And those complex traffic situations are the main challenge for autonomous driving. Getting the AIs to do the right things before they get themselves into trouble is key.
Lidar is not a silver bullet. It helps a little bit, but not a whole lot. It's great when the car has to respond quickly to get it out of a situation that it shouldn't have been in to begin with. Avoiding that requires seeing and understanding and planning accordingly.
You can train a DL model to act like a LiDAR based on only camera inputs (the data collection is easy if you already have LiDAR cars driving around). If they could get this to work reliably, I'm sure the competition would do it and ditch the LiDAR, but they don't, so that tells us something.
For anyone who understands sensor fusion and the Kalman filter, read this and ask yourself if you trust Elon Musk to direct the sensor strategy on your autonomous vehicle: https://www.threads.com/@mdsnprks/post/DN_FhFikyUE
For anyone wondering, to a sensors engineer the above tweet is like sayin 1 + 1 = 0 -- the truth (and science) is the exact opposite of what he's saying.
edit: no, it was ultrasonic sensors. But this was likely object detection, and now it's gone.
If you look at the statistics on fatal car accidents, 85%+ involve collisions with stationary objects or other road users.
Nobody's suggesting getting rid of machine vision or ML - just that if you've got an ML+vision system that gets in 1 serious accident per 200,000 miles, adding LIDAR could improve that to 1 serious accident per 2,000,000 miles.
It’s already better at X-rays and radiology in many cases.
Everything you are talking about is just a matter of sufficient learning data and training.
The most important thing is that Tesla/Elon absolutely had no way to know, and no reason to believe (other than as a way to rationalise a dangerously risky bet) that machine vision would be able to solve all these issues in time to make good on their promise.
> this tells me volumes about what's going on in the computer vision system
Emphasis:
> computer vision system
It wasn't Elon's but Karpathy's.
The position against lidar was that it traps you in a local max, that humans use vision, that roads and signs are designed for vision so you're going to ultimately have to solve that problem and when you do lidar becomes a redundant waste. The investment in lidar wastes time from training vision and may make it harder to do so. That's still the case.
I love Waymo, but it's doomed to be localized to populated areas with high-res mapping - that's a great business, but it doesn't solve the general problem.
If Tesla keeps jumping on the vision lever and solves it they'll win it all. There's nothing in physics that makes that impossible so I think they'll pull it off. His model is all this sort of first principles thinking, it's why his companies pull off things like starship. I wouldn't bet against it.
Elon is being foolish and weirdly anthropomorphic.
LLMs have shown the general public how AI can be plain wrong and shouldn't be trusted for everything. Maybe this influences how they, and regulators, will think about self driving cars.
And the general public?! No way. Most are completely unaware of the foibles of LLMs.
No the don't. You're making a straw man rather than trying to put forth an actual argument in support of your view.
If you feel can't support your point, then don't try to make it.
I responded to this parent comment:
"LLMs have shown the general public how AI can be plain wrong and shouldn't be trusted for everything."
You take issue with my response of:
"loads of DEVs on here will claim LLMs are infallible"
You're not really making sense. I'm not straw-manning anything, as I'm directly discussing the statement made. What exactly are you presuming I'm throwing a straw man over?
It's entirely valid to say "there are loads of supposed experts that don't see this point, and you're expecting the general public to?". That's clearly my statement.
You may disagree, but that doesn't make it a strawman. Nor does it make it a poorly phrased argument on my part.
Do pay better attention please. And your entire last sentence is way over the line. We're not on reddit.
The irony of telling someone not to be rude while being absolutely insufferable. Peak redditor behavior.
Please provide examples. Thank you!
All of that combined is probably closer to $1k than to $140.
And, again, that's - what - 10 years after Tesla originally made the decision to go vision only.
It wasn't a terrible idea at the time, but they should've pivoted at some point.
They could've had a massive lead in data if they pivoted as late as 3 years ago, when the total cost would probably be under $2.5k, and that could've led to a positive feedback loop, cause they'd probably have a system better than Waymo by now.
Instead, they've got a pile of garbage, and no path to improve it substantially.
They might be!
But I doubt it.
I don't know enough about Tesla's cameras, but it's not implausible to think there are LIDARs of low enough quality that you'd be better off with a good quality camera for your sensor.
Again, I doubt this is the case with BYDs cameras.
But it's still worth pointing out, I think.
My point is, BYD's LIDAR system costing $x is only one small part of the conversation.
>The EX90's LiDAR enhances ADAS features like collision mitigation and lane-keeping, which are active and assisting drivers. However, full autonomy (Level 3) is not yet available, as the Ride Pilot feature is still under development and not activated.
Solid-state LIDAR is still a fairly new thing. LIDAR sensors were big, clunky, and expensive back when Tesla started their Autopilot/FSD program.
