1. So-and-so Tesla Youtuber/Blog/News/Influencer said that they pressed the accelerator/manipulated results which invalidates the results and shows the test are untrustworthy.
Let us examine that claim. Why would pressing the accelerator or other manipulations invalidate the results? Nobody would argue that turning on the windshield wipers or setting the cabin temperature would invalidate the results. What is special about the accelerator or other stated manipulations? The answer is that they could change the results which would invalidate a test. The actual claim the Tesla influencers are making is:
"The only reason the car would run down a child in that situation is because the accelerator was pressed/car was manipulated."
The video evidence for the last year has proven that is a bold-faced lie. Every claimed manipulation that can be caught on video has been thoroughly proven to be a outright fabrication. The car will run down children while FSD is enabled in simple circumstances when operated normally and has demonstrably done so for over a year.
The videos the Tesla influencer community point to are earlier videos that did not have sufficient clarity yet to disprove their lies so they thought they would not get caught. The newer videos with increased clarity prove them to be unrepentant liars who will fabricate whatever is convenient as long as they think they can not get caught. If any of the Tesla mouthpieces show up in the comments here to regurgitate the newest "manipulation" I will show what I mean.
2. The clearly visible child facing the road stepped out into the road at the last second, which children never do, so the laws of physics make it impossible for a car, that did not show caution and slow down when clearly visible children are by the road, to stop. So, the car had no choice but to run down the child and then continue for a few seconds to roll over the child while it is on the ground.
> The videos the Tesla influencer community point to are earlier videos that did not have sufficient clarity yet to disprove their lies so they thought they would not get caught
Haven't kept up with the drama in a while, but to my understanding the issue people took with the Dawn Project's initial video was not that it lacked clarity - but that it had the clarity to see that FSD was not engaged.[0]
Could be explained, such as if the Dawn Project did the tests legitimately and then went back to capture footage for the ad, but I wouldn't assign so much malice ("unrepentant liars who will fabricate whatever is convenient as long as they think they can not get caught") to people pointing it out.
Except that it was engaged as stated in the Safety Test Report [1] and as seen in the raw footage [2]. Rather than acknowledging that a incomplete investigation with conflicting information demands waiting for confirmatory evidence such as the raw footage which would have demonstrated their assumptions on the validity of the test were false, they chose to immediately begin declaring it a smear campaign based on false tests.
They are not liars because they point out FSD was not engaged in portions of the ad. They are liars because they then straight up ignored evidence to the contrary and then assert, in confidence, that the tests were certainly fabricated and that FSD would never operate as demonstrated even though the video evidence that was forthcoming proved their assertions were completely and utterly wrong.
As to being unrepentant liars. Is a legitimate test demonstrating a critical safety defect in a consumer product a smear campaign? No. Have they retracted their false accusation that correctly exposing a critical safety defect is a smear campaign since they jumped the gun on their accusations before they did even a token investigation? No. Then it is a lie and they are unrepentant.
As to being willing to fabricate whatever is convenient. As soon as it became clear that FSD was enabled during those tests, the party line became that the only reason it happened was because the accelerator was pressed. This is false as can be seen in the current video. Then it became the only reason it happened was because the cones prevented a swerve. This is false as can be seen in the current video. Then it became the only reason it happened was because they steered into the child. This is false as can be seen in the current video. Every single one of these accusations was made because they thought the video could not disprove it and every single one has been proven to be false once the video setup was improved so it could disprove them.
The fact that people, to this day, still deliberately point back to the earliest videos because they can not find any way to lie about the current videos without getting caught is bad faith argumentation. To then argue the Dawn Project is untrustworthy because of false accusations made by liars is true grade A bad faith argumentation that needs to be nipped in the bud.
> Except that it was engaged as stated in the Safety Test Report [1] and as seen in the raw footage [2].
> They are not liars because they point out FSD was not engaged in portions of the ad. They are liars because they then straight up ignored evidence to the contrary
I'm not seeing evidence in video [2] contrary to the idea that FSD was not engaged in a portion of the original ad presented as being in FSD. Unless you have a timestamp that I missed, it doesn't seem to contain the relevant segment at all.
I think the most likely explanation for FSD not being enabled in the ad is that the Dawn Project did the tests properly (albeit probably with some cherry-picking) then went back to film some more dramatic shots and accidentally messed up enabling FSD on one of them. I'm not sure what to make of the fact that, as Electrek points out, the speeds in video [2] don't match what they wrote in the report. Do you know if the Dawn Project acknowledged/explained either of these?
> As to being unrepentant liars. Is a legitimate test demonstrating a critical safety defect in a consumer product a smear campaign? No. Have they retracted their false accusation that correctly exposing a critical safety defect is a smear campaign since they jumped the gun on their accusations before they did even a token investigation? No. Then it is a lie and they are unrepentant.
