Octopedia Elon’s billion-dollar blind spot: Why Tesla’s war on LiDAR is automotive hubris
In April 2019, Elon Musk stood before investors at Tesla’s Autonomy Day and delivered what may go down as one of the most spectacularly wrong predictions in automotive history. “LiDAR is a fool’s errand,” he declared with characteristic bravado. “Anyone relying on LiDAR is doomed. Doomed!”¹ He punctuated this proclamation by comparing the sensors to “expensive appendices” - unnecessary organs that evolution should have discarded. At the time, he argued that self-driving cars should rely solely on cameras and computer vision, because after all, humans drive with their eyes. Why shouldn’t machines be able to do the same?
It was a bold statement, and very on-brand. But it also revealed a fundamental misunderstanding of both biology and engineering.
The irony, of course, is that while Musk was publicly denouncing LiDAR as technological dead weight, Tesla was secretly purchasing over $2.1 million worth of LiDAR sensors from Luminar Technologies², becoming their largest customer. But who needs consistency when you have confidence?
The seductive myth of “pure vision”
The idea that you can build a safe navigation system using only vision is as naïve as it is seductive. Yes, humans rely heavily on sight, but our visual system is not an isolated camera feed. It is calibrated and cross-checked by a web of other senses: proprioception, vestibular balance, tactile feedback, even auditory cues.
Strip those out, and vision alone becomes fragile. Anyone who has experienced vertigo knows this. When the inner ear misfires, your eyes suddenly become unreliable, and your brain struggles to maintain orientation.
It’s funny when people talk about “pure vision systems” as though any animal actually navigates that way. In reality, vision has never been a standalone sense. For animals in motion, visual input is constantly supervised, corrected, and stabilized by other sensory systems. Humans can’t even walk straight if we lose proprioception or inner-ear balance, and we only know where our eyes are pointing because our brains integrate head orientation, vestibular feedback, and touch.
So when engineers slap a camera on a car and declare vision to be “enough,” they are effectively trying to reinvent a biological system while discarding the very sensory integration that makes it robust. Evolution solved this problem over hundreds of millions of years. Silicon Valley thinks it can do it with a few billion frames of training data.
This contradiction perfectly encapsulates Tesla’s approach to autonomous driving: bold public declarations masking a fundamentally flawed understanding of safety-critical engineering. While Waymo has quietly accumulated 100 million fully driverless miles³ with an 85% reduction in injury-causing crashes compared to human drivers⁴, Tesla’s “Full Self-Driving” remains perpetually supervised, accumulating crashes at an alarming rate - 467 Autopilot-involved crashes investigated by NHTSA since 2018, resulting in 14 deaths and 54 injuries⁵. Yet somehow, the man who compared LiDAR to vestigial organs continues to insist that cameras alone can solve autonomous driving, as if the laws of physics bend to the will of Silicon Valley disruption.
The tragedy isn’t just that Musk is wrong - it’s that he’s wrong in precisely the same way engineers have been catastrophically wrong before. The Tacoma Narrows Bridge collapsed because designers relied on successful patterns from shorter bridges without understanding aerodynamic flutter⁶. NASA normalized O-ring erosion patterns until Challenger exploded, with one official declaring that “statistics don’t count for anything… they have no place in engineering anywhere”⁷. Boeing’s 737 MAX killed 346 people because MCAS relied on a single sensor, with the company calculating the probability of failure as “virtually inconceivable” - once every 223 trillion flight hours⁸. Each disaster stemmed from the same fatal flaw: mistaking statistical patterns for causal understanding. Tesla’s vision-only approach represents the apotheosis of this hubris - the belief that enough data and clever algorithms can overcome the fundamental limitations of passive optical sensors trying to reconstruct a three-dimensional world.
The physics that pattern matching can’t solve
Here’s what Musk either doesn’t understand or won’t admit: cameras are passive sensors that must infer depth from two-dimensional images, a computationally intensive process prone to catastrophic failure modes. When the sun glares off wet pavement, when fog reduces visibility, when a white truck blends into an overcast sky - these aren’t edge cases to be solved with more training data. They’re fundamental limitations of relying on reflected light to understand the world. NHTSA is currently investigating 2.4 million Tesla vehicles after crashes in these exact conditions, including a pedestrian fatality in fog⁹. Meanwhile, LiDAR actively measures distance using time-of-flight calculations - it doesn’t guess, it knows. Waymo has demonstrated this repeatedly, with their sensors detecting pedestrians preparing to enter roadways when cameras couldn’t see them, identifying humans behind stopped buses invisible to optical systems.