I googled a bit and found a DFR1030 solid-state LIDAR unit for 267 DKK (for one). It has a field of view of 108 degrees and an angular resolution of 0.6 degrees. It has an angle error of 3 degrees and a max distance of 300mm. It can run at 7.5-28 Hz.
Clearly fine for a floor-cleaning robot or a toy. Clearly not good enough for a car (which would need several of them).
The cold hard truth is that LIDARs are a crutch, they're not strictly necessary. We know this because humans can drive without a LIDAR, however they are a super useful crutch. They give you super high positional accuracy (something that's not always easy to estimate in a vision-only system). Radars are also a super useful crutch because they give really good radial velocity. (Little anecdote, when we finally got the Radars working properly at work it made a massive difference to the ability for our car to follow other cars, ACC, in a comfortable way).
Yes machine learning vision systems hallucinate, but so do humans. The trick for Tesla would be to get it good enough to where it hallucinates less than humans do (they're nowhere near yet - human's don't hallucinate very often).
It's also worth adding that last I checked the state of the art for object detection is early fusion where you chuck the LIDAR and Radar point clouds into a neural net with the camera input so it's not like you'd necessarily have the classical methods guardrails with the Lidar anyway.
Anyway, I don't think Tesla were wrong to not use LIDAR - they had good reasons to not go down that route. They were excessively expensive and the old style spinning LIDARs were not robust. You could not have sold them on a production car in 2018. Vision systems were improving a lot back then so the idea you could have a FSD on vision alone was plausible.
The hard truth is there is no reason to limit machines to only the tools humans are biologically born with. Cars always have crutches that humans don't possess. For example, wheels.
Turns out, when there's demand for LIDAR in this form factor, people invest in R&D to drive costs down and set up manufacturing facilities to achieve economies of scale. Wow, who could have predicted this‽
Western countries might not be smart enough to keep R&D because Wall Street sees it as a cost center.
[1] https://www.tomshardware.com/tech-industry/artificial-intell...
You know what else used to be expensive? Structured light sensors. They cost $$$$ in 2009. Then Microsoft started manufacturing the Kinect for a mass market, and in 2010 price went down to $150.
You know what's happened to LIDAR in the past decade? You guessed it, costs have come massively down because car manufacturers started buying more, and costs will continue to come down as they reach mass market adoption.
The prohibitive cost for LIDAR coming down was always just a matter of time. A "visionary" like Musk should have been able to see that. Instead he thought he could outsmart everyone by using a technology that was not suited for the job, but he made the wrong bet.
This should be expected when someone who is *not* an experienced engineer starts making engineering decisions.
FOSS is the obvious counterexample to your absurdly firm stance, but so are many artistic pursuits that use engineering techniques and principles, etc.
My intent was to exclude research efforts, which is fundamentally different from engineering, which is a practical concern and not a “get it to just work” concern.
In a true self-driving utopia, all of the cars are using multiple methods to observe the road and drive (vision, lidar, GPS, etc) AND they are all communicating with each other silently, constantly, about their intentions and status.
Why limit cars to what humans can do?
The reason this is clear is because, except for a brief period in late 2022, Teslas have included some combination of radar and ultrasonic sensors. [0]
[0] https://en.m.wikipedia.org/wiki/Tesla_Autopilot_hardware
A system that's only based on cameras is only as good as its ability to recognize all road hazards, with no fall back if that fails. With LIDAR, the vehicle might not know what's the solid object in front of the vehicle using cameras, but it knows that it's there and should avoid running into it.
This is a good example of why sensor fusion is good.
Crutch for what? AI does not have human intelligence yet and let’s stop pretending it does. There is no shame in that as the word crutch implies.
If a sensor provides additional data, why not use it? Sure, humans can drive withot lidars, but why limit the AI to using human-like sensors?
Why even call it a crutch? IMO It's an advantage over human sensors.
Quality of additional data matters. How often does a particular sensor give you false positives and false negatives? What do you do when sensor A contradicts sensor B?
“3.6 roentgen, not great, not terrible.”
Probably comes down to lidar (and Ai) failure modes.
Also don’t forget that as a human you can move your head any which way, and also draw on your past experiences driving in that area. “There is always an old man crossing the road at this intersection. There is a school nearby so there might be kids here at 3pm.” That stuff is not as accessible to a LIDAR.
That's because our stereoscopic vision has infinitely more dynamic range, focusing speed and processing power w.r.t. a computer vision system. Periphery vision is very good at detecting movement, and central view can process tremendous amount of visual data without even trying.
Even a state of the art professional action camera system can't rival our eyes in any of these categories. LIDARs and RADARs are useful and shall be present in any car.
This is the top reason I'm not considering a Tesla. Brain dead insistence on cameras with small sensors only.
