In some sense, the visionaries in this space are not thinking big enough. I want visions of mobility with a totally different size, look, speed, etc. autonomous Golf carts? tuktuks? A moving autonomous bicycle carrier? etc
Like imagine a low speed, electric, autonomous, golf-cart-only lane at every train station, for the last mile.
The lead that Waymo has acquired in perceiving its driverless car's environment will be almost impossible to kill. In about 5 years, it'll be like NVidia and CUDA. Tesla's choice to abandon lidar will be one of the biggest oof in business history.
Driverless personal transportation is the unsolved problem.
Even in Japan, half of commutes are by car and that number has been growing.
Neal Stephenson wrote a short essay on path dependence that I really like-- https://slate.com/technology/2011/02/space-stasis-what-the-s....
I do see people driving horse-drawn carriages, ATVs (probably illegally), snowmobiles (legally in some parts of MI during Winter or condition-dependent), and riding mowers (probably illegally) in and around towns, though. Very rarely, I see someone driving an e-bike; this rareness is mostly because they aren't allowed on the sidewalks here and there's no bike lane, so you need to drive and signal like a car, which is pretty awkward given how many e-bikes don't even come with real brake lights (though many falsely advertise red rear running lights as a brake light, which'd be illegal to drive unless you hand-signal whenever you brake).
Well, I guess you are not as well traveled in the Midwest as you think.
Sun City, Arizona, though these are golf communities/mega-master-planned communities. Coronado is a better example of a mixed vehicle environment with golf carts bopping around all the time on the same streets.
Pretty convenient when I unexpectedly find myself needing to use a parking garage and such. The scooter can take me out of the parking garage and into the building with no issue. And then I can keep it with me in the building until it's time to get back to the car.
It's also probably cheaper than a golf cart - mine was just about $3,600 brand new. Though used carts are probably cheaper still, and there are also much cheaper scooters.
I actually used to use only an electric scooter for transit, but then I got hit by a pickup truck who didn't check the bike lane before turning. So I did driver's ed, got my license and leased a BEV.
What you envision might happen in 2100+
https://www.cnet.com/home/internet/google-fibers-secret-weap...
https://gizmodo.com/when-google-fiber-abandons-your-city-as-...
And the HN discussion
If anything, it's the opposite: most people in this space (Elon, George Hotz, Demis, etc.) have been saying for a very long time that autonomous driving is just the first step, and that their objective is to build world models.
Also, Tesla started FSD in 2016. The very core of their strategy was (and is) to sell $40k car with hardware capable of running FSD.
Cameras are super cheap, FSD chip is reasonably inexpensive. Lidar is not. Maybe today the cost isn't completely prohibitive (I think it still is, because you need multiple lidars) but it certainly was for the first 8 years of FSD program.
Tesla just didn't have the luxury of adding $50k to the cost of the car for the hardware, the way Waymo did. And they didn't have sugar daddy (Google) willing to burn several billions a year for many years.
So the Waymo approach was not an option for Tesla.
And given that in Austin they just reached parity with Waymo (i.e. completely unsupervised robotaxi service), they are not doing badly.
There is no unsupervised robotaxi service in Austin and there won't be, for years, if ever. Just like the way "FSD" is not fully self driving and likely never will be.
https://electrek.co/2026/01/22/tesla-didnt-remove-the-robota...
However the area it operates is extremely small, and they are still only allowing Tesla bros to try it.
Why does anyone want to do business with a person or company like that? I genuinely do not understand.
It's unsurprising someone so out of the loop on its true status is so hyped on it's future...
If robotaxi was doing legit rides, Elon would be posting about it 20 times a day.
Elon has been blatantly lying about FSD for years, and yet the fans still take whatever he says as gospel. And yet the skeptics are the ones with EDS? lol, ok.
A year? They'll be gone in two weeks!
Seriously, what portion of your financial and emotional net worth is tied up in TSLA?
That's a highly ironic statement given your position on "cost per mile".
With a small amount of business acumen, you'd know that betting on technology staying expensive is a bad idea. This is seen in all industries, but especially electronics, where there are many competitors continuously optimizing for cost. E.g., we're at the point now where an internet enabled phone is basically disposable, costing people ~ a few hours of wages.
History has shown that technology costs decrease over time, and rapidly if it's a critically important technology. If you don't agree, share a counter example.
That is a lot of confidence. Do you work in the autonomous vehicle space?
What makes you so certain?
I guess they understand that computer vision is a fast-moving target and their paper might become obsolete the next day.
When do you think it will be reliable enough?
If Tesla had been smart they would have used regular cameras and event based cameras where the pixels send a signal whenever their brightness changes enough. These can have microsecond latency. And multi spectral cameras. Combined this data would provide very rich data for neural networks.
Tesla is far behind Waymo on all meaningful measures.
Waymo sells more than 450k rides every week. Tesla is nowhere near that number.
Waymo offers rides in six cities. Tesla does two.
According to https://robotaxitracker.com/ Tesla has ~250 taxis in total. Waymo has +2500.
The bottom line is cost per mile and Waymo can't complete here, there is also style, Waymo's vehicles are extremely ugly looking cars vs the Cybercab. Tesla also has integrated everything from the chip up. Waymo is a cobbled together solution from multiple third party (very expensive) components.
Is the consumer going to pick a more expensive, ugly, non integrated vehicle for their trip?
I am only refuting the claim that Tesla has reached parity with Waymo in Austin. They are nowhere near.
Because Tesla has a history of over-promising and under-delivering, I will want to see Tesla scale up the robotaxi business to the level of Waymo (which is currently far ahead) before I proclaim them the winner.
You are not really backing your claims with facts or numbers, just opinion and future predictions which may or may not come true.
Tesla's market cap is $1.3 trillion. Granted the company itself doesn't have access to all of that, but surely if they wanted to spend, say, $10 billion per year on something big like FSD, they could have.
