Hacker News

Based on that list it boils down to 2 things it seems:

- cost (no longer a problem)

- too much code needed and it bloats the data pipelines. Does anyone have any actual evidence of this being the case? Like yes, code would be needed, but why is that innately a bad thing? Bloated data pipelines feels like another hand-wave when I think if you do it right it’s fine. As proven by Waymo.

Really curious if any Tesla engineers feel like this is still the best way forward or if it’s just a matter of having to listen to the big guy musk.

I’ve always felt that relying on vision only would be a detriment because even humans with good vision get into circumstances where they get hurt because of temporary vision hindrances. Think heavy snow, heavy rain, heavy fog, even just when you crest a hill at a certain time of day and the sun flashes you

The points linked repeatedly focus on cost and complexity as justification, even explicitly stating musks desire to minimise components in Kaparthy’s list.

They don’t focus on safety or effectiveness except to say that vision should be ‘sufficient’. Which is damning with faint praise imho.

If that link was to try and argue that the removal of sensors makes perfect sense i have to point out that anyone that reads that would likely have their negative viewpoint hardened. It was done to reduce cost (back when the sensors were 1000’s) and out of a ridiculous desire by Musk for minimalism. It’s the same desire that removed the indicator stalk i might add.

Instead of betting on RADAR and LIDAR HW getting better and cost going down, they went with vision only approach. Everybody in this field knows the strengths and weakness of each system. Multi-modal sensor fusion is the way to go for L4 autonomy. There is no other way to reduce the risk. Vision only will never be able to achieve L4 in all the weather conditions. Tesla may try to demonstrate L4 in limited geography and in good weather conditions but it won't scale.

I don't think that's the reasoning.

The reasoning was simply that LIDAR was (and incorrectly predicted to always be) significantly more expensive than cameras, and hypothetically that should be fine because, well, humans drive with only two eyes.

Musk miscalculated on 1) cost reduction in LIDAR and 2) how incredible the human brain is compared to computers.

Having similar sensors certainly doesn't guarantee your accidents look the same, so I don't think your logic is even internally sound.

Until a lawyer points out other cars see that. My car already has various sensors and in manual driving sounds alarms if there is a danger I seem not to have noticed. (There are false alarms - but most of the type I did notice and probably should have left more safety margin even though I wouldn't hit it)

also regulators gather srastics and if cars with something do better they will mandate it.

Very recent issue with Waymo https://dmnews.co.uk/waymo-robotaxi-spotted-unable-to-cross-.... This is 17 years after they bet the farm on LIDAR, with no signs its ever going to be cost effective or that it's better than multiple cameras, with millisecond reaction 360 degrees, that never gets tired, drunk, distracted, and also has other cheaper sensors and NN trained on Billions or real world data.

It is sound to think that cameras plus an accelerometer, plus data about about the car and environment (that you get from your ears) ought to be able to mimic and improve on human driving. However humans general purpose spatial awareness and ability to integrate all kinds of general information is probably really hard to replicate. A human would realize that an orange fluid spilling across the road might be slippery, guess the way a person might travel from the way their eyes are pointing...

It may just be faster to make lidar cheap. And lidar can do things humans can't.

IIUC, the cameras in a Tesla have worse vision (resolution) at far distances than a human. So while in the abstract your argument sounds fine; it'll crumble in court when a lawyer points out a similar driver would've needed corrective lens.

Most accidents happen because people are human, aren't paying attention, are inebriated, not experienced enough drivers, or reckless.

It's not fair to say that vision based models will "make the same mistakes people do" as >99% of the mistakes people make are avoidable if these issues were addressed. And a computer can easily address all those issues

This is a new and flawed rationale that I haven't heard before. Tesla cameras are worse (lower resolution, sensitivity, and dynamic range) than human eyes and don't have "ears" (microphones).

Pretty hard to do if your whole selling point is ‘better and safer than human’ however?

> lidars don't work well in rain/snow/fog.

Neither do cameras, or eyeballs.

No, it isn't "smarter." Camera-only driving is the product of a stubborn dogmatic boss who can't admit a fundamental error. "Just make it work" is a terrible approach to engineering.

Why does this matter? You have to slow down in rain/snow/fog anyway, so only having cameras available doesn't hurt you all that much. But then in clear weather lidar can only help.