You’re not considering them even though they have the best adas on the market lmao suit yourself
(and that's not even addressing that human vision is fundamentally a weird sensory mess full of strange evolutionary baggage that doesn't even make sense except for genetic legacy)
The assumption was that with similar sensors (or practically worse - digital cameras score worse than eyeballs in many concrete metrics), ‘AI’ could be dramatically better than humans.
At least with Tesla’s experience (and with some fudging based on things like actual fatal accident data) it isn’t clear that is actually what is possible. In fact, the systems seem to be prone to similar types of issues that human drivers are in many situations - and are incredibly, repeatedly, dumb in some situations many humans aren’t.
Waymo has gone full LiDAR/RADAR/Visual, and has had a much better track record. But their systems cost so much (or at least used to), that it isn’t clear the ‘replace every driver’ vision would ever make sense.
And that is before the downward pressure on the labor market started to happen post-COVID, which hurts the economics even more.
The current niche of Taxis kinda makes sense - centrally maintained and capitalized Taxis with outsourced labor has been a viable model for a long time, it lets them control/restrict the operating environment (important to avoid those bad edge cases!), and lets them continue to gather more and more data to identify and address the statistical outliers.
They are still targeting areas with good climates and relatively sane driving environments because even with all their models and sensors, heavy snow/rain, icy roads, etc. are still a real problem.
When the argument was Phoenix is too pleasant I could buy that. Most places aren't Phoenix. But SF and LA are both much more like a reasonable place other humans live. It rains, but not always, it's misty, but not always. Snow I do accept as a thing, lots of places humans live have some snow, these cities don't really have snow.
However for ice when I watch one of those "ha, most drivers can make this turn in the ice" videos I'm not thinking "I bet Waymo wouldn't be able to do this" I'm thinking "That's a terrible idea, nobody should be attempting it". There's a big difference between "Can it drive on a road with some laying snow?" and "Can it drive on ice?".
Both SF and LA climates are super cushy compared to say, Northern Michigan. Or most of the eastern seaboard. Or even Kansas, Wyoming, etc. in the winter.
In those climates, if you don’t drive in what you’re calling ‘nobody should be attempting it’ weather, you - starve to death in your house over the winter. Because many months are just like that.
Self driving has a very similar issue with the vast majority of, say, Asia. Because similarly “this is crazy, no one should be driving in these conditions” is the norm. So if it can’t keep up, it’s useless.
Eastern and far Northern Europe has a lot of kinda similar stuff going on.
Self driving cars are easy if you ignore the real world haha.
I live in the middle of a city, so, no, in terrible weather just like great weather I walk to the store, no need to "starve to death" even if conditions are too treacherous for people to sensibly drive cars. Because I'm an old man, and I used to live somewhere far from a city, I have had situations where you can't use a car to go fetch groceries because even if you don't care about safety the car can't go up an icy hill, it loses traction, gravity takes over, you slide back down (and maybe wreck the car).
Because as an old man who has actually lived in all these places - and also has ridden in Waymos before and has had friends on the Waymo team in the past, your comments seem pretty ridiculous.
A lot of the large population centres in the US are in these what you're calling "super cushy" zones where there's not much snow let alone ice. More launches in cities in Florida, Texas, California will address millions more people but won't mean more ice AFAIK. So I guess for you the most interesting announcement is probably New York, since New York certainly does have real snow. 2026 isn't that long, although I can imagine that maybe a President who thinks he's entitled to choose the Mayor of New York could mess that up.
As to the "But people in some places are crazy drivers" I saw that objection from San Francisco before it was announced. "Oh they'll never try here, nobody here drives properly. Can you imagine a Waymo trying to move anywhere in the Mission?". So I don't have much time for that.
and propellers on a plane are not strictly necessary because birds can fly without them? The history of machines show that while nature can sometimes inspire the _what_ of the machine, it is a very bad source of inspiration for the _how_.
When was the last time you had full attention on the road and a reflection of light made you super confused and suddenly drive crazy? When was the last time you experienced objects behaving erratically around you, jumping in and out of place, and perhaps morphing?
We were somewhere around Barstow on the edge of the desert when the drugs began to take hold. I remember saying something like, “I feel a bit lightheaded; maybe you should drive . . .”And suddenly there was a terrible roar all around us and the sky was full of what looked like huge bats, all swooping and screeching and diving around the car, which was going about 100 miles an hour with the top down to Las Vegas. And a voice was screaming: “Holy Jesus! What are these goddamn animals?” [0]
When was the last time you saw a paper bag blown across the street and mistook it for a cat or a fox? (Did you even notice your mistake, or do you still think it was an animal?)