> didn't have the luxury of adding $50k to the cost of the car for the hardware
A little more extreme, but: Tesla has sold something like 8.5 million cars total. If they simply dumped an extra $50K of material into every single one of those cars without raising the price a dime, that would be only $425 billion. That's a ridiculous sum of money, but still <checks notes> substantially less than $1.3 trillion.
> If they simply dumped an extra $50K of material into every single one of those cars without raising the price a dime, that would be only $425 billion. That's a ridiculous sum of money, but still <checks notes> substantially less than [their market cap of] $1.3 trillion.
That is an apples to dishwasher comparison. Money is fungible only when it's the same kind of money on both sides. You can't compare market cap like that. (Even for a company whose market cap is seemingly divorced from reality like Tesla's)
TSLA market cap was about 50B for the first several years of their FSD effort.
I think they'd choose lidar if they started now.
I'm giving ballpark numbers because I am in this space and don't want to dox myself.
Parity is not defined by how willing one is to let their robots kill the general public.
Nobody is talking about any of this using past tense. It is 2026 now, not 2016.
Tesla's design team prioritized form over function. Lidars definitely look ugly; they didn't want them on their cars, so as a consequence, they shoot themselves on the foot.
I don't know who will end up being right in the long term, however I don't think this was a form choice, I think they believe a pure camera system will be more functional.
Autonomous driving is a solved problem. The fact that self-driving cars are not permitted on most of the world’s roads is 100% the fault of regulators and those who vote for them.
> However, accidents involving Advanced Driving Systems occur more frequently than Human-Driven Vehicle accidents under dawn/dusk or turning conditions, which is 5.25 and 1.98 times higher, respectively.
Maybe a part of the fault is also the self driving systems themselves that keep crashing and killing people. I don’t have data to say with certainty whether self driving cars are safer or less safe than humans(I think they are less safe especially if taken out of their “trial” zones/sampling) but I can tell you that is less acceptable for self driving cars to kill people than it is for humans for “obvious” reasons especially when the self driving cars do this due stupid mistakes that a human driver would not (I.e goes straight into another car in plain day).
Still too big tho maybe. What about a Segway-sized vehicle, or even smaller.
Which is why their strategy (purely vision/photons in, controls out) seems to be more widely applicable and scalable over time.
And waymo seems to be arriving there too as they keep reducing the equipment (it would seem)
This yes.
> The lead that Waymo has acquired in perceiving its driverless car's environment will be almost impossible to kill.
This, I don't think so. I think it'll be more like the space race. Or the LLM race. Anytime money or data is all that's required, you won't hold the lead forever. The reason big tech holds their leads today is not innovation, but critical mass combined with user entrapment. Waymo is not positioned right now for either since their space is primarily focused on taxis, whereas the real winners (in auto) will be whoever does it best (and there may be a few) for consumer auto ownership.
We can talk about robots all day, but we haven't gotten to mass robots yet because of cost and reliability. It'll be a bit still for those to work and it won't surprise me if robots end up in homes and wars sooner than factories, since those former use cases are shockingly more fault tolerant than a high paced environment.
And regulatory capture by the incumbent. Reach the top then push for regulation behind you. Thats’s one big additional obstacle to overcome for a new player.
OpenAI was so willing to support regulating AI just as soon as they thought they’ve gained enough of an advantage over the competition and they can burn the bridge behind them.
Why? They have started unsupervised taxi rides in Austin. One of their goals was affordability, and their cars are massively more affordable.
Waymo is too deep in their complex hardware stack to do a hard about face at the moment.
Waymos are geofenced, restricted to certain roads, and have remote humans in the loop at lot more than most people assume.
Tesla's current robotaxi deployment is also geofenced and monitored by humans -- moreso than Waymo, even -- but of course Tesla superfans always conveniently leave that out of the narrative.
They will turn of the supervision requirements soon, and suddenly there will be hundreds of thousands of teslas that can drive themselves.
The skepticism is hard to get.
(Disclaimer: former Cruise employee)
If you look at it from an outside point of view, right now Tesla is worth $1.6T, Waymo is worth $130B, and GM is worth $72B. If Cruise were actually a third viable competitor in this race, it would probably be worth more than the rest of GM. Self-driving is just a far more valuable business than car-making.
So from that point of view it would make sense to say, don't worry about the rest of GM too much, you should be willing to sacrifice all of that to increase the changes of making Cruise work.
It's hard to change the culture at a place like GM though. Does the GM CEO really want to take a huge amount of risk? Would they be willing to take a 50-50 shot where they either 10x the company's value or lose it all? Or would they prefer to pay a few billion dollars to avoid that risk.
I think the bigger issue is that Cruise was not succeeding at building the driver.
Cruise was shutdown after a safety incident, same as Uber.
This has been a pretty consistent pattern -- Cruise was always less transparent about its safety data than Waymo, and its claims tended to be opaque and non-measurable, whereas Waymo was partnering with insurance companies to get hard data.
Waymo is going to have incidents, too, but I think they have made the (correct) decision that being open and transparent about safety stuff is the way they move past those; Cruise made a decision in the opposite direction, and it killed them.
At the same time, if Musk went away, the stock would crash back to reality but a non-idiot leader could just do impossible, crazy, hard stuff, like ... working on obvious new models and basic steady improvements.
Tesla PE is 398 today (after a drop). Toyota's PE is 13. Toyota at the least is not hemoraging market share, sales, revenue, profits. Tesla is losing on all thoes things. Tesla would need a 30x price reduction to get down to much much more stable and profitable toyota. It's gets worse because Tesla's sales and profit keep going down each quarter.
There's no doubt value in self driving but the overall value is questionable. If there are many companies providing it, and at least waymo is doing great, plus there are many many other companies in China in good shape, the value multiple won't be there.