The academic consensus is damning. Philip Koopman at Carnegie Mellon, one of the foremost autonomous vehicle safety experts, puts it bluntly: “Machine learning runs out of steam before you get to common sense”¹⁰. MIT researchers found Tesla “replaced a proven safety system with an unproven one”¹¹, while papers across the field demonstrate that vision sensors fail catastrophically in harsh weather conditions, producing “erroneous and misleading outputs” or complete sensor failure¹². Even Missy Cummings, former Navy fighter pilot and NHTSA senior safety advisor, observed a “statistical correlation with the sun going behind clouds” triggering Tesla vision system failures¹³. But perhaps the most damning admission comes from Tesla’s own Head of Autopilot, Ashok Elluswamy, who acknowledged Tesla is “lagging by a couple years” behind Waymo and that their vision-only approach is “technically challenging”¹⁴.
The industry votes with its sensors
While Musk predicted competitors would “dump LiDAR, mark my words,” the exact opposite has occurred. Every single major autonomous vehicle company except Tesla - Waymo, Cruise, Aurora, Zoox, Baidu Apollo - uses comprehensive sensor fusion. Waymo’s sixth-generation driver employs 13 cameras, 4 LiDAR sensors, and 6 radar units¹⁵, achieving genuinely driverless operations across five cities. Baidu Apollo has completed 130 million autonomous kilometers with 1.1 million rides in Q4 2024 alone¹⁶, expanding globally while Tesla’s Robotaxi remains limited to a tiny, geofenced area of Austin with safety drivers. Even Chinese manufacturers, whom Musk once dismissed as copycats, have driven LiDAR costs down to $200 per unit¹⁷ while advancing the technology beyond Tesla’s capabilities.
The cost argument that Musk clings to has evaporated. Luminar’s sensors have dropped from thousands to hundreds of dollars, while Waymo’s entire sensor suite costs roughly $12,650 - a premium that disappears when you consider the liability of crashes. Tesla’s camera-only system costs about $400, but when your “Full Self-Driving” can’t actually drive itself and accumulates the highest crash rate among ADAS systems, you’re not saving money - you’re externalizing costs onto society in the form of accidents, injuries, and deaths. Consumer Reports and IIHS have both removed safety endorsements from Tesla vehicles after radar removal¹⁸, while insurance companies report Tesla’s systems create higher injury rates when accidents occur.
When vision fails, people die
The real-world consequences of Musk’s technological absolutism are written in crash reports. There’s the Tesla that slammed into an overturned truck because the vision system couldn’t recognize an unusual orientation. The multiple collisions with emergency vehicles because flashing lights confused the cameras. The phantom braking incidents - 1,845 complaints for Model 3/Y alone¹⁹ - where shadows, overpasses, or lighting changes trigger sudden deceleration. Each incident represents a failure mode that sensor fusion would have caught. When Waymo’s LiDAR detects an obstacle, it doesn’t matter if the cameras are confused by glare or the radar is getting returns from a metal sign - the system has redundancy. Tesla’s approach is equivalent to flying a commercial airliner with a single altimeter because “pilots have eyes.”
Former Tesla employees have blown the whistle on these dangers. Lukasz Krupski, who won a landmark case against Tesla for retaliation, described the situation starkly: “It affects all of us because we are essentially experiments on public roads”²⁰. He documented how Tesla’s vision system problems correlated with weather changes, finding systematic failures when environmental conditions departed from California sunshine. Another whistleblower revealed that cameras on Tesla vehicles aren’t properly calibrated, causing the system to literally see things that aren’t there - or miss things that are²¹. These aren’t bugs to be patched; they’re fundamental architectural flaws.
The regulatory reckoning approaches
The walls are closing in on Tesla’s vision-only gambit. NHTSA has opened multiple investigations covering millions of vehicles, focusing on crashes in low-visibility conditions that LiDAR would handle without issue²². European regulators have effectively banned FSD in its current form, with deployment delayed until at least 2028²³. China suspended Tesla’s FSD trial after just one week²⁴, demanding evidence that simply doesn’t exist. Even the Trump administration’s regulatory loosening can’t overcome physics - crashes will continue, investigations will mount, and eventually, the weight of evidence will become undeniable.