Do you naturally drive faster on wide streets, slower on narrow streets, because the distance to the side of the road changes your subconcious feeling of how fast you're going? Do you even know, or are you limited to your memories rather than a dashcam whose footage can be reviewed later?
etc.
Now don't get me wrong, AI today is, I think, worse than humans at safe driving; but I'm not sure how much of that is that AI is more hallucinate-y than us vs. how much of it is that human vision system failures are a thing we compensate for (or even actively make use of) in the design of our roads, and the AI just makes different mistakes.
Self-driving is probably “AI-hard” as you’d need extensive “world knowledge” and be able to reason about your environment and tolerate faulty sensors (the human eyes are super crappy with all kinds of things that obscure it, such as veins and floaters).
Also, if the Waymo UI accurately represents what it thinks is going on “out there” it is surprisingly crappy. If your conscious experience was like that when you were driving you’d think you had been drugged.
The human brain's vision system makes pretty much the exact opposite mistake, which is a fun trick that is often exploited by stage magicians: https://www.youtube.com/watch?v=v3iPrBrGSJM&pp
And is also emphasised by driving safety awareness videos: https://www.youtube.com/watch?v=LRFMuGBP15U
I wonder what we'd seem like to each other, if we could look at each other's perception as directly as we can look at an AI's perception?
Most of us don't realise how much we mispercieve because it doesn't feel different in the moment to percieve incorrectly; it can't feel different in the moment, because if it did, we'd notice we were mispercieving.
Robots are supposed to make up for our limitations by doing things we can't do, not do the things we can already do, but differently. The latter only serves to replace humans, not augment them.
So maybe LIDAR isn't necessary but also if Tesla were actually investing in cameras with a memory bus that could approximate the speed of human vision I doubt it would be cheaper than LIDAR to get the same result.
One of the ways it's better is that humans can sense individual photons. Not 100% reliably, but pretty well, which is why humans can see faint stars on a dark night without any special tools even though the star is thousands of light years away. On the other hand, our resolution for most of our field of vision is pretty bad - this is compensated for by changing what we're looking it when we care about details we can just look directly at it and the resolution is better right in the centre of the picture.
This was only plausible to people who had no experience in robotics, autonomy, and vision systems.
Everyone knew LIDAR was the enabling technology thanks to the 2007 DARPA Urban challenge.
But the ignoramus Elon Musk decided he knew better and spent the last decade+ trashing the robotics industry. He set us back as far as safety protocols in research and development, caused the first death due to robotic cars, deployed them on public roads without the consent of the public by hoisting around his massive wealth, lied consistently for a DECADE about the capabilities of these machines, defrauded customers and shareholders while becoming richer and richer, all to finally admit defeat while he still maintains the growth story of for Tesla's future remains in robotics. The nerve of this fucking guy.
I agree that Tesla may have made the right hardware decision when they started with this. It was probably a bad idea to lock themselves into that path by over-promising.
The correct move for Tesla would have been to split the difference and add LIDAR to some subset of their fleet, ideally targeted in the most difficult to debug environments.
Somewhat like Google/Waymo are doing with their Jaguars.
Don't LIDAR 100% of Teslas, but add it to >0%.
Reportedly, they no longer use this widely - but they still have some LIDAR-equipped "scout vehicles" they send into certain environments to collect extra data.
Tesla would subsidize them and offer them at the same price as non-LIDAR models, to select customers in target areas.
And yes, you answered the second part of your own question.
In any case, thanks, TIL!
But Tesla didn't do this.
There is none with Musk's "vision only" approach. Vision can fail for a multitude of reasons --- sunlight, rain, darkness, bad road markers, even glare from a dirty windshield. And when it fails, there is no backup plan -- the car is effectively driving blind.
Driving is a dynamic activity that involves a lot more than just vision. Safe automated driving can use all the help it can get.
Both LIDAR and vision have edge cases where they fail. So you ideally want both, but then the challenge is reconciling disagreements with calibrated, and probabilistic fusion. People seem to be under the mistaken impression that vision is dirty input and LIDAR is somehow clean, when in reality both are noisy inputs with different strengths and weaknesses.
I guess my point is: Yes, 100% bring in LIDAR, I believe the future is LIDAR + vision. But when you do that, early iterations can regress significantly from vision-only until the fusion is tuned and calibration is tight, because you have to resolve contradictory data. Ultimately the payoff is higher robustness in exchange for more R&D and development workload (i.e. more cost).
The same reason why Tesla needed vision-only to work (cost & timeline) is the same reason why vision+LIDAR is so challenging.
It's the ability to detect sensor disagreements at all.
With single modality sensors, you have no way of truly detecting failures in that modality, other than hacks like time-series normalizing (aka expected scenarios).
If multiple sensor modalities disagree, even without sensor fusion, you can at least assume something might be awry and drop into a maximum safety operation mode.