On balance of probabilities I don't think the Dawn Project faked the tests, but I can see how the discrepancies (and some motivated reasoning) would lead someone to legitimately believe that they did.
To clarify, I was stating it was engaged during the tests used to inform the test report and the campaign. The claim in the article is that "they never engaged FSD Beta during the test". That is demonstrably false.
You could state: "FSD Beta was not engaged during portions of the ad footage". That is well supported. However, Fred Lambert read the Safety Test Report [1], as he pointed to specific sections, but was apparently too busy to view the raw footage linked on page 5 (unfortunately that link is now dead, but did point to the same footage I linked previously) or accept affidavits sworn under penalty of perjury before affirmatively declaring that FSD was not enabled during the referenced tests. He did not even make minimal effort to ascertain the actual truth before spouting vile and false accusations.
To further add on to his point, he fabricated nonsensical justification for why it must have been disabled like: "The explanation is likely quite simple. Tesla FSD Beta still relies on map data, and this test was performed on a closed course at Willow Springs International Raceway in Rosamond, California."
As a "journalist", he made no effort to ascertain the actual truth before promulgating a story. The story was venomous, throwing around the most inflammatory and sensationalist accusations possible such as claiming it was a "smear campaign" and arguing that they were either intentionally lying about the system being enabled or too stupid to do so and thus should be completely ignored. The consequences of his "journalism" are that his false accusations continue to be successfully used as ammunition to quash whistleblowing and he has made no effort to remedy his errors, making only a limp-wristed reluctant, non-retraction when confronted with the magnitude of his error that is insufficient to dispel the clear disinformation he presented as truth.
As to the data in the safety report, it is corrected in the report here [2] which has live links to the raw footage. The discrepancies from the original report are immaterial as they are only slight differences in child collision speed, there exists no qualitative error, and the footage speaks for itself as a authoritative source.
In any event, you can just look to the more recent tests which completely invalidate these questions anyways. The system is clearly engaged, there is no driver input, the test environment is simple, and the error is egregious and reproducible.
> or accept affidavits sworn under penalty of perjury before affirmatively declaring that FSD was not enabled during the referenced tests. He did not even make minimal effort to ascertain the actual truth before spouting vile and false accusations.
The driver's sworn word doesn't really contradict the theory that they failed to realize they didn't engage FSD - supported by the footage where the driver attempts to engage FSD, and acts as if it has been activated, but actually failed to do so.
Lambert says his theory is "based on their own footage in the ad", which I think is fair. The theory turns out to be wrong because the ad presents footage as part of the test that was actually recorded as a re-enactment afterwards (or at least, I believe it does - haven't seen the Dawn Project admit to this). It's understandable that they did so to get better shots for the ad and I don't think there was any malice or intentional misrepresentation, but the driver's mistake in this extra footage is the root of the error.
> As to the data in the safety report, it is corrected in the report here [2] which has live links to the raw footage. The discrepancies from the original report are immaterial as they are only slight differences in child collision speed
24mph -> 17mph is a notable discrepancy when the exact speed is visible in the released recording on the dashboard. It seems to be silently altered to match the released video, as opposed to explained, as far as I can see. But may just be a case of "we took the driver's estimate rather than checking the footage" or "we paused the video very inaccurately, sometimes before and sometimes after the collision".
> In any event, you can just look to the more recent tests which completely invalidate these questions anyways
If someone believes the Dawn Project are manipulating the tests, the release of new different footage where they're careful not to make the same mistake they were called out for previously doesn't really affect anything. An explanation of why there were discrepancies in the previous footage, with supporting evidence if possible, may help.
I believe the existence of the unexplained (but not unexplainable) discrepancies is why "people, to this day, still deliberately point back to the earliest videos".
Lambert is a "journalist". It is unacceptable for him to willfully ignore easily accessible evidence and loudly and aggressively smear whistleblowers on a hunch to protect the company that owes him 500,000$ worth of cars [1][2]. That you linked to his objectively wrong article shows Fred Lambert's smear campaign worked and continues to work to this day.
24 mph -> 17 mph is not a notable discrepancy. The car should not ever hit a child in that situation at a speed that sends them flying, period. The video is clear, but the old report has inaccuracies that were fixed in the new report is not grounds for dismissing the critical safety defect of running down children. That people laser focus on about 7 mph child collision speed differences so they can distract from the nearly trillion dollar company releasing a "full self driving" product that could not detect stationary children until somebody ran a national PSA about it is insane.
Of course the new tests invalidate their false accusations. As I said in my initial post, a claim of manipulation is only valid if the manipulation would change the outcome of the test. A clear test demonstrating the claimed manipulation is not present is evidence that their accusation that manipulation is the only way to achieve the stated outcome is false. Further evidence is that all major people in the Tesla influencer community has been invited to observe, participate, and even use their own cars in the tests and every single one of them has chickened out.
