You need LIDAR

[mobada]

Doesn’t sound like controversy; sounds like Elon didn’t like the results. Which is fine, but not the evidence you were hoping for.

Tesla claimed there was 20% battery left and there was no need to push the car back.

Secondly, this is exactly why jeremy clarkson would make a review of the cybercab and lie about it. He has had beef with elon for almost 20 years

Tesla is currently operating way too many cars to have any proof this didn't happen

When tesla was running 20 or 30 cars they maybe recorded everything, but at 1300 not anymore

 

[mobada]

Right now if you actually look at tesla's robotaxi accident rate (and cause of accidents), if you filter out hitting thin chains and parking lot things at 1mph, and remove not preventable accidents like rear rending at low speed, etc. tesla's accident rate is extremely low.

This suggests that at least tesla's on-road accident avoidance and awareness of other cars is very, very good.

 

[HC_79]

Right now if you actually look at tesla's robotaxi accident rate (and cause of accidents), if you filter out hitting thin chains and parking lot things at 1mph, and remove not preventable accidents like rear rending at low speed, etc. tesla's accident rate is extremely low.

This suggests that at least tesla's on-road accident avoidance and awareness of other cars is very, very good.

So if we remove the crashes, the crash rate is excellent . That’s like saying the Titanic had a great safety record if you filter out icebergs

 

[mobada]

So if we remove the crashes, the crash rate is excellent . That’s like saying the Titanic had a great safety record if you filter out icebergs

hitting a 1mph thin chain is not a safety issue and waymo has issues with thin chains as well. Waymo uses HD maps so more of this stuff is strictly baked into the map

Tesla has gotten better at these things with v15.

https://twitter.com/x/status/2117778384923382075

By comparison, here's not preventing an obvious preventable accident even if technically not at fault.

 

[TahoeMichael]

From the very article you posted:


Doesn’t sound like controversy; sounds like Elon didn’t like the results. Which is fine, but not the evidence you were hoping for.

I'm not seeing the comment you're replied to, since I've made use of the ignore button. But I'll note that Musk has well documented instances of lying in response to press reports he didn't like. The one that immediately comes to mind is when Reuters reported that Tesla had canceled their low cost "Model 2" vehicle program, and Musk denied it. Within a few months it became clear that the Reuters report was correct. I wouldn't go so far as saying that everything Musk/Tesla says is false, but they also don't deserve a presumption of honesty at this point.

 

[HC_79]

hitting a 1mph thin chain is not a safety issue and waymo has issues with thin chains as well. Waymo uses HD maps so more of this stuff is strictly baked into the map

Tesla has gotten better at these things with v15.

https://twitter.com/x/status/2117778384923382075

By comparison, here's not preventing an obvious preventable accident even if technically not at fault.

That Waymo crash looks bad. If it had room to stop and didn’t, that’s a real failure and Waymo should own it. See how easy that was? I have no skin in the Waymo game, so admitting that cost me nothing. Unless you’re on Tesla’s payroll, admitting FSD has flaws should cost you nothing too. Your turn.

Because here’s your thread so far.

Post #12: FSD does not run red lights.
Post #15: okay, about 1 in 100,000 miles.
Post #19: fine, v13 did it, but it’s fixed now.

That’s not a defense, that’s a retreat with a press release. And the source for “fixed” is you. So a Tesla column is “total bullshit” and hitting chains is “not a safety issue,” but one Waymo tweet is evidence. Again, pick a standard.

Elon’s going to be thrilled at your performance review this year. “Exceeds expectations in deflection, needs improvement in acknowledging reality.

 

[mobada]

That Waymo crash looks bad. If it had room to stop and didn’t, that’s a real failure and Waymo should own it. See how easy that was? I have no skin in the Waymo game, so admitting that cost me nothing. Unless you’re on Tesla’s payroll, admitting FSD has flaws should cost you nothing too. Your turn.

Because here’s your thread so far.

Post #12: FSD does not run red lights.

FSD does not run red lights in the way he described it which was blatantly running a red light. In FSD v13 it would run a red light after stopping for some time, and it would take off if the coast was clear. This did not happen in a city as it would only happen if no other cars were around and it was an extremely rare event anyways. Whatever caused it, tesla completely fixed it by the time robotaxi came out. This issue with v13 also never happened in FSD v12. I can't remember there ever being a version of FSD at least from v11 up that ever blatantly ran a red light

Post #15: okay, about 1 in 100,000 miles.

I'm saying that if it ever ran a red light, it would be a very rare event for some obvious reason beyond it just doing it. The good thing about these computers is they don't make mistakes

That’s not a defense, that’s a retreat with a press release. And the source for “fixed” is you. So a Tesla column is “total bullshit” and hitting chains is “not a safety issue,” but one Waymo tweet is evidence. Again, pick a standard.

Hitting a chain is not a safety issue. What is a safety issue is a thing that puts other people who riders at risk of injury.

if a robotaxi scraped the same bollard every day until the end of time at 1mph that is not a safety issue.

Waymo never claimed their accident rate was low. It's not. They only say their injury causing accident rate is low.