What's the market value of all taxi compannies combined in the us? It was about $230 billion in 2024 (https://www.skyquestt.com/report/taxi-market). Will tesla get 100% of the us self driving business in the future? No, waymo at least will be a serious market competitor, tesla's service doesn't really work.
Because there are going to be muiltiple competitors with working products (we'll see if/when tesla ever gets there), Tesla's huge valuation will never make sense. Robots are much farther behind than robotaxis (there's no brain, no prototype of a learning system, maybe one day).
This got way too long, I think GM just saw it as a money sink. I think that was a big mistake, though.
That said, you could be right! Maybe self-driving will never be worth more than that. It's really hard to tell what business models will be like in the future. But this is the cultural mismatch, it seemed like GM leadership did not want to be in a risky business where they were betting billions of dollars on the success of self-driving. Clearly, to some people, that seemed like a really good bet to make. Time will tell.
Having experience and capability to manufacturer cars has approximately zero benefit to create a self-driving software/sensor stack. It would make more sense for Adobe to create a self-driving car than GM.
Instead they chopped it up for spare parts, specifically, sending some Cruise personnel to work on deadend GM driver assistance tech and firing the rest. Baffling.
(Also former Cruise employee)
(Another former employee)
(Another former).
Cruise was always destined to be "like Waymo, but worse". Tesla, on the other hand, is taking a very different path than Waymo, they have a chance at beating Waymo at their own game and even if they don't beat Waymo, they can be a winner in some specific niche. (For the record, I'm a fan of Waymo.)
2. Going directly for vision-only, no geofence system. Waymo's strategy has been to start with a proof-of concept and gradually expand geography and capabilities.
That's how we get Uber, Lyft, DiDi, Grab, Bolt, WeRide, BlackWolf...
The winner-takes-all aspect comes from economies of scale for example. If Waymo is several years ahead and has better economics, how can Cruise catch up? They will have lower cost of capital. Top people will want to work at Waymo rather than Cruise. It will be hard to close the R&D gap. When Cruise is where Waymo is today, Waymo will already have a lightweight package that will be used by Ford, VW or Toyota and capable to drive everywhere in any weather conditions.
I'd love a really good driver assistive system for my car (ala FSD today) but I likely wouldn't actually get driven around a lot more unless the economics were more compelling than seem likely anytime soon.
People who yell "share" forget that most people are driving during rush hour, and so the car will be idle the rest of the day anyway. As such there isn't that much money to save by sharing a car.
Absent driving yourself at all, you're going to be very limited most places. Which may be OK for continuing a college lifestyle but mostly doesn't work unless you have a partner who handles the driving.
Car arrives. I get in. The car is sitting there getting ready to depart but not moving. After a few minutes I hit the button to call support. Someone tells me it's about ready to go. Ten minutes later it starts leaving.
I have no idea why it took so long to start but it wasn't a great experience.
If you (or anyone else from Cruise) can explain what was going on, that would settle the difference in experience to me.
Check out their history of EV or hybrid vehicles or even the history of Saturn - they stumble onto something awesome that people love and it's the company mission to destroy it.
GM pulled the rug on us a day or two before announcing. The current Cruise CEO wasn't aware at all either. I have my own conspiracies of why GM did this, but GM also has a long history of fumbling the ball.
[1] https://www.cbsnews.com/sanfrancisco/news/nhtsa-robotaxi-cru...
[2] https://www.theautopian.com/here-are-five-times-gm-developed...!
He said they were pretty awful and would constantly mess up.
Nice abbreviation.
If a leaf lands on your windshield, you can look beside it or move your head to see around it. If a leaf lands on a camera lens, it blocks it.
A pair of cameras mounted in the same place as human eyes, with the freedom to move a bit would be a fairer comparison. (The cameras would probably see better…)
It's weird how many people there are like that still.
But what they mean is that they are putting the new release into production (without backup drivers). They have been fully autonomous for many years.
Fleet response: Lending a helpful hand to Waymo’s autonomously driven vehicles
Much like phone-a-friend, when the Waymo vehicle encounters a particular situation on the road, the autonomous driver can reach out to a human fleet response agent for additional information to contextualize its environment. The Waymo Driver does not rely solely on the inputs it receives from the fleet response agent and it is in control of the vehicle at all times. As the Waymo Driver waits for input from fleet response, and even after receiving it, the Waymo Driver continues using available information to inform its decisions. This is important because, given the dynamic conditions on the road, the environment around the car can change, which either remedies the situation or influences how the Waymo Driver should proceed. In fact, the vast majority of such situations are resolved, without assistance, by the Waymo Driver.
https://waymo.com/blog/2024/05/fleet-response/
In other words, much like Waymo tries to put a nice spin on it, their cars are not fully autonomous and despite the wording of the article above, they are not "operating a fully autonomous service". Nor can the Waymo Driver "confidently navigate the "long tail" of one-in-a-million events" it "regularly encounter[s] when driving millions of miles a week".
They have remote safety drivers. Not fully autonomous. "Fully autonomous" is their aspiration marketing, but not their current reality.
1. They're not "safety drivers" in the sense that most people understand, ie. someone dedicated to watching the car
2. What's with the fixation on defining "fully autonomous" to mean 0% human intervention ever? If a vending machine works 99% of the time, and 1% of the time needs some technician to come to get a drink unstuck does it make sense to get up and arms about how it's not "fully automated" or whatever? In all contexts why people would care (eg. unit economics, safety, customer experience), there's no meaningful difference between 99% autonomous and 100% autonomous.
Yeah, good point. If Waymo were honest they'd say their system is "autonomous". Fully autonomous implies 100% autonomy. Otherwise, how is it "fully"?
But, hey, don't ask me. Write a paper with robot that is 99% autonomous but a human has to take control every once in a while and see how easy you can get that past any reviewer in robotics or AI.