Nothing works perfectly in all conditions and scenarios. Sensor fusion has been the most logical approach now, and into the foreseeable future.

Computer vision does not work exactly like human vision, closely equating the two has tended to work out poorly in extreme circumstances.

High performance fully automated driving that relies solely on vision is a losing bet.

Why does that strategy absolutely require the lidar to be absent from the car? When was less technology the solution to a software problem?

Limited resources? Billions per year are being thrown at the base technology. We have the capital deployed to exhaust every path ten times over.

The Swiss cheese model would like to disagree.

When you have sensor ambiguity sounds like the perfect time to fail safely and slow to a halt unless the human takes over.

Do cameras work well in those conditions? Nope. Also cameras don't work well with certain answer of glare, so as a consumer I'd rather have something over-engineered for my safety to cover all edge cases...

Evidence clearly shows otherwise.

Also, military sensor use shows the best answer is to have as many different types of sensors as possible and then do sensor fusion. So machine vision, lidar, radar, etc.

That way you pick up things that are missed by one or more sensor types, catches problems and errors from any of them, and end up with the most accurate ‘view’ of the world - even better than a normal human would.

It’s what Waymo is doing, and they also unsurprisingly, have the best self driving right now.

This is silly. Cameras are cheap. Have both. Sensors that sense differently in different conditions is not an exotic new problem. The kalman filter has existed for about a billion years and machine learning filters do an even better job.

Stereo cameras? My 2015 Subaru has them to detect obstacles and it works great.

Human/animal vision uses way more than parallax to judge distances and bearings - it uses a world model that evolved over millions of years to model the environment. That's why we can get excellent 3D images from a 2D screen, and also why our depth perception can be easily tricked with objects of unexpected size. Put a human or animal in an abstract environment with no shadows and no familiar objects, and you'll see that depth perception based solely on parallax is actually very bad.

Human eyes are much better than cameras at dealing with dynamic range. They’re also attached to a super-computer which has been continuously trained for many years to determine distances and classify objects.

> Human eyes do not have distance information

Single human eyes do resolve depth perception. Not as good as binocular vision, but you don't loose all depth perception of you lose an eye.

https://en.wikipedia.org/wiki/Monocular_vision

I don’t like the comparison between humans and humans. Humans don’t travel around at 100mph in packs of other humans. Why not use every sensor type at our disposal if it gives us more info to make decisions? Yes I understand it’s more complicated, but we figure stuff out.

Let me know when you have a camera package with human eye equivalency.

Because they don't have superhuman LIDAR. They never did. Nobody ever did. LIDAR input is not completely reliable so what do you do then?

How do you explain the reports of Robotaxis running into fixed objects? If what you are saying is true that shouldn't be able to happen.

https://electrek.co/2026/02/17/tesla-robotaxi-adds-5-more-cr...

The phantom braking issue with auto pilot tells me it can’t. A shadow from a tree doesn’t trigger your brakes locking up at 70+ mph when there’s a lidar sensor to tell you it’s not a physical object.

“Just buy FSD” isn’t a reasonable answer to a problem literally no other automaker suffers from.

Luckily everyone else in the comments is an expert. And also doesn't recognize that Tesla's already drive themselves and did not need Lidar. They also mischaracterize the reasoning.

> I'm not sure what would be wrong with the vision system that lidar needs to fix.

This conversational disconnect is as old as the hills:

1. Person 1 asks "what's wrong" (if it ain't broke don't fix it)

2. Person 2 wants to make something better

My meta-goal here on HN (and many places where people converse) is for people to step back and recognize the conversational context and not fall into the predictable patterns that prevent us from making sense of the world as best as we can.

Yeah it's BS. Tesla uses lidar where it makes sense: They have a small lidar fleet to collect ground truth depth data for better vision estimation. This part is long solved.

You're way off if you think that Waymo and FSD are anywhere close.

1. Tesla is not competitive with Waymo, they're not even in the same class. Waymo is 10 years ahead at least. I understand you can't buy a Waymo, but still.