And injury causing accident rate ignores some pretty serious failures. Waymo stopped inbetween gate arms and a train track. They had a car wash away for 4 days in flood waters. Those are deadly-level failures.

 

[MJAir23TheGoat]

Clarkson is a clown … clickbait
He’s been this way with his car reviews as well
Is his drawing card now
My daughter drives 60 miles daily to and from college in her M3 ever touches the steering wheel.

 

[mobada]

What people don't realize is waymo does almost all of their "validation" of their vehicles in simulation. There is a sim 2 real gap where you can't get confidence that a problem is fixed until you test it in the real world.

This is where tesla does something no legit robotaxi can do. They can make changes to the neural network, push an update out every 3 days and get pretty clear confidence that an issue is fixed. Testing on real scenarios is always better than simulation.

There is another gap if you train on synthetic data, and if you look at waymo's synthetic lidar data they have demonstrated, it is very inaccurate and it's like genAI images where you see hands with too many fingers, etc.

We're talking about the 99.999%+ level which is what you really need to operate unsupervised vehicles

 

[ButMuhElectrons]

The evidence is clear as day. FSD being used for billions of miles and it simply does not run red lights

This is not a scientific statement.
More an emotional one.
Do you have research or raw car data on "F"SD?

 

[mobada]

This is not a scientific statement.
More an emotional one.
Do you have research or raw car data on "F"SD?

Everything you see here is emotions and believing something that doesn't happen.

Tesla literally has a dash cam (sentry mode) that records 100% of the time while you drive. These failures of FSD are posted online

FSD is being used for billions of miles so you can tell when an issue still exists or not

There was never evidence of FSD just blatantly running a red light or misidentifying a red light as green. In FSD 13 there was that rare issue where it would take off after sitting at a red light for about 10 seconds. It would do that at like 2am when nobody was around and not sure what caused it. That was fixed with v14

The scientific evidence is literally there in the NHTSA SGO reports which look good in robotaxi favor after doing 6 million miles.

Then the goalposts will get moved again when tesla passes 100 million or 200 million miles. Tesla could do a billion miles and people will still say that camera only is not safe.

 

[ButMuhElectrons]

I think Tesla has a long way to go before they could be considered trustworthy.
(Not that other automakers are that far ahead)

https://www.washingtonpost.com/technology/2025/08/29/tesla-autopilot-crashes-evidence-testimony-wrongful-death/

 

[mobada]

The most ridiculous thing of all is we have 20 brands in china with self driving.

They have the compute, lidar, radar, "HD-like maps", etc.

And they have all been working on their systems for 5 years. The best systems have poor reaction time. Poor safety. They don't dodge debris on roads. Can't perform unprotected turns without making dangerous errors. Have to be taken over to prevent a crash during simple uturns. Nonexistent accident avoidance.

The best one, xpeng, claims that they will do unsupervised in 2028. They don't know. If you give a date more than 1 year out you don't know what you're doing.

Now rivian, who is like 3 years behind them, is going to deliver unsupervised robotaxi the same time as xpeng? And now lucid who has a poor adas system of their own is going to deliver the world's first consumer level 4 system?

And somehow it's the lidar and the compute that was the issue?

meanwhile it is trivial for tesla to sell a cybercab to a consumer and then consumer level 4 is already here

 

[mobada]

I think Tesla has a long way to go before they could be considered trustworthy.
(Not that other automakers are that far ahead)

https://www.washingtonpost.com/technology/2025/08/29/tesla-autopilot-crashes-evidence-testimony-wrongful-death/

That's literally an autopilot crash where tesla was not at fault. The person pressed the accelerator pedal with their hand which overrode AEB and crashed into that person while they were on the floor looking for their phone.

Also back when this accident happened, tesla did not collect video of accidents, only for FSD training. This was before the 2021 SGO update which has different reporting requriements.

The fact that you repeat this without any knowledge of what happend shows how much your emotions are getting in the way of facts.

The reason tesla was 35% at fault or whatever is they argued that autopilot driver assist should not have been enabled on city streets.

 

[borski]

Tesla claimed there was 20% battery left and there was no need to push the car back.

Cool. That’s not evidence unless they published data.

Secondly, this is exactly why jeremy clarkson would make a review of the cybercab and lie about it. He has had beef with elon for almost 20 years

That’s not evidence; it is circumstantial at best, and is not evidence of his having done anything dishonest here.

Tesla is currently operating way too many cars to have any proof this didn't happen

That’s nonsense; if it doesn’t happen, it should be trivial to find evidence of that. Instead, it’s trivial to find evidence it does happen. You are incorrect.

When tesla was running 20 or 30 cars they maybe recorded everything, but at 1300 not anymore

They still record everything.

 

[borski]

Right now if you actually look at tesla's robotaxi accident rate (and cause of accidents), if you filter out hitting thin chains and parking lot things at 1mph, and remove not preventable accidents like rear rending at low speed, etc. tesla's accident rate is extremely low.

Yes, if you filter out most accidents, nobody has many accidents!

This suggests that at least tesla's on-road accident avoidance and awareness of other cars is very, very good.