It's like that time with Facebook and MySpace. A while ago now. I was in a student group at uni and this student, call her Alice, asked me for my Facebook. I said I don't have one, I don't like Facebook, and the conversation continued. Later another student came in, call him Bob. Alice told Bob "Where were you, we just had a big fight about Facebook over Myspace". I asked when that happened since I was there and didn't remember it and Alice said "that was me and you. We had a big fight about it. Did you forget?". I said, nonplussed, that I didn't think we had a fight. "But you said you don't like Facebook. So you like MySpace". Said Alice. Oh Alice.
From that I understand that you, like Alice, must be a very astute observer of human behaviour. No hidden motive stays hidden for long, with you, does it? Well done. You got me. I'm a Tesla fanboi. That's what I am, through and through.
> The Waymo Driver does not rely solely on the inputs it receives from the fleet response agent and it is in control of the vehicle at all times.
…
> The Waymo Driver evaluates the input from fleet response and independently remains in control of driving.
Waymo prevaricates about the "influence" the human operator has on the path taken by the Waymo Driver [1] but it is clear there are situations that the Waymo Driver cannot choose point A and point B on its own, at least safely, otherwise Waymo would not be paying for humans to do it. They'd let the system do it on its own. It can't. It's not "fully autonomous".
We can play with words and accept whatever terminological obfuscation Waymo wants to impose in order to pimp its wares, or we can accept that current systems have limitations, and choose to understand the real SOTA over marketing.
_____________
[1] Fleet response can influence the Waymo Driver's path, whether indirectly through indicating lane closures, explicitly requesting the AV use a particular lane, or, in the most complex scenarios, explicitly proposing a path for the vehicle to consider idib.
So we just made driving a million times more efficient for human labor input
Even if human controllers actually could pay attention 100% of the time, they'd struggle to respond in time to a lot of dangerous situations. Most accidents happen when one of the cars (or their drivers) fails to react in time with what is effectively a split second decision.
Autonomous driving (with or without a controller) means a computer takes essentially all of those decisions for the simple reason that any human controller would probably be too late way too often.
Once you accept that simple logical reality, the role of that controller becomes more clear: they are there to step in and provide instructions to the car when it encounters some challenging situation and slows down, pulls over, or stops in a safe place. This probably doesn't involve any joysticks or steering wheels.
Controller responses are not real time critical. They can't be. It would fail to work too often. Also, most controllers probably need to monitor more than one car. Which only makes the problem worse. And they might have to juggle two stuck cars at once.
Mostly autonomous cars are pretty good at object detection and not crashing into stuff (all the real time stuff). It's object classification and interpreting complex situations where cars get stuck or might sometimes still do dangerous/illegal/sub optimal things. Getting stuck or slowing down is fine. The human controller can fix that. Doing the wrong thing is more problematic.
Not 99% of a chauffeur, 100%. (or 99.99999%)
The roll out of this is clearly limited by the number of remote employees that are filling in the 1%.
In this very thread plenty of people are saying that what Tesla are doing now in Austin is NOT fully autonomous, but you assert Waymo doing the same for many years is?
Waymo had remote operators who could take over when needed for a long time.
Nice dig at Tesla.
There's a partnership with Toyota related to this: https://waymo.com/blog/2025/04/waymo-and-toyota-outline-stra...
“Vision is all you need.”
But I note that human drivers do not rely on LIDAR or radar or a high res map of the road.
Fun fact, people can see light polarization.
Human drivers also have a brain that computers have struggled to replicate for decades.
Also, many of the Chinese vehicle makers already include LIDARs in their vehicles for their current ADAS
Sure, Wayve did some experiments with just radar + vision, but now incorporate lidar across their fleet. Waymo, Pony.ai, and WeRide have always used it.
You can shout "you don't need it!" as loud as you like, but the people who are actually building and running these systems seem to disagree.
They might get there one day but, since they've been breaking promises about it for longer than those other companies have been in existence, at this point it's reasonable to treat what they have to say with extreme scepticism.
It would be an absolute catastrophe for the developers of autonomous vehicles to decide there's no need to do better than human drivers.
I read the whole thing, but, idk, surprised they didn't include a picture or clarify if this is strictly hardware, or hardware + software changes (with the software changes maybe back propagating to existing Drivers)
"Because we are focused on building a Driver and not a vehicle, we’ve designed a versatile, integrated autonomous driving system that can be adapted to various platforms and use cases over time. Our versatile hardware approach allows us to reconfigure our sensors and generalize our AI to meet each platform's unique needs—whether it is the Ojai or the Hyundai IONIQ 5—providing the Waymo Driver an optimal view of its surroundings while streamlining for efficiency."
ie this is a sensor+software package for any vehicle that they install on.
My understanding of the text is that, to get this to run on the existing fleet, they'd need to go into the shop for sensor/computer replacement, but the text isn't explicit about that.
Waymo announcements tend to be very incremental, each one is only a small change from a prior known state. They seem to operate an attitude of least possible surprise, probably to avoid spooking anyone about scary robocars.
The image caption reads:
Compared to a traditional automotive camera (right), the 6th-generation Waymo Driver camera (left) delivers significantly higher resolution at cost parity, allowing the system to make better-informed driving decisions.
> Our experience as the only company operating a fully autonomous service at this scale has reinforced a fundamental truth: demonstrably safe AI requires equally resilient inputs. This deep understanding of real-world requirements is why the Waymo Driver utilizes a custom, multi-modal sensing suite where high-resolution cameras, advanced imaging radar, and lidar work as a unified system. Using these diverse inputs, the Waymo Driver can confidently navigate the "long tail" of one-in-a-million events we regularly encounter when driving millions of miles a week, leaving nothing to the imagination of a single lens.
The fact that people still trust him on literally anything boggles my mind.
No single sensor can ever give you that kind of resilience. Sure, it is easy in that you never have ambiguity, but that means that when you're wrong there is also nothing to catch you to indicate something might be up.