Meanwhile, competitors using sensor fusion are achieving what Tesla promised but can’t deliver. Waymo is expanding to Miami and Washington DC with fully autonomous vehicles - no safety driver, no supervision, no “beta” excuse. Their safety data shows dramatic improvements over human driving across every metric²⁵. Baidu achieved 100% driverless operations across their fleet. Even GM’s troubled Cruise, despite its setbacks, achieved more autonomous miles than Tesla before suspending operations²⁶. The industry consensus isn’t just different from Tesla’s approach - it’s succeeding where Tesla is failing.
The tragedy of preventable failures
What makes this situation particularly galling is its preventability. The lessons from Tacoma Narrows, Challenger, Quebec Bridge, and Boeing 737 MAX are clear: when you prioritize pattern matching over causal understanding, when you dismiss fundamental physics in favor of statistical confidence, when you remove redundancy because you believe your algorithms are infallible - people die. The Royal Commission investigating the Quebec Bridge disaster noted that chief engineer Cooper’s expertise “became the sole factor that was relied upon for assuring structural integrity”²⁷. Replace “Cooper” with “Musk” and “structural integrity” with “autonomous driving safety,” and you have Tesla’s current predicament.
Musk’s response to criticism has been to double down, calling critics “haters” and lawsuits “grifting” while continuing to promise unsupervised FSD “next year” - a promise he’s made for ten consecutive years²⁸. In 2016, he predicted 90% autonomous driving by 2019. In 2019, he promised a million robotaxis by 2020. In 2020, he claimed Level 5 autonomy was months away²⁹. Each failed prediction hasn’t tempered his confidence or modified his approach. Like the NASA officials who normalized O-ring erosion because “it happened before and we survived,” Musk has normalized overpromising and underdelivering while his customers serve as unwitting test subjects for a fundamentally flawed architecture.
The appendix that isn’t useless
The supreme irony of Musk’s “expensive appendix” analogy becomes even more delicious when you understand modern medical science. For over a century, the appendix was dismissed as a vestigial organ - evolutionary baggage with no purpose. Medical textbooks called it a “evolutionary accident” and doctors routinely removed healthy appendixes as a precautionary measure during other surgeries.
But science has a way of humbling our assumptions. In the past two decades, researchers have discovered that the appendix plays a crucial role in maintaining gut health and immune function. It serves as a “safe house” for beneficial bacteria, helping to repopulate the intestinal microbiome after illness or antibiotic treatment. The appendix produces immune cells, manufactures infection-fighting chemicals, and may even help train the immune system to distinguish between helpful and harmful bacteria.
Studies now show that people without an appendix have different gut microbiomes and may be more susceptible to certain infections and autoimmune conditions. The organ we once considered useless turns out to be an important component of our body’s integrated defense systems - much like the sensory redundancy that Tesla foolishly discarded.
So when Musk compared LiDAR to an “expensive appendix,” he inadvertently made the perfect argument for why autonomous vehicles need multiple sensor types. Just as evolution preserved the appendix because it serves vital functions we didn’t initially understand, the autonomous vehicle industry has overwhelmingly embraced sensor fusion because redundancy saves lives. The appendix isn’t evolutionary waste - it’s evolutionary wisdom. And neither is LiDAR.
The expensive appendix that saves lives
Similarly, LiDAR isn’t a crutch or a fool’s errand - it’s a critical safety component that provides information cameras simply cannot capture. When visibility degrades, when unusual scenarios arise, when milliseconds matter, active sensing saves lives. Every other serious player in autonomous vehicles understands this. The question isn’t whether Tesla will eventually add LiDAR - their secret Luminar purchases suggest they already know the answer. The question is how many crashes, investigations, and preventable tragedies will occur before Musk’s ego yields to engineering reality.
The path forward is obvious to everyone except Tesla: sensor fusion combining cameras, LiDAR, and radar provides the redundancy and reliability required for genuine autonomous driving. This isn’t a matter of opinion or disruption or thinking differently - it’s physics. Light has properties, sensors have limitations, and safety-critical systems require redundancy. Until Tesla acknowledges these facts, their “Full Self-Driving” will remain what it’s always been: a beta test running on public roads, accumulating crashes that proper engineering could prevent, all to protect one man’s pronouncement that anyone using LiDAR is “doomed.”