Anybody pointing to old, immature tests is just propagating a mindworm meant to sound persuasive when it is actually a completely moronic argument. Arguing that improvements in the camera setup and test transparency are some sort of tachyonic evidence of malfeasance is asinine. The only way you would think the current videos are manipulated is if you come in with the assumption that they are manipulated. That is why the Tesla community always makes sure to point back to the earliest videos because they will get caught if they try to lie about anything new, so they need to lie about the past when the cameras were not good enough to catch their hands in the cookie jar. Yet, for some reason, nobody applies that same level of scrutiny to the nearly trillion dollar company that has killed and injured hundreds.
> That you linked to his objectively wrong article shows Fred Lambert's smear campaign worked and continues to work to this day.
Say I claim to have tight-rope walked between two buildings. You notice an error in my video that suggests I didn't - maybe an incorrectly-keyed green-screen on one of the clips claimed to be the actual walk. I think it's fair for you to say "based on the footage in your promo video, you did not tight-rope walk between two buildings".
Maybe it turns out I did, and the shot was actually filmed afterwards just for the promo video. In this case your theory was wrong, but the error stems from my own (non-malicious, but faulty) misrepresentation of the clip as something it wasn't.
> 24 mph -> 17 mph is not a notable discrepancy. The car should not ever hit a child in that situation at a speed that sends them flying, period.
I'm very obviously not saying that it is okay to hit a child with a car, regardless of speed. I'm saying that the released clips do not align with the stated results of the test, with the numbers being different enough that it's not within regular measurement error.
Some take that to mean the clips do not show the original tests. I believe instead that the Dawn Project messed up repeatedly on recording the speeds in some way - but I don't think the Dawn Project have admitted as such.
> Of course the new tests invalidate their false accusations
> [...]
> That is why the Tesla community always makes sure to point back to the earliest videos
Returning to the tight-rope example: releasing new correctly-keyed footage, absent an explanation of why the previous footage contained the error, does not invalidate someone's reason for concern about me having manipulated the footage. If I did indeed manipulate the original footage, I would very likely avoid visibly making the same mistake that I was called out for in future footage. For someone to debunk my claims, I don't think it'd be invalid for them to point to earlier footage containing unexplained errors from before I got sneakier.
Say I claim to have tight-rope walked between buildings. In my official report I have a continuous video clearly showing that. Simultaneously, I release a promo video that the report does not reference.
You watch the promo video and then read the report, but then choose to not watch the actual video of the tests that are literally in the report you just read. You then smear whistleblowers and do not retract your statements when you are proven to be wrong.
And you know what, if I did smear whistleblowers, I would likely avoid visibly smearing in the same way in future footage that would be able to call out my lies. For someone to debunk my smears, I don’t think it would be invalid to point to later footage and show how my claimed manipulations are demonstrably absent and unnecessary for it to run down children now that the videos are clearer and more transparent.
It is frankly kind of ridiculous that you are applying a significantly higher standard toward whistleblowing about a dangerous and defective product rather than the product itself and the people who are known cheerleaders and shills. That both they and the manufacturer have actively rejected every invitation to reproduce the results with their own cars should make it clear who is scared of the burden of proof here.
> Simultaneously, I release a promo video that the report does not reference. You watch the promo video and then read the report, but then choose to not watch the actual video of the tests that are literally in the report you just read.
As far as I can tell:
* 2022-08-09 13:50PT (timestamp on video) - The ad is released, containing no mention or link to a report
* Lambert contacts the Dawn Project for comment and receives the affidavit and screen photograph
* 2022-08-10 08:56PT (timestamp on article) - Lambert's article is released. It uses press images provided by the Dawn Project ([0], [1]), but not anything directly taken by Lambert from the report, which I do not believe he has seen at this point
* Between 2022-08-09 20:43PT and 2022-08-10 14:27PT (closest Wayback machine snapshots with and without the campaign[2]) - the Dawn Project added the campaign to their website, including the link to the report
* Later in the day, Lambert updates the article addressing the newly released footage
I could well be missing something, but I don't see evidence that Lambert had read the report at the time of writing the initial article as you claim.
> You then smear whistleblowers [...]
> [...] you are applying a significantly higher standard toward whistleblowing about a dangerous and defective product rather than the product itself [...]