But we'd think that the budget config of the Boeing 737 MAX would have taught us that tying safety critical systems to single sources of truth is a bad idea... (in that case, critical modality / single physical sensor)
Also, this is probably when Waymo calls up a human assistant in a developing-country callcentre.
"A man with a watch always knows what time it is. If he gains another, he is never sure"
Most safety critical systems actually need at least three redundant sensors. Two is kinda useless: if they disagree, which is right?
EDIT:
> If multiple sensor modalities disagree, even without sensor fusion, you can at least assume something might be awry and drop into a maximum safety operation mode.
This is not always possible. You're on a two lane road. Your vision system tells you there's a pedestrian in your lane. Your LIDAR says the pedestrian is actually in the other lane. There's enough time for a lane change, but not to stop.
What do you do?
> This is not always possible. You're on a two lane road. Your vision system tells you there's a pedestrian in your lane. Your LIDAR says the pedestrian is actually in the other lane. There's enough time for a lane change, but not to stop.
> What do you do?
Go into your failure mode. At least you have a check to indicate a possible issue with 2 signals.
People are underweighting the alternative single system hypothetical -- what does a Tesla do when its vision-only system erroneously thinks a pedestrian is one lane over?
The issue is not recognising that optimising for Ux at the expense of safety here is the wrong call, motivated likely by optimism and a desire for autonomous cars, more than reasonable system design. I.e. if the sensors disagree so often that it makes the system unusable, maybe the solution is “we’re not ready for this kind of technology and we should slow down” rather than “let’s figure out non-UX breaking edge case heuristics to maintain the illusion of autonomous driving being behind the corner”.
Part of this problem is not even technological - human drivers tradeoff safety for UX all the time - so the expectation for self driving is unrealistic and your system has to have the ethically unacceptable system configuration in order to have any chance of competing.
Which is why - in my mind - it’s a fools endeavour in personal car space, but not in public transport space. So go waymo, boo tesla.
They don't work by merely taking a straw poll. They effectively build the joint probability distribution, which improves accuracy with any number of sensors, including two.
> You're on a two lane road. Your vision system tells you there's a pedestrian in your lane. Your LIDAR says the pedestrian is actually in the other lane. There's enough time for a lane change, but not to stop.
Any realistic system would see them long before your eyes do. If you are so worried, override the AI in the moment.
Lots of safety critical systems actually do operate by "voting". The space shuttle control computers are one famous example [1], but there are plenty of others in aerospace. I have personally worked on a few such systems.
It's the simplest thing that can obviously work. Simplicity is a virtue when safety is involved.
You can of course do sensor fusion and other more complicated things, but the core problem I outlined remains.
> If you are so worried, override the AI in the moment.
This is sneakily inserting a third set of sensors (your own). It can be a valid solution to the problem, but Waymo famously does not have a steering wheel you can just hop behind.
This might seem like an edge case, but edge cases matter when failure might kill somebody.
1. https://space.stackexchange.com/questions/9827/if-the-space-...
I see in many domains a tendency to oversimplify decision making algorithms for human understanding convenience (eg vote rather that develop a joint probability distribution in this case, supply chain and manufacturing in particular seem to love rules of thumb) rather than use better algorithms that modern compute enables higher performance, safety etc
I will not pretend to be an expert. I would suggest that "human understanding convenience" is pretty important in safety domains. The famous Brian Kernighan quote comes to mind:
> Everyone knows that debugging is twice as hard as writing a program in the first place. So if you're as clever as you can be when you write it, how will you ever debug it?
When it comes to obscure corner cases, it seems to me that simpler is better. But Waymo does seem to have chosen a different path! They employ a lot of smart folk, and appear to be the state of the art for autonomous driving. I wouldn't bet against them.
No, when it comes to not killing people, I'd say that safer is usually better.
Remember the core function of the system is safety, simplicity is nice to have, but explicitly not as important.
That said, beware of calling something 'complicated' just because you don't understand it, especially if you don't have training and experience in that thing. What's more relevant is whether the people building the systems think it is too complicated.
This is completely different from systems that cover different domains, like vision and lidar.
Cars can stop in quite a short distance. The only way this could happen is if the pedestrian was obscured behind an object until the car was dangerously close. A safe system will recognize potential hiding spots and slow down preemptively - good human drivers do this.
"Quite a short distance" is doing a lot of lifting. It's been a while since I've been to driver's school, but I remember them making a point of how long it could take to stop, and how your senses could trick you to the contrary. Especially at highway speeds.
I can personally recall a couple (fortunately low stakes) situations where I had to change lanes to avoid an obstacle that I was pretty certain I would hit if I had to stop.
While it's true they don't stop instantaneously at highway speeds, cars shouldn't be driving highway speeds when a pedestrian suddenly being in front of you is a realistic risk.