This goes for any system where you have such a limited set of inputs that you never reach quorum the basic idea is to have enough sensors that you always have quorum, and to treat the absence of quorum as a very high priority failure.
Fail safe, always. That's what I tried to get at with 'absence of quorum', it means you are in uncharted territory.
You have an extremely detailed world model including a mental model of the drivers and other road users around you. You rely on sight, sound, experience and lots of knowledge. You are aware of the social contracts at work when dealing with shared resources and your brain is many orders of magnitude more powerful than any box full of electronics.
What you can do with 'just vision' misses the fact that you are part of the hardware.
Admitting this would be admitting their Tesla will never be self driving.
Having more sensors is complicating the matter, but yes sure you can do that if you want to. But just using vision simplifies training a huge amount. The more you think about it, the stronger this argument is. Synthesising data is a lot easier if you’re dealing with one fairly homogenous input.
But the real point is that cameras are cheap, so you can stick them in many many vehicles and gather vast amounts of data for training. This is why Waymo will lose - either to Tesla or more likely a Chinese car manufacturer.
I do not like Elon because I do not think nazi salutes or racism are cool, but I do think Tesla are correct here. Waymo wins for a while, then it dies.
So the "we can train cheaply because of lots of cameras" falls down when, for example, BYD has all of its cars with lidar for ADAS but can collect the data for training as well as the vision from cameras and whatever other sensors like tyre pressures and suspension readings and all the other sensors that are on a modern car.
The argument that we can make the cars cheaper in the future by not collecting the additional data now has been proven wrong by the CN and KR manufacturers.
That's also independent of the whole EV side of things.
It's just that the cost of lidars are falling like crazy, with new automotive lidars using phased-array laser optics instead of what waymo started with (mechanically scanned lidars)
Which seems like a very bad assumption, I'm not even sure it was ever true and is getting less and less true.
Waymo gets limited data from very limited locations, and will have a harder time synthesising data than others.
But (some) humans have the ability to handle difficult situations, and no autonomous system gets anywhere close to that. So this is more of a "robots handle the easy 80% better, but fail hard on the rest of the 20%". Humans have a possibly worse 80% performance, but shine in the 20%.
If you include minor fender-benders and unreported incidents, estimates drop to around 100,000–200,000 miles between any collision event.
This is cataclysmically bad for a designed system, which is why targets are super-human, not human.
So while he might turn out to be wrong, I don't think his opininon is uninformed.
However, if you think about this for 2 seconds with even a rudimentary understanding of sensor fusion, more hardware is always better (ofc with diminishing marginal value).
But ~10y ago, when Tesla was in a financial pinch, Musk decided to scrap as much hardware as possible to save on operational cost and complexity. The argument about "humans can drive with vision only, so self-driving should be able to as well" served as the excuse to shareholders.
In May 2016, Tesla Model S driver Joshua Brown died in Williston, Florida, when his vehicle on Autopilot collided with a white tractor-trailer that turned across the highway. The Autopilot system and driver failed to detect the truck's white side against a brightly lit sky, causing the car to pass underneath the trailer.
Our eyes are supported by our brain's AGI which can evaluate the input from our eyes in context. All Tesla had is a camera, and it didn't perform as well as eyes + AGI would have.
When you don't have AGI, additional sensors can provide backup. LiDAR would have saved Joshua Brown's life.
This field is extremely complex, it's often better to pick a sensor and stick with it rather than trying to figure out how to piece together data from very dissimilar sources.
> This field is extremely complex, it's often better to pick a sensor and stick with it rather than trying to figure out how to piece together data from very dissimilar sources.
Whether sensor fusion makes sense is a highly domain specific question. Guidance like "pick a sensor and stick with it" might have been correct for the projects you've worked on, but there's no reason to think this translates well to other domains.
For what it's worth, sensor fusion is extremely common in SLAM type applications.
The cameras on Teslas only really lose visibility when dirty. Especially in winter when there's salt everywhere. Only the very latest models (2025+?) have decent self-cleaning for the cameras that get very dirty.
This vision clearly doesn't scale to more complex scenarios.
The thing to remember about cameras is what you see in an image/display is not what the camera sees. Processing the image reduces the dynamic range but FSD could work off of the raw sensor data.
There is an argument for sure, about how many sensors is enough / too much. And maybe 8 cameras around the car is enough to surpass human driving ability.
I guess it depends on how far/secure we want the self-driving to be. If only we had a comprehensive driving test that all (humans and robots) could take and be ranked... each country lawmakers could set the bar based on the test.
Meanwhile Waymo is doing half a million rides a week, and Tesla is doing what, a few dozen? Maybe? Maybe zero? Who knows, because they lie and obfuscate about everything. Meanwhile I can go take a Waymo right now in cities all over America.
Given the history on this topic: https://motherfrunker.ca/fsd/
It's not unreasonable to distrust anything Elon says, especially about Tesla/self driving.
It's literally linking to direct quotes from Elon, can you explain the problem with that?
I would firmly disagree with that.
What Musk has done is bring money to develop technologies that were generally considered possible, but were being ignored by industry incumbents because they were long-term development projects that would not be profitable for years. When he brings money to good engineers and lets them do their thing, pretty good things happen. The Tesla Roadster, Model S, Falcon 9, Starlink, etc.
The problem with him is he's convinced that he is also a good engineer, and not only that but he's better than anyone that works for him, and that has definitively been proven wrong. The more he takes charge, the worse it gets. The Model X's stupid doors, all the factory insanity, the outdoor paint tent, etc. Model 3 and Model Y arguably succeeded in spite of his interference, but the Dumpstertruck was his baby and we can all see how that has basically only sold to people who want to associate themselves closely with his politics because it's objectively bad at everything else. The constant claims that Tesla cars will drive themselves, the absolute bullshit that is calling it "Full Self Driving", the hilarious claims of humanoid robots being useful, etc. How are those solar roofs coming? Have you heard of anyone installing a Powerwall recently? Heard anything about Roadster 2.0 since he went off claiming it would be able to fly? A bunch of Canadian truckers have built their own hybrid logging trucks from scratch in the time since Tesla started taking money for their semis and we still haven't seen the Tesla trucks haul more than a bunch of bags of chips.