The Dawn Project is really Dan O'Dowd, a self-described "billionaire entrepreneur" whose company provides for competing automated driving systems[3]. According to his website[4], he's also:
* "Your Source for Software that Never Fails and Can't Be Hacked"
* The "World's leading authority on cybersecurity", "World's leading authority on software safety", "World's leading authority on software reliability", and "World's leading authority on software performance"
* Designer of "the only operating system that never fails and can't be hacked"
* The one who "rescued the Macintosh and Apple Computers at Steve Jobs' request"
* "I can make any computer system unhackable", "I can program any system so it will never fail"
* "I can find every bug in any program I write", "The last time I couldn’t find a bug, it was 1975"
* "I can make your software run up to 4 times faster than anyone else. No one else even claims to be an expert in software performance."
Not to turn this into just ad hominem about O'Dowd's quackery, but he shouldn't be presented as some legitimate impartial organisation or sincere whistleblower - it's one rich schmuck with a business interest and vendetta against another rich schmuck. That doesn't mean O'Dowd's interests can't happen to align with the public's, or that he can't perform valid tests, but I do think a large dose of scrutiny for his claims is appropriate.
An Internet debate/investigation of any kind is a far lower standard than the rigorous testing Tesla's "FSD" should be subject to - don't take my comments as arguing that should be lowered.
> [...] and the people who are known cheerleaders and shills. That both they and the manufacturer have actively rejected every invitation to reproduce the results with their own cars should make it clear who is scared of the burden of proof here.
Is there some specific invitation from O'Dowd you're referring to? I can see plenty of people testing it, even including a couple testing on real children[5] which is horrendous.
No, it is just bad at detecting small humans. Given that Tesla only uses monocular vision and thus does not have binocular depth perception, it is possible that it determines that children are just far away adults, though that is just speculation on my part. It is Tesla that should affirmatively determine how their system acts in the presence of children before releasing it to their in-house testers, let alone before they give it to their customers so they can book revenue.
It definitely sees it in the later ones. I’m wondering if the confidence is too low to trigger the system, to prevent a phantom brake, maybe from the “child” moving in an impossible way (statically sliding across the ground). I would assume this is similar to the crash test setup though. It would be really interesting to get a deep dive into the cause.
It probably can't be 'fixed' - vision only + 'end-to-end NN'. I don't believe you can get it to be really reliable with that setup, at least not in this decade.
It's a very good setup for awesome demo of the technology and vague promises in the future. Just like in 2016 (every Tesla will become a Robotaxi with an OTA update) and 2019 (we will have 1 million Robotaxis on the road in 2020).
Is it a lie or not ? Even the usual influencers are saying it's not taking pedestrian gestures into account, based on their tests. But, the system is an NN - so it's impossible to prove it - just like it's impossible to prove that it does take it into account. Hell, it takes everything it sees on the camera into account !
Those are tests of automatic emergency braking (AEB) where advanced driver assist systems (ADAS) such as Tesla FSD are explicitly not enabled.
In actual ADAS tests, such as the recent IIHS Partial Automation Safeguard Ratings [1] Tesla FSD scores the lowest (6 Poor, 2 Acceptable) and Tesla Autopilot scores the second lowest (6 Poor, 1 Acceptable, 1 Good) out of 14 evaluated systems with the third lowest being the Genesis Highway Driving Assist 2 (6 Poor, 1 Marginal, 2 Good) and the highest being Lexus Teammate (1 Marginal, 3 Acceptable, 4 Good) with the average (to the extent that means anything) being around (3 Poor, 1 Marginal, 1 Acceptable, 3 Good).
Or if we wish to look at the Euro NCAP Assisted Driving Gradings [2] Tesla scored 36/100 in assistance competence, the lowest score by far out of 26 evaluated systems. The second lowest is 52/100 and the average over all models excluding Tesla being ~67/100.
Directly from your Euro NCAP link on the assisted driving test:
>The Model 3 responds to avoid a collision in all the ACC test scenarios and requires AEB interventions only in the most critical cut-in and
cut-out tests. The driver is supported through the S-Bend, staying centred in the lane at all test speeds. The vehicle has an Active
Blindspot system designed to prevent lane changing into adjacent vehicles. A lane-change assist function is also provided. In case of
an unresponsive driver, the Tesla performs a controlled stop in lane. If the radar or camera are blocked the Model 3 provides a timely
warning and prevents system activation.
>The Tesla Model 3 excels in the level of Vehicle Assistance but fails to balance that high level of support with a similar level of
Driver Engagement leading to possible overreliance. Euro NCAP tested the latest version of AutoPilot available at the time of
testing. Tesla can improve the functionality of the system remotely by over-the-air software updates.
The autopilot system itself got very high points, the 36% is from driver engagement features. Driver Engagement / 36%. Pretty disingenuous to use this driver engagement score which is measuring things like the driver monitoring system and not the autopilot performance.