2. Other car companies are properly valued, Tesla is overinflated.

3. Other cars, even basic Hondas, have the same level of self driving as Teslas.

4. Other car companies don't lie to their customers about their capabilities or what they're buying.

There is certainly some truth that "some company" overpromised and underdelivered. They advertise "full self driving" but then hide in the fine-print that "oh jk, not really, but its still full self driving if anyone asks ;) ;) ;)"

I think the frustration stems from the obvious falsehoods in the advertising, and the doubling-down on the tech, despite the well-documented weaknesses of the implementation.

No, cameras cannot create reliable distance measurements in real-world conditions. Parallax is not a great way to measure distance for fast, unpredictably moving objects (such as cars on the road). And dirt or misalignment can significantly reduce accuracy compared to lab conditions.

Note that humans do not rely strictly on our eyes as cameras to measure distances. There is a huge amount of inference about the world based on our internal world models that goes into vision. For example, if you put is in a false-perspective or otherwise highly artifical environment, our visual acuity goes down significantly; conversely, people with a single eye (so no parallax-based measurement ability) still have quite decent depth perception compared to what you'd naively expect. Not to mention, our eyes are kept very clean, and maintain their alignment to a very high degree of precision.

Stereo cameras are useless against repeating patterns. They easily match neighboring copies. And there are lots of repeating or repeating-like patterns that computers aren't smart enough to handle.

You can solve this by adding an emitter next to the camera that does something useful, be it just beaconing lights or noise patterns or phase synced laser pulses. And those "active cameras" are what everyone call LIDARs.

There are tons of evidence showing that cameras are alone are not safe enough and even Tesla has realized that removing lidar to save cost was a mistake.

'cameras can see in color, therefore lidar is unnecessary for self driving' is unconvincing

> ridiculous idea by a certain set of people that lidar is necessary.

"Necessary"? Seems like a straw man, don't you think? I strive to argue against the strongest reasonable claim someone is making.

Lots of reasonable people suggest LIDAR is helpful to fill in gaps when vision is compromised, degraded, or less capable.

People running businesses, of course, will make economic trade-offs. That's fine. But don't confuse, say, Elon's economic tradeoff with the full explanation of reality which must include an awareness that different sensors have different strengths in different contexts.

So, when one thinks about what sensor mix is best for a given application, one would be wise to ask (and answer) such questions as:

- What is the quality bar?

- What sensors are available?

- Wow well do various combinations of sensors work across the range of conditions that matter for the quality bar?

- WRT "quality bar": who gets to decide "what matters"? The company making the cars? The people that drive them? regulators that care about public safety. The answer: it is a complex combination.

It is time to dismiss any claim (or implication) that "technology good, regulation bad". That might be the dumbest excuse for a philosophy I've ever heard. It is the modern-day analogue of "Brawndo's got what plants crave." Smart people won't make this argument outright, but unfortunately, their claims sometimes reduce to this level of absurdity. Neither innovation nor regulation are inherently good nor bad. There are deeper principles in play.

Yes, some individuals would use their self-proclaimed freedom to e.g. drive without seatbelts at 100 mph at night with headlights off. An extreme example, but it is the logical extension of pure individualism run amok. Regulators and anyone who cares about public safety will draw a line somewhere and say "No. Individual stupidity has a limit." Even those same people would eventually come to their senses after they kill someone, but by then it is too late.

I have a suspicion here on HN. When criticizing big tech, especially Google and FB, at a certain time of the day a specific cohort comes online and downvotes. Suspiciously, that is a time when one could conclude, that now people in the US start working or come online. Either fanboys, employees or an organized group of users trying to silence big tech criticism.

I have no proof of course and it might be coincidence, or just difference of mindset between US citizens and Europe citizens. It happened a few times already and to me looks sus.

But if they actually read and not just ctrl+f <company name>, then of course not writing the company name, but hinting at it in an obvious way is no more helpful either.

April 2019: https://www.youtube.com/live/Ucp0TTmvqOE?t=9220s

There are probably even earlier statements from him against lidar...

WTF was their calculus on the break-even liability point? The "if we do this, we save X amount of money, but stand to lose Y in lawsuits for cases where the usage of LIDAR could have otherwise prevented it."

You're forgetting the nervous system and the brain connected to those eyes (and vestibular system).

> Why make things more complicated than they need to be? Humans don't have lidar and we are the only intelligence that can reliably drive.