The more Musk is personally involved with a project the worse it is. The man is useful for two things: Providing capital and blatantly lying to hype investors.
If he had stuck to the first one the world as a whole would be a better place, Tesla would probably be in a much better position right now.
SpaceX was for a long time considered to be further from his influence with Shotwell running the company well and Musk acting more as a spokesperson. Starship is sort of his Model X moment and the plans to merge in the AI business will IMO be the Cybertruck.
I never claimed he‘s a good engineer, nor that he has high EQ, nor that he is honest, nor that he has sole responsibility for the success of his companies.
Long-distance amateur psychology question: I wonder if he's convinced himself that he's a smart guy, after all he's got 12 digits in his net worth, "How would that have been possible if I were an idiot?".
Anyway, ego protection is how people still defend things like the Maga regime, or the genocide; it's hard for someone to admit that they've been stupid enough to have been fooled to vote for "Idi Amin in whiteface" (term coined by Literature Nobel Prize winner Wole Soyinka), or that the "nation's right to self-defense" they've been defending was a thin excuse for mass murder of innocents.
At all the other topics he couldn't even name the field. The only thing he is good at is scamming people dumb enough to fall for this.
This is - to me - entirely separate from the fact that his companies routinely revolutionize industries.
One of his latest, on the topic of rain/snow/mist/fog and handling with cameras:
"Well, we have made that a non-issue as we actually do photon counting in the cameras, which solves that problem."
No, Elon, you don't. For two reasons: reason one, part A, the types of cameras that do photon counting don't work well for normal 'vision'/imagery associated with cameras, and part B, are not actually present in your cars at all. And reason two, photon counting requires the camera being in an enclosed space to work, which cars on the road ... aren't.
What Elon has mastered the art of is making statements that sound informed, pass the BS detector of laypeople, and optionally are also plausibly deniable if actually called out by an SME.
I shit on Tesla and Elon on any opportunity, and it's a shame they basically have the software out there doing things when it probably shouldn't, but I don't think they're that far behind Waymo where it really matters, which is the thing actually working.
I love this argument because it is so obviously wrong: how could any self aware person seriously argue that hearing, touch, and the inner ear aren't involved in their driving?
As an adult I can actually afford a reliable car, so I will concede that smell is less relevant than it used to be, at least for me personally :)
Not to mention possibly the most complex structure in the known universe, the human brain: 86 billion neurons, 100 trillion connections.
If it makes you happy, you can read "only vision" as "no lidar or radar." Cars already have microphones and IMUs.
2. since this is in context of Tesla: tesla cars do have microphones and FSD does use it for responding to sirens etc.
Beating human sensors wasn't hard for over a decade now. The problem is that sensors are worthless. Self-driving lives and dies by AI - all the sensors need to be is "good enough".
Well, in TFA the far more successful manufacturer of self driving cars is saying you're wrong. I think they're in much better position to know than you :)
But I also don't think we can take anything from what Waymo claims about the feasibility of vision-only.
A favorite of mine: https://x.com/dmitri_dolgov/status/1900219562437861685
The DMV required them to submit their safety data before allowing the vehicles on the road. Waymo claims that the data the DMV needed about the public safety of their cars, and the emails they exchanged about it with regulators, were entirely "trade secrets" to keep them hidden from the public who understandably felt like they should be able to access that information since those cars are going to be on their streets.
For all the impressive technological advances Waymo makes (and don’t get me wrong, they are impressive), their cars are still a constant obnoxious menace to drivers.
The places I see the most human error are the places Waymos don’t operate - interstates, highways and older narrow neighborhoods.
Maybe there's a way to tell Waymo that they keep using an illegal no stopping zone?
I believe they are waiting for passengers - they usually have some kind of LED display with what looks like initials on the topmost sensor when they're doing this.
>> The 6th-generation Waymo Driver is the product of seven years of safety-proven service amassed from driving nearly 200 million fully autonomous miles across the densest cores of 10+ major cities and an expanding network of freeways. Our experience as the only company operating a fully autonomous service at this scale has reinforced a fundamental truth: demonstrably safe AI requires equally resilient inputs. This deep understanding of real-world requirements is why the Waymo Driver utilizes a custom, multi-modal sensing suite where high-resolution cameras, advanced imaging radar, and lidar work as a unified system. Using these diverse inputs, the Waymo Driver can confidently navigate the "long tail" of one-in-a-million events we regularly encounter when driving millions of miles a week, leaving nothing to the imagination of a single lens.
Waymo uses remote safety drivers that they call "fleet response agents", probably to deflect from the fact that they are, indeed, remote safety drivers.
Much like phone-a-friend, when the Waymo vehicle encounters a particular situation on the road, the autonomous driver can reach out to a human fleet response agent for additional information to contextualize its environment. The Waymo Driver does not rely solely on the inputs it receives from the fleet response agent and it is in control of the vehicle at all times. As the Waymo Driver waits for input from fleet response, and even after receiving it, the Waymo Driver continues using available information to inform its decisions. This is important because, given the dynamic conditions on the road, the environment around the car can change, which either remedies the situation or influences how the Waymo Driver should proceed. In fact, the vast majority of such situations are resolved, without assistance, by the Waymo Driver.
In the most ambiguous situations, the Waymo Driver takes the lead, initiating requests through fleet response to optimize the driving path. Fleet response can influence the Waymo Driver's path, whether indirectly through indicating lane closures, explicitly requesting the AV use a particular lane, or, in the most complex scenarios, explicitly proposing a path for the vehicle to consider. The Waymo Driver evaluates the input from fleet response and independently remains in control of driving. This collaboration enhances the rider experience by efficiently guiding them to their destinations.
