happens.
Your retort is then that you are getting downvoted because of anti-musk hate.
Have you considered that maybe you just don't have that good of an argument instead, and that your comment comes off as more of a tesla fanboy (regardless if you are) than a useful contribution?
I could write the literal opposite comment of what you did, in favor of Waymo. That would not be a useful contribution either for the same reasons.
Why not? They only do the integration, and traditional car and electronics manufacturers have solved the scaling problems long ago.
cost of Tesla FSD hardware add-ons : $5,000
Tesla benefits from vertical integration, which is super important in the innovation phase.
I'm still coming to this from an unbiased perspective. Personally, I'd be a lot richer if Waymo wins the FSD race.
2) Tesla has zero even remotely monitored, let alone autonomous, miles driven. So no, there is no reason to believe Tesla is close to true robotaxis.
These guys are all about tech and couldn't believe there were companies ahead of Tesla, what do you think the normies know?
Tesla has much lower costs. If they can beat Waymo on customer experience (better driving primarily, but also better in-car entertainment, better mobile app, and match Waymo's pricing) they'll win.
Waymo might have a regulatory advantage since a lot of politians don't like Musk.
Yes there can be cleaning fees and vetting, but all it takes is one or two people puking in your Tesla, and you'll have no interest in providing a Robotaxi for Musk's Mission.
> Yes there can be cleaning fees and vetting, but all it takes is one or two people puking in your Tesla, and you'll have no interest in providing a Robotaxi for Musk's Mission.
Plenty of Tesla owners (and owners of all sorts of other vehicles including high-end sports cars and luxury vehicles) already rent out their vehicles to strangers on Turo. Listing them on Tesla's platform instead won't be a large change. A lot of owners might actually prefer it if it allows more fine-grained options (eg. no rides past 9pm, no rides more than 50 miles from home) instead of renting it out for 24 hour periods.
The entry bar for renting a vehicle on Turo is quite a bit more substantial than hailing an Uber. I looked at making my cars available.
You can restrict the age of the driver (anything vaguely high end I see required you to be 30+), require substantial deposits ($750), proof of your own insurance, mileage and location limitations, and more. Versus "Downloaded the Robotaxi app and put in a credit card number".
The Tesla-owned fleet will likely have the same requirements as riding an Uber. If the rumors are to be believed, those will be dedicated robotaxi vehicles with no steering wheel.
Of course for peer-to-peer you will be able to set your own requirements based on what you're comfortable with, or even just not join the network at all so your car stays exclusively yours.
They only have a car that can drive with a human partner behind the wheel who's ready at any second to take over and prevent it from crashing.
Even that is a faulty system.
I really don't understand how people can think Tesla and Waymo are anywhere close to each other. Making a car drive itself is a much harder problem to solve than scaling production. If you look at what Tesla has, today, it is a vehicle that can sometimes drive on its own with the need for a human behind the wheel. The ability to remove the human from the front seat is the other 95% of the work. In systems like this where human lives are at stake, getting "mostly there" with FSD is essentially worthless.
The problems of logistics and scaling are well-known and well traveled. The invention of a self-driving car is not. Once Waymo has solved the issue of entering new cities and quickly gathering data to feed into their model, they've won.
I think it's possible, but it's definitely less trivial than scaling production of more self driving cars.
- Elon (and his pro-analysts) heavily weight the future of the co's valuation on their ability to deploy a taxi network and has been promising it just around corner for years
- Alphabet via Waymo seems to have "solved" robotaxis for city-proximity driving and has deployed as a business.
Beyond the obvious "reality distortion field" argument is Tesla actually in a position to win here due to their manufacturing capability / current deployment of Tesla's?
Disclaimer - I am an Alphabet & Tesla shareholder