Because we want self driving cars to be safer than human driven cars.

If humans had built in lidar we would use it when driving.

Self driving cars are not equipped with human brains so this doesn’t really make sense.

“We should achieve self driving cars via replicating the human brain” strikes me as an incredibly inefficient and difficult way to solve the problem.

> we are the only intelligence that can reliably drive.

Science would like to point out that rats also can learn to drive

https://theconversation.com/im-a-neuroscientist-who-taught-r...

The bitter lesson I think is a great way of explaining the logic behind Tesla's strategy. People aren't getting it.

Whether or not it'll actually work remains to be seen, but it's a perfectly reasonable strategy. One counterargument would be that the bitter lesson can be applied to LIDAR too; you don't have to use that data for feature engineering just because it seems well suited for it.

Humans can drive with eyes only, but we are better drivers when we can also use other senses like hearing. If humans has lidar we would use it when driving.

Don't cars already use a ton of sensors that don't reproduce human senses and ways of doing things?

This knee-jerk reply is old and tired, and the counterarguments are well-trod at this point. Even if cameras-only can build a car that’s as good as humans, why should we settle for “as good as“ humans, who cause 40,000 fatalities a year in the US? If we can do better than humans with more advanced sensors, we are practically morally obligated to do that.

In a saner world Teslas would be running Waymo's self-driving stack instead of the half-baked "might kill you at any time" not quite-FSD.

It would be smarter to take that approach. Google's core competency is technology, technical infrastructure, and research. More mundane things like manufacturing and customer service are... shall we say, less of a core competency. Take the high value add, leave other things to automakers to duke it out. Also good for avoiding attracting even more regulatory attention.

That is one plausible outcome. Waymo is experimenting with partnerships with ride hailing apps on the one hand, and building their software into Toyotas on the other hand. So far they have built a few thousand vehicles in a factory run by Magna, which specializes in low volume vehicles. Hyundai wants to sell Waymo tens of thousands of vehicles. That's going to look different in fundamental ways.

With the price declines in ev we are talking about 1 million ev even with all the waymo tech for $50 billion soon. approximate Annual Revenue of a private hire car is $50+k ie $50-60 billion a year for a million cars. But total taxi driver population is 350-400k in the US. I think people are underestimating the electric tech + ai/automation to hit soon.

Why sell cars to everyone?

People on here used to buy servers themselves (very few of us still do), most now rent via cloud.

Why should transportation be different?

Alphabet wants drivers on their devices looking at ads instead of driving.

Do OEMs want to manage their own ride-share platforms? 10+ apps/providers?

> but does Alphabet actually intend Waymo to be a trillion dollar retail car business itself, selling cars to everyone?

Google doesn't do retail other than Chromecast and Pixel phones, and that is already annoying to them as it is because it involves something Google is notoriously bad at - actual customer support.

Starting up a car brand is orders of magnitude worse.

For one, people actually need to trust your brand to survive for at least five to ten years - cars are an investment, and a car that I can't trust to get safety-relevant spare parts (brake rotors, brake pads, axle bearings) all of a sudden is essentially an oversized paperweight. For a company such as Google, this alone (remember Killed By Google) is a huge obstacle to overcome.

Then, you need production. Sure, you can go to Magna or other contract manufacturers, or have an established large brand build vehicles for you, or you say you have to go the Tesla route and build everything from scratch. Either way has associated pros and cons.

And then, you need a nationwide network of spare parts, dealerships, repair shops and technicians that can fix the issues that people will get alone because the wide masses abuse cars in ways you might not even dare think about while testing, or because other people run into your cars and so your cars need repairs.

Even being a derivative of an established car brand can be a royal PITA. Let's take Mercedes Benz as an example with the 2003-2009 Mercedes-Benz SLR McLaren. On paper, it's a Mercedes vehicle, with a lot of the parts actually originating from stock Mercedes cars - but most dealerships will refuse to work on it. Either because they lack the support to even properly jack the car up, or because they lack the specialized tools for the AMG engine, or because they cannot even order the parts as Mercedes gates repairs for that thing to special shops. Or, again Mercedes, with Maybach luxury cars. The situation isn't as bad as with the McLaren, but their cars are challenging in another way - the S 650 Pullman weighs around 3 metric tons empty and is 6.50 meters long. Good luck finding a jack even capable of lifting that beast, most Mercedes sports-car shops don't carry jacks that are normally used to lift Mercedes Vito transporters!