From: https://waymo.com/blog/2024/05/fleet-response/
Note the language: the Waymo Driver "remains in control of driving" but a Fleet Response Agent "proposes" the path.
In other words, Waymo is not "operating a fully autonomous service", nor does it seem anything has changed now, with the "sixt-generation fully autonomous Waymo Driver". It still needs human brains to take it by the hand and help it when it gets stuck in ambiguous situations that arise despite the claim that it "can confidently navigate the "long tail" of one-in-a-million events we regularly encounter when driving millions of miles a week".
Dunno, you're engaging in the same "quibbling about semantics" and "arbitrary definitions", when for all intents and purposes, Waymo is fully autonomous. Is there any meaningful difference between 99% autonomous and 100% autonomous? If we're sending a space probe to alpha centauri, then the last 1% would be very important, because help is light-years away, but the same isn't true for autonomous cars. What's more misleading, claiming that 99% autonomy is "fully autonomous", or claiming that waymo "uses safety drivers" while omitting the fact that they're rarely used, in contrast to other vendors (eg. uber, when they were doing tests) where they also have "safety drivers" monitoring the car all the time?
To clarify, how Waymo's remote safety drivers operate is that they are called to make decisions that the Waymo Driver can't make, when it gets in situations that are not covered by its training. Here's the relevant text from further down the Waymo blog post:
In the most ambiguous situations, the Waymo Driver takes the lead, initiating requests through fleet response to optimize the driving path. Fleet response can influence the Waymo Driver's path, whether indirectly through indicating lane closures, explicitly requesting the AV use a particular lane, or, in the most complex scenarios, explicitly proposing a path for the vehicle to consider. The Waymo Driver evaluates the input from fleet response and independently remains in control of driving. This collaboration enhances the rider experience by efficiently guiding them to their destinations.
Note the wording "The Waymo Driver ... remains in control of driving". That has a very precise meaning.
In plain English, the Waymo Driver asks a human to choose a destination for it and then it autonomously follows a path to that destination. That's autonomous path planning, but not autonomous navigation. Path planning is common as sparrows [1] and it's certainly not enough to qualify a system as "fully autonomous", despite Waymo's marketing copy to the contrary.
If Waymo Driver was "fully autonomous" even to the "99%" extent you say I'm quibbling about (I'm not) then it would be capable of autonomous navigation without any support from humans. If it didn't manage to do it right all the time we'd say it has some error, not that it's not fully autonomous. Like when a blender fails to blend my bastard sword, I won't say the blander can't blend, I'll just say it can't blend my bastard sword.
Btw I deliberately left the first sentence in the quote above where Waymo says "the Waymo Driver takes the lead" just in case leaving it out created the impression I'm trying to downplay the system's capabilities. It's rather the other way around. "Takes the lead" is a completely useless anthropomorphizing of an automated system made in order to create an impression of capabilities that the system doesn't have. Most likely the procedure that "takes the lead" is a hand-coded routine that calls home when some set of variables crosses a threshold. More like a tripwire than "takes the lead".
The Waymo Driver is not "fully autonomous". It's an impressively well-functioning and safe system, to the extent we can trust the scant data shared by Waymo, which we can't totally because that data is itself only shared for the purpose of marketing, but nobody needs to accept Waymo's press releases as reality.
__________________
[1] Really. There's dozens upon dozens of different approaches. I'm on my other computer and I don't have access to my papers-to-read folder but there's just one survey I often quote that lists three different categories each with many sub-categories. It's not so much an open problem as one where it's difficult to choose a solution, just because there are so many around. One way or another, no system is "fully autonomous" just because it can autonomously follow a path between points A and B, while avoiding obstacles. I even worked on one myself.
They appear to be distinguished externally by the vehicle model and new sensor design. Currently, the production fleet in Phoenix and elsewhere consists of the Jaguar I-PACE.
The sixth generation would have been in testing phases -- closed tracks, simulations, and supervised driving. Now they're deploying on the Waymo Ojai (Zeekr) and Hyundai's IONIQ 5.
This is way more fun than I ever had in my Corolla.
From the company who got the world "go-goo"ing like infants, I, for one, can't wait to say "O HAI" to my new ride, or "Isn't it IONIQ, don't you think?"
Unlike human driven cars, 100% of Waymos are fully filmed. The proof is on every drive. And there is one entity responsible for all of them, Waymo.
I've enjoyed my trips so far, but want them to stop breaking the law.
When the general public does.
Autonomous cars that abide by the law at the expense of violating the norms of expected traffic behavior like a 16yo in a driver's ed car (which is plastered in signs for exactly that reason) are not a scalable way of sharing roads with the general public.
As an aside, the venn diagram between people who complain about normal traffic behavior being unlawful and people who resist tweaking the law to make what is normal also lawful is far too close to a circle for my taste.
https://youtu.be/040ejWnFkj0?si=-iffWU43sxwviD5t
[EDIT] Most of you seem unwilling to spend an hour to watch a youtube video (although I believe it's worth your time esp if you're from North America) so here's a summary I attempted in another comment:
"Autonomous cars will clog up existing cities by cruisnig around looking to pick up rides or deliver shit and mill around endlessly or occupy every piece of parking in prime real estate to make sure they are quickly available wherever demand is high (i.e. where people want to or have to be). In time they will phase out human driven cars which will lead to higher speed limits and more infrastrcuture that supports autonomous driving. Meaning fewer "difficult" intersections, straighter roads, no bike lanes or pedestrian sidewalks. Everything optimized for autonomous cars to endlessly mill around. People will be blocked from being near autonomous cars as those will be going too fast for human reflexes to cope with so areas where cars drive will not have sidewalkss nor bike lanes. This will lead to urban areas that are even more car dependent with only pockets of urbanism that support human scale. To get anywhere one will need to hail one of those autonomous taxis and then zoom in it to a destination where it's again safe to walk in whatever pocket of human activity. Since cars need a lot more land area than humans the urban infrastructure will mostly cater to them and not to people because the expectation and argument will be that you can always get your ass shuttled to wherever you need to be."