No, 36% is the overall Assistance Competence score as ADAS systems demand driver engagement to be operated safely. In case you did not know, Tesla ADAS systems must only be used with a fully attentive driver and require you to "keep your hands on the steering wheel at all times [1]". Failure to ensure driver attention is how you kills tens to hundreds of people as Tesla admits is the cause for the numerous deaths while using their system.
But you know what is really disingenuous? Pointing at IIHS tests that do not even have Autopilot or FSD enabled and using that to justify selling a system that will run down children in basic circumstances.
That is false. They make operating systems and development tools for embedded applications such as aerospace and automotive. Some automotive companies may use the tools to make advanced driver assistance systems (ADAS) or automated driving systems (ADS), but that is as much a competitor with Tesla as GM's hammer supplier is a competitor with Tesla.
> They make operating systems and development tools for embedded applications such as aerospace [...]
Specifically a "Platform for Safe and Secure Automated Driving Systems". Could argue it's still not a FSD competitor in its own right, but it's more than just the implied "they make developement tools, which happen to be used by some automotive companies, which happen to make ADAS".
28 comments
[ 3.2 ms ] story [ 78.4 ms ] thread1. So-and-so Tesla Youtuber/Blog/News/Influencer said that they pressed the accelerator/manipulated results which invalidates the results and shows the test are untrustworthy.
Let us examine that claim. Why would pressing the accelerator or other manipulations invalidate the results? Nobody would argue that turning on the windshield wipers or setting the cabin temperature would invalidate the results. What is special about the accelerator or other stated manipulations? The answer is that they could change the results which would invalidate a test. The actual claim the Tesla influencers are making is:
"The only reason the car would run down a child in that situation is because the accelerator was pressed/car was manipulated."
The video evidence for the last year has proven that is a bold-faced lie. Every claimed manipulation that can be caught on video has been thoroughly proven to be a outright fabrication. The car will run down children while FSD is enabled in simple circumstances when operated normally and has demonstrably done so for over a year.
The videos the Tesla influencer community point to are earlier videos that did not have sufficient clarity yet to disprove their lies so they thought they would not get caught. The newer videos with increased clarity prove them to be unrepentant liars who will fabricate whatever is convenient as long as they think they can not get caught. If any of the Tesla mouthpieces show up in the comments here to regurgitate the newest "manipulation" I will show what I mean.
2. The clearly visible child facing the road stepped out into the road at the last second, which children never do, so the laws of physics make it impossible for a car, that did not show caution and slow down when clearly visible children are by the road, to stop. So, the car had no choice but to run down the child and then continue for a few seconds to roll over the child while it is on the ground.
I think I will let that point speak for itself.
Haven't kept up with the drama in a while, but to my understanding the issue people took with the Dawn Project's initial video was not that it lacked clarity - but that it had the clarity to see that FSD was not engaged.[0]
Could be explained, such as if the Dawn Project did the tests legitimately and then went back to capture footage for the ad, but I wouldn't assign so much malice ("unrepentant liars who will fabricate whatever is convenient as long as they think they can not get caught") to people pointing it out.
[0]: https://electrek.co/2022/08/10/tesla-self-driving-smear-camp...
They are not liars because they point out FSD was not engaged in portions of the ad. They are liars because they then straight up ignored evidence to the contrary and then assert, in confidence, that the tests were certainly fabricated and that FSD would never operate as demonstrated even though the video evidence that was forthcoming proved their assertions were completely and utterly wrong.
As to being unrepentant liars. Is a legitimate test demonstrating a critical safety defect in a consumer product a smear campaign? No. Have they retracted their false accusation that correctly exposing a critical safety defect is a smear campaign since they jumped the gun on their accusations before they did even a token investigation? No. Then it is a lie and they are unrepentant.
As to being willing to fabricate whatever is convenient. As soon as it became clear that FSD was enabled during those tests, the party line became that the only reason it happened was because the accelerator was pressed. This is false as can be seen in the current video. Then it became the only reason it happened was because the cones prevented a swerve. This is false as can be seen in the current video. Then it became the only reason it happened was because they steered into the child. This is false as can be seen in the current video. Every single one of these accusations was made because they thought the video could not disprove it and every single one has been proven to be false once the video setup was improved so it could disprove them.
The fact that people, to this day, still deliberately point back to the earliest videos because they can not find any way to lie about the current videos without getting caught is bad faith argumentation. To then argue the Dawn Project is untrustworthy because of false accusations made by liars is true grade A bad faith argumentation that needs to be nipped in the bud.
[1] https://dawnproject.com/wp-content/uploads/2022/08/The_Dawn_...
[2] https://dawnproject.com/wp-content/uploads/2022/08/raw-foota...