Even Tesla, and they've been at it for the better part of two decades, still struggles with that. Their shitty spare parts logistics actually drive up not just insurance prices for their own customers, but for everyone - hit a Tesla with your Dodge and be at fault, and now your insurance has to pay out for months of a rental car because Tesla can't be arsed to provide the body shop the Tesla ends up at with spare parts in any reasonable time.

Established car brands however have all of that ironed out for many, many decades now. American, Asian, European, doesn't matter. And the spare parts don't even have to be made for cars: ask your local Volkswagen dealer to order a few pieces of "199 398 500 A" and one piece of "199 398 500 B" and you'll probably have a lead time of less than a day, at least in Germany - for the uninitiated: that part number belongs to the famous sausage, the second one to the accompanying curry ketchup, with more sausages being sold each year than actual cars.

And established car brands also bring something to the table: their own experiences with integrating smart technology. Yes, particularly German carmakers are notoriously bad in that regard, but for example Mercedes Benz was the first car brand in the world to get a certified Level 3 system on the road [1] and are now working on a Level 4 certification [2]. That kind of experience in navigating bureaucracy, integration and testing cannot be paid for in money.

tl;dr: I see no way in which Waymo goes to general availability regarding selling cars. They will run their own autonomous car fleets in select markets where they can fully control everything, but seeing Waymo tech generally available will be as part of established car brands.

[1] https://group.mercedes-benz.com/technologie/autonomes-fahren...

[2] https://group.mercedes-benz.com/technologie/autonomes-fahren...

> Before Waymo deploys in a new city, it deploys a huge fleet of cars that spend months of driving completely supervised, presumably to construct a detailed LIDAR map of the city.

Not entirely true. From their recent "road trips" last year, the trend is they just deploy less than 10 cars in a city for a few weeks (3-4 weeks from what I recall) for mapping and validating. Then they come back after a few months to setup infrastructure for ride hailing (depot, charging, maintenance, etc.) and start service.

Suborbital ballistic "travel" beyond continental distances, is almost as expensive as orbital. If you can make it to the antipode, you're basically almost orbital.

Suborbital "trips" straight up, beyond the atmosphere, are very cheap.

Maybe. But Google has been there in a sensor laden car, overhead with an airplane, and buying all the access that is available in satellite imagery, and fusing that together in a continually updated model. Plus real time data from a billion maps and navigation users. I pity the fool going up against that.

Yeah, imagine having, say, two of these LIDAR sensors, each pointed towards the car's blind spots. Comma already does well with the car's built-in radar + vision on straight freeway runs, but can't reliably change lanes on its own. The built in blind spot detectors on most cars are a binary "there is/not a car present", which doesn't reliably determine if it's safe to actually do a lane change.

I think of Tesla autopilot as sophisticated cruise control. Can perform most driving tasks better than I can, saves a lot of cognitive work, still needs close of my 100% attention.

Is this comment from 2010? Maybe I'm missing your point, but it seems you would be shocked by what modern cars are capable of.

I had to look this up, because I had never heard of it. How could a lens be damaged by infrared lasers?

It turns out it’s the sensors that are easily damaged by high powered lidar lasers.

https://spectrum.ieee.org/amp/keeping-lidars-from-zapping-ca...

Is there any deeper study on long term effects regarding retinal damage?

I would imagine, even with safe dosages, there would be some form of cumulative effect in terms of retinal phototoxicity.

More so if we consider the scenario that this becomes a standard COTS feature in cars and we are walking around a city centre with a fleet of hundreds of thousands of these laser sources.

I wish this was true. It'd immediately be the best way to fight surveillance systems like Flock

iPhones have had lidar for years, have cameras been affected?

Could be a gain for privacy ;-)

TIL!

Thanks! What a headache

we'd likely see new coatings and sensor designs that avoid it, not trivial but also not the end of the world

What? Please explain!

I think the difference is that these are intended for automotive use and have a much longer range than the ones in your Roomba.