I can appreciate this technology might negatively impact other countries more heavily, but, for me, it's easily the most exciting tech I interact with and I'm rooting for it whole-heartedly. I'm at around 1000 miles logged on Waymo and am part of their beta tester program for freeway usage.
I also think that post-Covid remote work has probably damaged incentives for increasing the density of cities more so than anything autonomous vehicles will do. San Francisco is actively cutting bus routes, bus density, and threatening to significantly cut BART stops due to budget constraints and reduction in ridership.
It's odd because I do get where you're coming from, and I feel like I should be your target audience, but, for me, the ship sailed so long ago that I struggle to relate to your position.
Now of course sometimes I’m not content staying within this 15-minute circle. Then I simply choose the fastest method of transport to get there. Is BART or Muni faster than the Waymo trip? Then yes I’ll take pubic transportation. That’s what good transit is for.
> "Autonomous cars will clog up existing cities..."
Congestion charges. Limited licensing for TNCs. Dedicated public or private holding areas rather than "milling about". All of these have solutions.
> Meaning fewer "difficult" intersections, straighter roads, no bike lanes or pedestrian sidewalks.
It is already best practice in urban design to separate cars that need to quickly transit an area without interacting with it into completely independent routes where there are no bikes or pedestrians, and combine transit/bikes/walking into livable mixed mode streets where cars are not allowed. NotJustBikes has many examples of this, most commonly around Europe.
> To get anywhere one will need to hail one of those autonomous taxis and then zoom in it to a destination where it's again safe to walk in whatever pocket of human activity.
This is what already happens in places that don't have usable, safe, or car-competitive transit, modulo autonomous, including currently most of North America. The solution to needing fewer cars -- self driving or not -- is investment in transit and in ground-up overhaul of existing cities to optimize for transit and deprioritization of cars.
I had tuned in to some channels for analysis and insightful commentary, for example, film and TV series.
But every one devolved into “Worst episode ever!” and “<studio> has RUINED <franchise>!”
So to sum up, the YouTube recommendations algorithm has ruined independent criticism and there is nothing on anymore. Join my Patreon, “UnJustLikes” for the deep dive!
Just say what you want to say ... with words!
Autonomous cars will clog up existing cities by cruisnig around looking to pick up rides or deliver shit and mill around endlessly or occupy every piece of parking in prime real estate to make sure they are quickly available wherever demand is high (i.e. where people want to or have to be). In time they will phase out human driven cars which will lead to higher speed limits and more infrastrcuture that supports autonomous driving. Meaning fewer "difficult" intersections, straighter roads, no bike lanes or pedestrian sidewalks. Everything optimized for autonomous cars to endlessly mill around. People will be blocked from being near autonomous cars as those will be going too fast for human reflexes to cope with so areas where cars drive will not have sidewalkss nor bike lanes. This will lead to urban areas that are even more car dependent with only pockets of urbanism that support human scale. To get anywhere one will need to hail one of those autonomous taxis and then zoom in it to a destination where it's again safe to walk in whatever pocket of human activity. Since cars need a lot more land area than humans the urban infrastructure will mostly cater to them and not to people because the expectation and argument will be that you can always get your ass shuttled to wherever you need to be.
I regularly bike, which is why I'm hugely in favor of self driving cars; they're way safer for me when biking than human driven cars.
In time, human driving will be phased out and that will precipitate removal of speed limits and traffic lights as autonomous cars will be able to use vehicle to vehicle messaging to negotiate intersections. Of course pesky pedestrians and cyclists could still be in the way. That's where lobbying comes in to restrict the pedestrian areas to pockets where cars and people never share the same space. But since cars require much more space than peeople the result will be more sprawl and less walkable places as it will be people who will get pushed aside.
All the incentives you described exists today. On any given road any space devoted to sidewalks or bike lanes means less space for cars, and you already need separation between car lanes and sidewalks. You also have the same incentive to "restrict the pedestrian areas to pockets where cars and people never share the same space", because any controlled access roadway increases speed and throughput. Finally if you restrict pedestrians to certain areas (we all live in megatowers?), that actually makes taxis (including robotaxis) less attractive relative to public transit, because their whole value proposition is that they take you exactly where you want to go. Therefore it's unclear how automated cars would make things worse.
But - I'm just not sure your analysis is right. Someone who drives a self-owned car will park it in a downtown area for hours. Someone who takes a taxi of any sort will use much, much less amortized parking spot space. New York is a pretty good example of this.
Good public transit beats the snot out of cars, but a dense taxi deployment seems to get more people moved per total car-dedicated-space than private car drivership does.
And if we can reduce the amount of space dedicated to parking, we can increase density, which reduces the need for driving.
So the problem will be if we have self-driving whatevers at the expense of public transit, but perhaps not if it's at the expense of private car drivership.
This implies that there is no moat. One company or another might be the first on the market with one system, but the others will catch up in the space of a couple of years. The first to market won't have much advantage over the others. We'll see a replica of what happened with LLMs: any latecomer will be able to replicate the results by putting a few billions on the table and hiring researchers from other companies, Chinese companies will develop a working version that runs on slightly less demanding hardware, open weights and open source will appear. Etc.
Eventually, it will become more of a commodity-level task, but by then most of the big incumbents will already be very established.
From: "Using these diverse inputs, the Waymo Driver can confidently navigate the "long tail" of one-in-a-million events we regularly encounter when driving millions of miles a week, leaving nothing to the imagination of a single lens."