I don't think these problems can just be assumed away.
This is why good redundant systems have at least 3... in your scenario, without a tie-breaker, all you can do is guess at random which one to trust.
For example jet aircraft commonly have three pitot static tubes, and you can just compare/contrast the data to look for the outlier. It works, and it works well.
If you tried to do that with e.g. LIDAR, vision, and radar with no common point of reference, solving for trust/resolving disagreements is an incredibly difficult technical challenge. Other variations (e.g. two vision + one LIDAR), does not really make it much easier either.
Tie-breaking during sensor fusion is a billion+ dollar problem, and will always be.
For example, say you have a pedestrian that's partially obscured by a car or another object, and maybe they're wearing a hat or a mask or wearing a backpack or carrying a kid or something, it may look unusual enough that either the camera or the lidar isn't going to recognize it as a person reliably. However, since the camera is generally looking at color, texture, etc in 2D, and the Lidar is looking at 3D shapes, they'll tend to fail in different situations. If the car thinks there's a substantial probability of a human in the driving path, it's going to swerve or hit the brakes.
But vision only hasn't worked --- not as promised, not after a decade's worth of timeline. And it probably won't any time soon either --- for valid engineering reasons.
Engineering 101 --- *needing* something to work doesn't make it possible or practical.
It was maybe a valid argument 10 years ago, but in 2025 many companies have shown sensor fusion works just fine. I mean, Waymo has clocked 100M+ miles, so it works. The AV industry has moved on to more interesting problems, while Tesla and Musk are still stuck in the past arguing about sensor choices.
The old ambition is dead.
[1] https://electrek.co/2025/05/16/tesla-robotaxi-fleet-powered-...
Infra-red of a few different wavelengths as well as optical light ranges seems like it'd give a superior result?
But we're far from plateauing on what can be done with vision - Humans can drive quite well with essentially just sight, so we're far from extinguishing what can be done with it.
Notably, sensor confusion is also an “unsolved” problem in humans, eg vision and vestibular (inner ear) conflicts possibly explaining motion sickness/vertigo <https://www.nature.com/articles/s44172-025-00417-2>
The results of both tournaments: <https://carnewschina.com/2025/07/24/chinas-massive-adas-test...> Counterintuitively, vision scored best (Tesla Model X)
The videos are fascinating to watch (subtitles are available): Tournament 1 (36 cars, 6 Highway Scenarios): <https://www.youtube.com/watch?v=0xumyEf-WRI> Tournament 2 (26 cars, 9 Urban Scenarios): <https://www.youtube.com/watch?v=GcJnNbm-jUI>
Highway Scenarios: “tests...included other active vehicles nearby to increase complexity and realism”: <https://electrek.co/2025/07/26/a-chinese-real-world-self-dri...>
Urban Scenarios: “a massive, complex roundabout and another segment of road with a few unsignaled intersections and a long straight...The first four tests incorporated portions of this huge roundabout, which would be complex for human drivers, but in situations for which there is quite an obvious solution: don’t hit that car/pedestrian in front of you” <https://electrek.co/2025/07/29/another-huge-chinese-self-dri...>
I keep reading arguments like this, but I really don't understand what the problem here is supposed to be. Yes, in a rule based system, this is a challenge, but in an end-to-end neural network, another sensor is just another input, regardless of whether it's another camera, LIDAR, or a sensor measuring the adrenaline level of the driver.
If you have enough training data, the model training will converge to a reasonable set of weights for various scenarios. In fact, training data with a richer set of sensors would also allow you to determine whether some of the sensors do not in fact contribute meaningfully to overall performance.
It's really hard to accept cost as the reason when Tesla is preparing a trillion dollar package. I suppose that can be reconciled if the venture is a vehicle (ha!) to shovel money from investors and buyers into Elon's pockets, I imagine the prospect of being the worlds first trillionare is appealing.
It's really hard to accept cost as the reason when Tesla is preparing a trillion dollar package. I suppose that can be reconciled if one considers the venture to be a vehicle (ha!) to shovel as much money as possible from investors and buyers into Elon's pockets, I imagine the prospect of being the worlds first trillionare is appealing.
So like a human driver. Problem is, automatic drivers need to be substantially better than humans to be accepted.
(For those who don't want to click through "LIDAR is a fool's errand, and anyone relying on LIDAR is doomed."
I am much more curious about the next ten years. If we can bring down the cost of a LIDAR unit into parity with camera systems[1], I think I know the answer. But I thought that 10 years ago and it did not happen so I wonder what is the real roadblock to make LIDAR cheap.
[1] Which it won't replace, of course. What it will change is that it makes the LIDAR a regular component, not an exceptionally expensive component.