I think we’re well past the point where mass surveillance was a technical challenge. Mass oppression through autonomous violence however…

LIDAR would be preferrable to cameras when it comes to privacy actually

The minute internet became widespread it was game over.

Pros and cons. :/

It'll never happen, but we need a bill of rights for privacy. The laypeople aren't well-versed or pained enough to ask for this, and big interest donors oppose it.

Maybe the EU and states like California will pioneer something here, though?

Edit: in general, I'm far more excited by cheap lidar tech than I am afraid of the downsides. We just need to be vigilant.

Humanity has never known a world without surveillance. Responsibility cannot exist without being watched. Primitive tribes lived under the constant eye of the group, and agricultural eras relied on the strict oversight of the clan. Modern states simply adopted new tools for an ancient necessity. A society without monitoring is a society without accountability, which only leads to the Hobbesian trap of endless conflict.

Cepton primarily operates B2B, as B2C demand for specialized LIDAR like this is pretty low. Anything you find on eBay is either a leftover dev kit or salvage. This is pretty much the case for MicroVision, Ouster etc.

Also, MicroVision, the company in OP's article bought the IP from Luminar. This feels like a circular venture capital scam. Luminar originally went public via SPAC and made a bunch of people very wealthy before ultimately failing.

The same Luminar from the Mark Rober video?

https://www.forbes.com/sites/bradtempleton/2025/03/17/youtub...

This is very wrong. LIDAR scanners have revolutionized surveying by enabling rapid, high-precision 3D mapping of terrain and infrastructure, capturing millions of data points per second. LIDAR can penetrate dense vegetation, allowing accurate, ground-level, mapping in forested or obstructed areas. Drone mounted LIDAR has become very popular. Tripod mounted LIDAR scanners are very commonly used on construction sites. Handhels LIDAR scanners can map the inside of buildings with incredible accuracy. This is very commonly used to create digital twins of factories.

Lidar is critical for any autonomous vehicle. It turns out a very accurate 3D point cloud of the environment is very useful for self driving. Crazy, I know.

Neato from San Diego has developed a $30 (indoor, parallax based) LIDAR about 20 years ago, for their vacuum cleaners [1].

Later, improved units based on the same principle became ubiquitous in Chinese robot vacuums [2]. Such LIDARs, and similarly looking more conventional time-of-flight units are sold for anywhere between $20-$200, depending on the details of the design.

[1] https://scholar.google.com/scholar?q=%22A+Low-Cost+Laser+Dis... [2] https://github.com/kaiaai/awesome-2d-lidars/blob/main/README...

Sounds like the quality isn't all that great but LD06 sensors look like they're about $20 and someone who works on libraries about this suggested the STL27L which seems to be about $160 and here's an outdoor scan from it: https://sketchfab.com/3d-models/pidar-scan-240901-0647-7997b...

Not sure if the ld06 is a scanner like this or if it's just a line (like you'd use for a cheaper robot vac).

What are the chances some non-trivial proportion of the millions of cars on the road will not have their LIDAR designed, built, installed or calibrated correctly? I suspect this is going to be a recognized public health issue in a decade or two. (It will likely be an issue well before that, but unrecognized...)

That's a lot of qualifiers. And replace "humans" with "cameras" and I'm reminded that despite their well-intentioned efforts Volvo has failed there already.

For example this one has 120m range with 1cm accuracy and its 15 euros: https://www.temu.com/bg-en/-digital-laser-distance-meter-50m...

Sure, these are the assumptions but silicon is silicon, copper is copper and solder is solder. They don't use easy melting electronics in vacuums and hardened stuff in cars, the tech is about the same unless it is supposed to work in highly radioactive environment. The plastics are different but car interiors are full of plastics, so its unlikely that the costs of temperature resistant plastics needed for this is more than a cupholder.

As for the range, again pretty powerful lasers are sold for sub 10SUD prices on retail. I am sure that there must be higher calibration and precision requirements as the distance increase but is it really order of magnitudes higher? 120 meters laser measurer with 1cm accuracy is 15 Euros on Temu and that thing has an LCD screen and a battery as a handheld device. How much distance do you actually need?

Not only that but vibrations play a big part as well, especially on ICE vehicles.