> They are not liars because they point out FSD was not engaged in portions of the ad. They are liars because they then straight up ignored evidence to the contrary
I'm not seeing evidence in video [2] contrary to the idea that FSD was not engaged in a portion of the original ad presented as being in FSD. Unless you have a timestamp that I missed, it doesn't seem to contain the relevant segment at all.
I think the most likely explanation for FSD not being enabled in the ad is that the Dawn Project did the tests properly (albeit probably with some cherry-picking) then went back to film some more dramatic shots and accidentally messed up enabling FSD on one of them. I'm not sure what to make of the fact that, as Electrek points out, the speeds in video [2] don't match what they wrote in the report. Do you know if the Dawn Project acknowledged/explained either of these?
> As to being unrepentant liars. Is a legitimate test demonstrating a critical safety defect in a consumer product a smear campaign? No. Have they retracted their false accusation that correctly exposing a critical safety defect is a smear campaign since they jumped the gun on their accusations before they did even a token investigation? No. Then it is a lie and they are unrepentant.
On balance of probabilities I don't think the Dawn Project faked the tests, but I can see how the discrepancies (and some motivated reasoning) would lead someone to legitimately believe that they did.
You could state: "FSD Beta was not engaged during portions of the ad footage". That is well supported. However, Fred Lambert read the Safety Test Report [1], as he pointed to specific sections, but was apparently too busy to view the raw footage linked on page 5 (unfortunately that link is now dead, but did point to the same footage I linked previously) or accept affidavits sworn under penalty of perjury before affirmatively declaring that FSD was not enabled during the referenced tests. He did not even make minimal effort to ascertain the actual truth before spouting vile and false accusations.
To further add on to his point, he fabricated nonsensical justification for why it must have been disabled like: "The explanation is likely quite simple. Tesla FSD Beta still relies on map data, and this test was performed on a closed course at Willow Springs International Raceway in Rosamond, California."
As a "journalist", he made no effort to ascertain the actual truth before promulgating a story. The story was venomous, throwing around the most inflammatory and sensationalist accusations possible such as claiming it was a "smear campaign" and arguing that they were either intentionally lying about the system being enabled or too stupid to do so and thus should be completely ignored. The consequences of his "journalism" are that his false accusations continue to be successfully used as ammunition to quash whistleblowing and he has made no effort to remedy his errors, making only a limp-wristed reluctant, non-retraction when confronted with the magnitude of his error that is insufficient to dispel the clear disinformation he presented as truth.
As to the data in the safety report, it is corrected in the report here [2] which has live links to the raw footage. The discrepancies from the original report are immaterial as they are only slight differences in child collision speed, there exists no qualitative error, and the footage speaks for itself as a authoritative source.
In any event, you can just look to the more recent tests which completely invalidate these questions anyways. The system is clearly engaged, there is no driver input, the test environment is simple, and the error is egregious and reproducible.
[1] https://dawnproject.com/wp-content/uploads/2022/08/The_Dawn_...
[2] https://dawnproject.com/the-dawn-projects-new-advertising-ca...
The driver's sworn word doesn't really contradict the theory that they failed to realize they didn't engage FSD - supported by the footage where the driver attempts to engage FSD, and acts as if it has been activated, but actually failed to do so.
Lambert says his theory is "based on their own footage in the ad", which I think is fair. The theory turns out to be wrong because the ad presents footage as part of the test that was actually recorded as a re-enactment afterwards (or at least, I believe it does - haven't seen the Dawn Project admit to this). It's understandable that they did so to get better shots for the ad and I don't think there was any malice or intentional misrepresentation, but the driver's mistake in this extra footage is the root of the error.
> As to the data in the safety report, it is corrected in the report here [2] which has live links to the raw footage. The discrepancies from the original report are immaterial as they are only slight differences in child collision speed
24mph -> 17mph is a notable discrepancy when the exact speed is visible in the released recording on the dashboard. It seems to be silently altered to match the released video, as opposed to explained, as far as I can see. But may just be a case of "we took the driver's estimate rather than checking the footage" or "we paused the video very inaccurately, sometimes before and sometimes after the collision".
> In any event, you can just look to the more recent tests which completely invalidate these questions anyways
If someone believes the Dawn Project are manipulating the tests, the release of new different footage where they're careful not to make the same mistake they were called out for previously doesn't really affect anything. An explanation of why there were discrepancies in the previous footage, with supporting evidence if possible, may help.
I believe the existence of the unexplained (but not unexplainable) discrepancies is why "people, to this day, still deliberately point back to the earliest videos".
24 mph -> 17 mph is not a notable discrepancy. The car should not ever hit a child in that situation at a speed that sends them flying, period. The video is clear, but the old report has inaccuracies that were fixed in the new report is not grounds for dismissing the critical safety defect of running down children. That people laser focus on about 7 mph child collision speed differences so they can distract from the nearly trillion dollar company releasing a "full self driving" product that could not detect stationary children until somebody ran a national PSA about it is insane.