2) make it right
3) make it fast (or cheap in this case)
Elon thinks his genius intellect allows him to skip to #3.
Anything except the lowest end car will cost $20K or more, so $200 is one percent of that price.
It’s nothing.
True self-driving is still a baby that needs to grow and cannot even compete against an adult human with 30+ years of experience. As self driving actually forms to that level the market will grown.
A two to four ton vehicle that can accelerate like a Ferrari and go over 100 mph, fully self-driving, and 'a few hundred dollars is way too much'.
Disagree. Even as they are dialing back the claims, which may or may not affect how people use the vehicles. These things respond too quickly for flaky senses based on human sensoriums.
Once a product starts to sell after initial design, time is take to reduce the development cost. Try to reuse parts or replace part A with B. A machine from early 2018 can be little different than ones going out the door late 2018. _Kaizen_ was coined for this.
My point of view of when mass reduction in cost will be when self-driving is cost effect secondary feature on all Toyota vehicles. I see that as the litmus test for knowing that self-driving has reached true utility.
Also well designed vehicles would need a multi-sensor system to operate in self-driving mode. A human operating a car is using multi-sensor intake. Lack of multi-sensor in humans prevent them from operating a vehicle. Blind people need a secondary sensory input like walking stick. Vehicles need a multi-sensor system to prevent harming, mutilating, and killing the passengers and pedestrians.
Supervised FSD is already safer than a human.
The revisions and updates are safety tested on roads for months before they are released. Tesla also has models that are too big to run on existing production hardware that perform better than the release versions in test cars.
Updates are not git pulls and no engineer would ever think that they were.
Based on what I've read over the years: it costs too much for a consumer vehicle, it creates unwanted "bumps" in the vehicle visual design, and the great man said it wasn't needed.
Yes, those reasons are not for technology or safety. They are based on cost, marketing, and personality (of the CEO and fans of the brand).
He's doing the exact same thing and worse to people he doesn't like.
What you say is plausible, I haven't directly seen the evidence for it but I'm not inclined to completely doubt you
But I think back to the time where everything surrounding covid was 'misinformation' and Musk (even if the broken clock is right twice a day) genuinely gave people a place to speak. Old Twitter would shut down people, so fast (even experts)
https://www.aljazeera.com/economy/2024/8/13/the-right-wing-l...
You can find this stuff searching 2 minutes on Google.
He's just a power hungry oligarch.
https://opg.optica.org/oe/fulltext.cfm?uri=oe-31-2-2013&id=5...
I'm in a 2025 with HW4, but it's dramatic improvement over the last couple of years (previously had a 2018 Model 3) increased my confidence that Elon was right to focus on vision. It wasn't until late last year where I found myself using it more than not, now I use it almost every drive point to point (Cupertino to SF) and it does it.
I think people are generally sleeping on how good it is and the politicization means people are under valuing it for stupid reasons. I wouldn't consider a non Tesla because of this (unless it was a stick shift sports car, but that's for different reasons).
Their lead is so crazy far ahead it's weird to see this reality and then see the comments on hn that are so wrong. Though I guess it's been that way for years.
The position against lidar was that it traps you in a local max, that humans use vision, that roads and signs are designed for vision so you're going to have to solve that problem and when you do lidar becomes a redundant waste. The investment in lidar wastes time from training vision and may make it harder to do so. That's still the case. I love Waymo, but it's doomed to be localized to populated areas with high-res mapping - that's a great business, but it doesn't solve the general problem.
If Tesla keeps jumping on the vision lever and solves it they'll win it all. There's nothing in physics that makes that impossible so I think they'll pull it off.
I'd really encourage people to here with a bias to dismiss to ignore the comments and just go in real life to try it out for yourself.
This is not a general solution, it is an SF one... at best.
Most humans also don't get in accidents or have problems with phantom breaking within the timeframe that you mentioned.
The Bay Area has massive traffic, complex interchanges, SF has tight difficult roads with heavy fog. Sometimes there’s heavy rain on 280. 17 is also non trivial.
What Tesla has done is not trivial and roads outside the bay are often easier.
People can ignore this to serve their own petty cognitive bias, but others reading their comments should go look at it for themselves.
Here outside of Los Angeles, about an hour east, they do not do well at all on their 'auto-pilot.'
Your area has the benefit of being one of the primary training areas, and thus the dataset for your area is good.
Try that here. I'll be more than happy to watch you piss yourself as the Tesla tries to take you into the HOV lane THROUGH THE BARRIERS.
To date, SpaceX has sent nothing to Mars. Not to understate the company's accomplishment, but "people on HN" are fed up exactly with statements like yours.
My point is people will still be calling him a fraud when they do get it to mars, no evidence is sufficient for the HN cynic that thinks their “above the fray” ethos makes them smart.