There is no evidence of any such organised campaign. The critical comments we see against that company and person are generally from known, established HN users, and align with frequently-expressed sentiments among the general public. And the complaint is just as often made that "anything remotely critical" about that company and person is flagged. If posts about the topic are being downvoted and flagged, it's mostly because that person and company are in the news so frequently that most commentary about them is repetitive, sensationalist and uninteresting, and thus off topic for HN.

Or everyone is just tired of tesla and their stubborn camera only tech that will fail in higher autonomy cases?

No no it's the cabal...

Could be lurkers triggered

It is a Class 1 laser. I can buy a Class 4 laser online that I can start fires with. Laser danger is not the reason.

If you want to compete with LIDAR, you need high resolution 4D (range, velocity, azimuth, and height) RADAR. Those are usually phased arrays with expensive phase sensitive electronics, and behind that a chip that can do a lot of Fourier transforms very quickly.

The cheap RADAR devices you're talking about usually only output range and velocity, sometimes for a handful of rather large azimuth slices. That doesn't compete with LIDAR at all.

The main damage risk from LIDAR is to retinal rods and cones. You just know some jerk is going to overclock his system and we know some people just don't care about the harm they cause so long as they get a benefit. As a combo that means I'll be wearing protective eyewear outdoors the day this tech comes to the roads.

I took the Waymo from San Jose airport to home on the peninsula. It took the 101 highway back for the most part, driving very conservatively at 65-55 mph, and in the right most lane. It still has a few quirks though. When there aren't any cars around it will speed up to 65 mph, but at on-ramps, it will slow down to 55 and then speed up once past. It will get stuck behind slow drivers being in the right most lane and patiently follow them a few car length behind them. On the plus side, the lidar stack field of view as shown on the internal display seems to see pretty far down the highway.

Tesla doesnt have Lidar?

Why wouldn't you trust a Telsa, millions of people let there Tesla drive them all over USA (not geofences like Waymo) without touching the wheel from parking spot to parking spot everyday. Have you tried it?

It's fair to point out that comma.ai is SAE level two system, however it's not geofenced at all, which is an SAE level 5 requirement. But really that brings up the fact that SAE's levels aren't the right ones, merely the ones they chose to define since they're the standards body. A better set of levels are the seven I go into more detail about on my blog.

As far as distinguishing shadows on the road, that's what radar is for. Shadows on the road as seen by the vision system don't show up on radar as something the vehicle will run into.

I wouldn't be surprised if this was a better solution. I think while radar might have a worse spatial resolution, it's depth perception, speed measurement capability, and general robustness to adverse weather might make it a better complementary sensor.

Lidar struggles with things like rain and snow way worse than cameras do.

Xpeng, Wayne, aiMotive to name three. Probably many others, who claim to use LIDAR but don’t actually rely on it. Because LIDAR is perceived as a prerequisite for autonomous safety, admitting to not needing it is a bad PR move — for now.

Tesla is using radars as well as cameras. No one is using only cameras

Nope...

Humans also don't have wheels, but we build objects with wheels. It is as if we can build objects that don't resemble humans for specific purposes. Crazy...

> Tesla are producing cyber cabs now which are 10th the price of Waymo's and can drive autonomously anywhere in the world.

My understanding is that cyber cabs still need safety drivers to operate, is that not the case?

> Tesla are producing cyber cabs now which are 10th the price of Waymo's and can drive autonomously anywhere in the world.

Wait what? when did they actually enter mass production?

> I mean humans have Lidar sensors

Real time slam is actually pretty good, the hard part is reliable object detection using just vision. Tesla's forward facing cameras are effectively monocular, which means that its much much harder to get depth (its not impossible but moving objects are much more difficult to observe if you only have cameras aligned on the same plane with no real parallax)

Ultimately Musk is right, you probably don't need lidar to drive safely. but its far more simple and easier to do if you have Lidar. Its also safer. Musk said "lidars are a crutch", not because he is some sort of genius, Its obvious that SLAM only driving is the way forward since the mid 00's (of not earlier). The reason he said it is because he thought he could save money not having lidar. The problem for him is that he didn't do the research to see how far away proper machine perception is to account for the last 1% in accuracy needed to make vision only safe and reliable.