No, it doesn’t. It suggests you filtered out a lot of accidents.

 

[borski]

hitting a 1mph thin chain is not a safety issue

Hitting anything is an issue.

and waymo has issues with thin chains as well.

Great. Still an issue. Nobody claimed Waymo was perfect.

Waymo uses HD maps so more of this stuff is strictly baked into the map

Nonsense, and a good indication that you don’t know how this technology works.

Rather than treating HD maps as a strict, brittle track the car is forced to follow, Waymo uses them as a powerful "prior"; analogous to a mental memory of the environment. The onboard computer combines data from its sensors (lidar, radar, and cameras) with the HD map to streamline navigation, especially in low-visibility or complex urban areas.

This setup allows the vehicle's real-time processing power to focus on dynamic or unexpected elements (such as temporary construction, detours, or sudden obstacles) rather than having to re-calculate basic road geometry from scratch every single second.

“Thin chains” are not baked into maps.

Tesla has gotten better at these things with v15.

Literally nobody has claimed Tesla has not improved. All ADAS has improved. Time, and effort, causes things to improve.

“Tesla has improved” is not the win you think it is. So has everyone else, and Tesla had no choice. They couldn’t “not improve” unless they quit. That’s just literally how progress works.

https://twitter.com/x/status/2117778384923382075

By comparison, here's not preventing an obvious preventable accident even if technically not at fault.

There are videos of every single rideshare company (robo and not) making mistakes and getting into accidents.

Nobody thinks Waymo is perfect. It would appear at least one of us thinks Tesla is pretty close to perfect though, which is precisely the issue.

 

[borski]

That Waymo crash looks bad. If it had room to stop and didn’t, that’s a real failure and Waymo should own it. See how easy that was? I have no skin in the Waymo game, so admitting that cost me nothing. Unless you’re on Tesla’s payroll, admitting FSD has flaws should cost you nothing too. Your turn.

Precisely.

 

[borski]

What people don't realize is waymo does almost all of their "validation" of their vehicles in simulation. There is a sim 2 real gap where you can't get confidence that a problem is fixed until you test it in the real world.

This is where tesla does something no legit robotaxi can do. They can make changes to the neural network, push an update out every 3 days and get pretty clear confidence that an issue is fixed. Testing on real scenarios is always better than simulation.

There is another gap if you train on synthetic data, and if you look at waymo's synthetic lidar data they have demonstrated, it is very inaccurate and it's like genAI images where you see hands with too many fingers, etc.

We're talking about the 99.999%+ level which is what you really need to operate unsupervised vehicles

You are literally just making things up now. Waymo does not exclusively validate with synthetic data. They test in the real world literally every day.

I understand you find it fun and useful to regurgitate press release and twitter content, but outright making things up is new, even for you.

 

[borski]

Everything you see here is emotions and believing something that doesn't happen.

No, it isn’t. Some of us have provided evidence, based on data. You have not.

I do not live in a “post-truth” world. I require facts, evidence, and a basis in reality, along with enough humility to be able to accept when you’re incorrect, because the goal should be getting to and finding “truth,” not “proving” a preconceived bias or opinion.

You came here to try and “win.” That was where you went wrong.

Try curiosity instead of brash confidence. You might find it fun and enlightening.

Tesla literally has a dash cam (sentry mode) that records 100% of the time while you drive. These failures of FSD are posted online

Sometimes.

FSD is being used for billions of miles so you can tell when an issue still exists or not

“You can tell” is not a scientific statement. If you want to make this assertion, you need to provide evidence.

You won’t.

There was never evidence of FSD just blatantly running a red light or misidentifying a red light as green.

There was, in that it happened to Clarkson, and it is not a thing that has “never happened” to Tesla.

Flawless.

And before you try again, yes, Waymo also has issues sometimes. I make no claim they’re perfect.

In FSD 13 there was that rare issue where it would take off after sitting at a red light for about 10 seconds. It would do that at like 2am when nobody was around and not sure what caused it. That was fixed with v14

“Had bug, got fixed” seems to be a common refrain from you. The natural implication is “and thus there are now no bugs,” which is, of course, absolute and utter nonsense.

But without that implication, your statement is just “Tesla still exists and is building software” which is just… true for every company building ADAS.

The scientific evidence is literally there in the NHTSA SGO reports which look good in robotaxi favor after doing 6 million miles.

The evidence for what? The way a discussion works is: you make a claim, you provide evidence for that claim (or state it is an opinion), and then we can discuss it.

You have made a claim, and then stopped listening. That’s not a discussion. If you want a soapbox, this is not the place.

Then the goalposts will get moved again when tesla passes 100 million or 200 million miles. Tesla could do a billion miles and people will still say that camera only is not safe.

Nobody here has moved goalposts. We have simply been asking for any kind of evidence for the assertions you keep making. When we ask or point out holes, you have moved the goalposts from version to version, implying that because some things have been fixed, those holes are not valid; and thus, there must be no more holes.

There are. I promise. You don’t have to believe me, but at this point, this discussion is over and has run its course. It is no longer providing useful value to anyone on this forum.