Tesla has had massive success despite the haters, the model y becoming the literally best selling car on earth and you wouldn’t know it from HN. FSD has gotten really good, good enough to use more than not as they continue to improve it.
The best thing about capitalism is the losers here don’t matter - the winners get rich and keep going.
Have you met any humans? Or seen people driving?
But what is the point to use it everywhere if you still need to pay attention to the road, keep hands on the steering wheel?
It’ll be nice when that’s not required anymore, but even today it’s way more comfortable.
How is it politicization when TESLA THE COMPANY is saying Full Self Driving doesn't mean "Full" "Self" Driving?
If it is as good as you claim, why doesn't Tesla claim it's Full Self Driving?
Why would anyone listen to the opinion of someone who bought a Tesla in 2025?
The only people still buying them are musk fanboys.
Check this for a reference of how well Tesla's vision-only fares against the competition, where many have LiDAR. Keep it simple wins the game. https://www.youtube.com/watch?v=0xumyEf-WRI
One analyst asked about the reliability of Tesla’s cameras when confronting sun glare, fog, or dust. Musk claimed that the company’s vision system bypasses image processing and instead uses direct photon counting to account for “noise” like glare or dust.
This... is horseshit. Photon counting is not something you can do with a regular camera or any camera installed on a Tesla. A photon counting camera doesn't produce imagery that is useful for vision. Even beyond that, it requires a closed environment so that you can you know, count them in a controlled manner, not an open outside atmosphere.
It's bullshit. And Elon knows it. He just thinks that you are too stupid to know it and instead think "Oh, yeah, that makes sense, what an awesome idea, why is only Tesla doing this?" and are wowed by Elon's brilliance.
But go ahead fight some weird strawman you built.
Did you even look at the video? Don't think you did.
tl;dr: you can use optics to determine if there's rain on a surface, from below, without having to use any fancy cameras or anything, just a light source and light sensor.
If you're into this sort of thing, you can buy these sensors and use them as a rain sensor, either as binary "yes its rained" or as a tipping bucket replacement: https://rainsensors.com
I rented a Tesla a while back and drove from the bay to the death valley. On clear roads with no hazards whatsoever, the car hit the brakes at highway speeds. It scared the bejeesus out of me! Completely off put by the auto drive and derailed plans to buy a Tesla.
The filters introduce the problem of incorrectly deleting something that really is there.
Careful. HN takes a dim view of puns.
Self-starting wipers uses some kind of current/voltage measure on the windshield right - unrelated to self-driving? It's been around longer than Tesla - or are you just saying it's another random failure?
LIDAR requires line-of-sight (LoS) hence cannot see around conner, but RADAR probably can.
It's interesting to note that the all time 2nd most popular post on Tesla is 9 years ago on its full self driving hardware (just 2nd after the controversial Cybertruck) [1].
>Elon's vision-only move was extremely "short-sighted"
Elon's vision was misguided because some of the technologists at the time including him seem to really truly believed that AGI is just around the corner (pun attended). Now most of the tech people gave up on AGI claim blaming on the blurry definition of AGI but for me the truly killer AGI application is always full autonomous level 5 driving with only human level sensor perceptions minus the LIDAR and RADAR. But the complexity of the goal is very complicated that I really truly believe it will not be achieved in foreseeable future.
[1] All Tesla Cars Being Produced Now Have Full Self-Driving Hardware (2016 - 1090 comments):
This is because the vision system thinks there is something obstructing its view when in reality it is usually bright sunlight -- and sometimes, absolutely nothing that I can see.
The wipers are, of course, the most harmless way this goes wrong. The more dangerous type is when it phantom-brakes at highway speeds with no warning on a clear road and a clear day. I've had multiple other scary incidents of different types (swerving back and forth at exits is a fun one), but phantom braking is the one that happens quasi-regularly. Twice when another car was right behind me.
As an engineer, this tells me volumes about what's going on in the computer vision system, and it's pretty scary. Basically, the system detects patterns that are inferred as its vision being obstructed, and so it is programmed to brush away some (non-existent) debris. Like, it thinks there could be a physical object where there is none. If this was an LLM you would call it a hallucination.
But if it's hallucinating crud on a windshield, it can also hallucinate objects on the road. And it could be doing it every so often! So maybe there are filters to disregard unlikely objects as irrelevant, which act as guardrails against random braking. And those filters are pretty damn good -- I mean, the technology is impressive -- but they can probabistically fail, resulting in things that we've already seen, such as phantom-braking, or worse, driving through actual things.
This raises so many questions: What other things is it hallucinating? And how many hardcoded guardrails are in place against these edge cases? And what else can it hallucinate against which there are no guardrails yet?
And why not just use LIDAR that can literally see around corners in 3D?