This is a weirdly tired counterpoint that Elon and Elonstans like to bandy about as if it's an apples to apples comparison. Humans have a weirdly ultra-high-dynamic-range binocular vision system mounted on an advanced ptz/swivel gimbal that allows for a great degree of freedom of movement, parallax effects, and a complex heuristic system for analyzing vision data.

The Tesla FSD system has... well, sure, a few more cameras, but they're low resolution, and in inconveniently fixed locations.

My alley has an occlusion at the corner where it connects to the main road: a very tall, very ample bush that basically makes it impossible to authoritatively check oncoming traffic to my left. I, a human, can determine that if I see the light flicker even slightly as it filters through the bushes, that the path is not clear: a car is likely causing that very slight change in light. My Tesla has no clue at all that that's happening. And worse, the perpendicular camera responsible for checking cross-traffic is mounted _behind my head_ on the b-pillar, in a fixed location that means that without nosing my car _into_ the travel lane, there is literally no way for it to be sure the path is clear.

This edge case is navigated near-perfectly by Waymo, since its roof-mounted lidar can see above and beyond the bush and determine that the path is clear. And to hit back on the "Tesla is making cheaper cars that can drive autonomously anywhere in the world": I mean, they still aren't? Not authoritatively. Not authoritatively enough that they aren't seeing all sorts of interventions in the few "driverless" trials they're doing in Austin. Not authoritatively enough when I have my Tesla FSD to glory. It works well enough on the fat part of the bell curve, but those edges will get you, and a vision only system means that it is extremely brittle in certain conditions and with certain failure modes, that a lidar/radar backup help _enhance_.

Moreover, Waymo has brought lidar development in-house, they're working to dramatically reduce their vehicle platform cost by reducing some redundant sensors, and they can now simulate a ground truth model of an absurd number of edge cases and odd scenarios, as well as simulate different conditions for real-world locations in parallel with their new world modeling systems.

None of which reads to me as "not going well for Waymo." Waymo completes over 450,000 fully autonomous rides per week right now. They're dramatically lowering their own barriers to new cities/geographies/conditions, and they're pushing down the cost per unit substantially. Yeah, it won't get to be as cheap as Tesla owning the entire means of production, but I'm still extremely bullish on Waymo being the frontrunner for autonomous driving for the foreseeable future.

> Yes, silly using just cameras, I mean humans have Lidar sensors, that why they can drive, why didn't new just copy that....oh wait.

Humans don't have wheels and cannot go 70MPH. Humans also don't have rear view cameras and cannot process video feeds from 8 cameras simultaneously. The point of these machines is to be better than humans for transportation. If adding LIDAR means that these vehicles can see better than humans and avoid accidents that humans do get into, then I for one want them in my vehicle.

The human brain is a product of millions of years of dealing with spatial problems for survival — and most individual humans are the product of thousands of hours of experience using it to navigate the physical world.

We're always getting closer at emulating this, but we're still a ways off from matching it.

I don't understand what you're saying.

Stereo based depth mapping is kind of bad, especially so if it is not IR assisted. The quality you get from Lidar out of the box is crazy good in comparison.

What you can do is train a model using both the camera and Lidar data to produce a good disparity and depth map but this just means you're using more Lidar not less.

>It all seriousness though, Tesla are producing cyber cabs now which are 10th the price of Waymo's and can drive autonomously anywhere in the world. I think we can see where this is going. (Hint: not well for Waymo)

This feels like a highly misleading claim that might technically be true in the sense that there are less restrictions, but a reduction in restrictions doesn't imply an increase in capability.

The comment about Waymo seems to be particularly myopic. Waymo has self driving technology and is operating as a financially successful business. There is no conceivable situation where the mere existence of competition with almost the same capabilities would shake that up. Why isn't it companies like Uber, who have significantly fallen behind, that are in trouble?

>Also the article is speculative 'MicroVision says its sensor could one day break the $100 barrier'. One day...

And so is the comment about Tesla cyber cabs.

Humans cannot drive safely. Human drivers kill someone every 26 seconds. Waymos have never killed a person.

Part of that is that humans are distractible, and their performance can be degraded in many ways, and that silicon thinks faster than meat.

But part of it is the sensor suite. Look at Waymo vs Tesla robotaxi accident rates.