Of course the new tests invalidate their false accusations. As I said in my initial post, a claim of manipulation is only valid if the manipulation would change the outcome of the test. A clear test demonstrating the claimed manipulation is not present is evidence that their accusation that manipulation is the only way to achieve the stated outcome is false. Further evidence is that all major people in the Tesla influencer community has been invited to observe, participate, and even use their own cars in the tests and every single one of them has chickened out.
Anybody pointing to old, immature tests is just propagating a mindworm meant to sound persuasive when it is actually a completely moronic argument. Arguing that improvements in the camera setup and test transparency are some sort of tachyonic evidence of malfeasance is asinine. The only way you would think the current videos are manipulated is if you come in with the assumption that they are manipulated. That is why the Tesla community always makes sure to point back to the earliest videos because they will get caught if they try to lie about anything new, so they need to lie about the past when the cameras were not good enough to catch their hands in the cookie jar. Yet, for some reason, nobody applies that same level of scrutiny to the nearly trillion dollar company that has killed and injured hundreds.
[1] https://www.thedrive.com/news/24025/electreks-editor-in-chie...
[2] https://old.reddit.com/r/teslamotors/comments/b0bljo/how_muc...
Say I claim to have tight-rope walked between two buildings. You notice an error in my video that suggests I didn't - maybe an incorrectly-keyed green-screen on one of the clips claimed to be the actual walk. I think it's fair for you to say "based on the footage in your promo video, you did not tight-rope walk between two buildings".
Maybe it turns out I did, and the shot was actually filmed afterwards just for the promo video. In this case your theory was wrong, but the error stems from my own (non-malicious, but faulty) misrepresentation of the clip as something it wasn't.
> 24 mph -> 17 mph is not a notable discrepancy. The car should not ever hit a child in that situation at a speed that sends them flying, period.
I'm very obviously not saying that it is okay to hit a child with a car, regardless of speed. I'm saying that the released clips do not align with the stated results of the test, with the numbers being different enough that it's not within regular measurement error.
Some take that to mean the clips do not show the original tests. I believe instead that the Dawn Project messed up repeatedly on recording the speeds in some way - but I don't think the Dawn Project have admitted as such.
> Of course the new tests invalidate their false accusations
> [...]
> That is why the Tesla community always makes sure to point back to the earliest videos
Returning to the tight-rope example: releasing new correctly-keyed footage, absent an explanation of why the previous footage contained the error, does not invalidate someone's reason for concern about me having manipulated the footage. If I did indeed manipulate the original footage, I would very likely avoid visibly making the same mistake that I was called out for in future footage. For someone to debunk my claims, I don't think it'd be invalid for them to point to earlier footage containing unexplained errors from before I got sneakier.
You watch the promo video and then read the report, but then choose to not watch the actual video of the tests that are literally in the report you just read. You then smear whistleblowers and do not retract your statements when you are proven to be wrong.
And you know what, if I did smear whistleblowers, I would likely avoid visibly smearing in the same way in future footage that would be able to call out my lies. For someone to debunk my smears, I don’t think it would be invalid to point to later footage and show how my claimed manipulations are demonstrably absent and unnecessary for it to run down children now that the videos are clearer and more transparent.
It is frankly kind of ridiculous that you are applying a significantly higher standard toward whistleblowing about a dangerous and defective product rather than the product itself and the people who are known cheerleaders and shills. That both they and the manufacturer have actively rejected every invitation to reproduce the results with their own cars should make it clear who is scared of the burden of proof here.
As far as I can tell:
* 2022-08-09 13:50PT (timestamp on video) - The ad is released, containing no mention or link to a report
* Lambert contacts the Dawn Project for comment and receives the affidavit and screen photograph
* 2022-08-10 08:56PT (timestamp on article) - Lambert's article is released. It uses press images provided by the Dawn Project ([0], [1]), but not anything directly taken by Lambert from the report, which I do not believe he has seen at this point
* Between 2022-08-09 20:43PT and 2022-08-10 14:27PT (closest Wayback machine snapshots with and without the campaign[2]) - the Dawn Project added the campaign to their website, including the link to the report
* Later in the day, Lambert updates the article addressing the newly released footage
I could well be missing something, but I don't see evidence that Lambert had read the report at the time of writing the initial article as you claim.
> You then smear whistleblowers [...]
> [...] you are applying a significantly higher standard toward whistleblowing about a dangerous and defective product rather than the product itself [...]
The Dawn Project is really Dan O'Dowd, a self-described "billionaire entrepreneur" whose company provides for competing automated driving systems[3]. According to his website[4], he's also:
* "Your Source for Software that Never Fails and Can't Be Hacked"
* The "World's leading authority on cybersecurity", "World's leading authority on software safety", "World's leading authority on software reliability", and "World's leading authority on software performance"
* Designer of "the only operating system that never fails and can't be hacked"
* The one who "rescued the Macintosh and Apple Computers at Steve Jobs' request"
* "I can make any computer system unhackable", "I can program any system so it will never fail"
* "I can find every bug in any program I write", "The last time I couldn’t find a bug, it was 1975"
* "I can make your software run up to 4 times faster than anyone else. No one else even claims to be an expert in software performance."
Not to turn this into just ad hominem about O'Dowd's quackery, but he shouldn't be presented as some legitimate impartial organisation or sincere whistleblower - it's one rich schmuck with a business interest and vendetta against another rich schmuck. That doesn't mean O'Dowd's interests can't happen to align with the public's, or that he can't perform valid tests, but I do think a large dose of scrutiny for his claims is appropriate.
An Internet debate/investigation of any kind is a far lower standard than the rigorous testing Tesla's "FSD" should be subject to - don't take my comments as arguing that should be lowered.
> [...] and the people who are known cheerleaders and shills. That both they and the manufacturer have actively rejected every invitation to reproduce the results with their own cars should make it clear who is scared of the burden of proof here.
Is there some specific invitation from O'Dowd you're referring to? I can see plenty of people testing it, even including a couple testing on real children[5] which is horrendous.
[0]: https://dawnproject.com/wp-content/uploads/2022/08/Willow-Sp...
[1]: https:/&#...
I wonder whether they will try to detect the test setup though and disable FSD, or pull a VW.
It's a very good setup for awesome demo of the technology and vague promises in the future. Just like in 2016 (every Tesla will become a Robotaxi with an OTA update) and 2019 (we will have 1 million Robotaxis on the road in 2020).
Another example - Tesla official account highlighting that FSD is 'taking pedestrian gestures into account': https://nitter.poast.org/Tesla/status/1776021631346835491#m
Is it a lie or not ? Even the usual influencers are saying it's not taking pedestrian gestures into account, based on their tests. But, the system is an NN - so it's impossible to prove it - just like it's impossible to prove that it does take it into account. Hell, it takes everything it sees on the camera into account !
Home-made tests are pretty easy to game, there could have been hundreds of attempts and these are the only that failed.
In actual ADAS tests, such as the recent IIHS Partial Automation Safeguard Ratings [1] Tesla FSD scores the lowest (6 Poor, 2 Acceptable) and Tesla Autopilot scores the second lowest (6 Poor, 1 Acceptable, 1 Good) out of 14 evaluated systems with the third lowest being the Genesis Highway Driving Assist 2 (6 Poor, 1 Marginal, 2 Good) and the highest being Lexus Teammate (1 Marginal, 3 Acceptable, 4 Good) with the average (to the extent that means anything) being around (3 Poor, 1 Marginal, 1 Acceptable, 3 Good).
Or if we wish to look at the Euro NCAP Assisted Driving Gradings [2] Tesla scored 36/100 in assistance competence, the lowest score by far out of 26 evaluated systems. The second lowest is 52/100 and the average over all models excluding Tesla being ~67/100.
[1] https://www.iihs.org/ratings/partial-automation-safeguards
[2] https://www.euroncap.com/en/ratings-rewards/assisted-driving...
>The Model 3 responds to avoid a collision in all the ACC test scenarios and requires AEB interventions only in the most critical cut-in and cut-out tests. The driver is supported through the S-Bend, staying centred in the lane at all test speeds. The vehicle has an Active Blindspot system designed to prevent lane changing into adjacent vehicles. A lane-change assist function is also provided. In case of an unresponsive driver, the Tesla performs a controlled stop in lane. If the radar or camera are blocked the Model 3 provides a timely warning and prevents system activation.
>The Tesla Model 3 excels in the level of Vehicle Assistance but fails to balance that high level of support with a similar level of Driver Engagement leading to possible overreliance. Euro NCAP tested the latest version of AutoPilot available at the time of testing. Tesla can improve the functionality of the system remotely by over-the-air software updates.
The autopilot system itself got very high points, the 36% is from driver engagement features. Driver Engagement / 36%. Pretty disingenuous to use this driver engagement score which is measuring things like the driver monitoring system and not the autopilot performance.
But you know what is really disingenuous? Pointing at IIHS tests that do not even have Autopilot or FSD enabled and using that to justify selling a system that will run down children in basic circumstances.
[1] https://www.tesla.com/support/autopilot
Specifically a "Platform for Safe and Secure Automated Driving Systems". Could argue it's still not a FSD competitor in its own right, but it's more than just the implied "they make developement tools, which happen to be used by some automotive companies, which happen to make ADAS".