The only thing doomed here is Tesla’s vision-only approach - and unfortunately, the people who trust it with their lives.
Bibliography
TechCrunch. “‘Anyone relying on lidar is doomed,’ Elon Musk says.” 2019. https://techcrunch.com/2019/04/22/anyone-relying-on-lidar-is-doomed-elon-musk-says/
InsideEVs. “Tesla Is Anti-Lidar. So Why Is It This Lidar Company’s Biggest Customer?” 2024. https://insideevs.com/news/718859/tesla-luminar-top-customer-2024/
Waymo. “Waymo significantly outperforms comparable human benchmarks over 7+ million miles of rider-only driving.” 2023. https://waymo.com/blog/2023/12/waymo-significantly-outperforms-comparable-human-benchmarks-over-7-million
Waymo. “New Data Hub Shows How Waymo Improves Road Safety.” 2024. https://waymo.com/blog/2024/09/safety-data-hub/
NBC News. “Federal regulator finds Tesla Autopilot has ‘critical safety gap’ linked to hundreds of collisions.” 2024. https://www.nbcnews.com/tech/tech-news/feds-say-tesla-autopilot-linked-hundreds-collisions-critical-safety-ga-rcna149512
Washington State Department of Transportation. “Tacoma Narrows Bridge history - Bridge - Lessons from failure.” https://wsdot.wa.gov/tnbhistory/bridges-failure.htm
IEEE Spectrum. “The Challenger Disaster: A Case of Subjective Engineering.” https://spectrum.ieee.org/the-space-shuttle-a-case-of-subjective-engineering
The Seattle Times. “The inside story of MCAS: How Boeing’s 737 MAX system gained power and lost safeguards.” https://www.seattletimes.com/seattle-news/times-watchdog/the-inside-story-of-mcas-how-boeings-737-max-system-gained-power-and-lost-safeguards/
PBS NewsHour. “U.S. opens new investigation into Tesla’s ‘Full Self-Driving’ system after fatal crash.” 2024. https://www.pbs.org/newshour/nation/u-s-opens-new-investigation-into-teslas-full-self-driving-system-after-fatal-crash
Carnegie Mellon University. “Philip Koopman - View Profile & Connect.” https://www.cmu.edu/news/experts/philip.koopman
MIT Professional Education. “Is LiDAR a ‘fool’s errand’?” https://professional.mit.edu/news/articles/lidar-fools-errand
PMC. “Sensor and Sensor Fusion Technology in Autonomous Vehicles: A Review.” https://pmc.ncbi.nlm.nih.gov/articles/PMC8003231/
George Mason University. “From ChatGPT to Tesla’s Autopilot, Mason professor Missy Cummings isn’t afraid to call out bad tech.” 2023. https://www.gmu.edu/news/2023-02/chatgpt-teslas-autopilot-mason-professor-missy-cummings-isnt-afraid-call-out-bad-tech
InsideEVs. “Tesla Admits FSD Is ‘Lagging By A Couple Years’ Behind Waymo.” https://insideevs.com/news/760336/tesla-couple-years-behind-waymo/
Automotive Dive. “A look inside the sixth-generation Waymo Driver.” https://www.automotivedive.com/news/waymo-6th-generation-driver-autonomous-driving-hardware-robotaxi-lidar-ai/725519/
Technology Magazine. “How Baidu’s Apollo Go Targets Global Robotaxi Expansion.” https://technologymagazine.com/articles/how-baidus-apollo-go-targets-global-robotaxi-expansion
Futurism. “This $500 lidar system could prove Elon Musk wrong.” https://futurism.com/the-byte/500-lidar-system-prove-elon-musk-wrong
Consumer Reports. “Tesla Model 3 Loses CR Top Pick Status and IIHS Safety Award After Dropping Features.” https://www.consumerreports.org/car-safety/tesla-model-3-loses-cr-top-pick-status-and-iihs-safety-award-a2421791602/
KGW. “Tesla ‘phantom braking’ more common than previously reported.” https://www.kgw.com/article/news/investigations/tesla-drivers-complain-unexpected-phantom-braking/283-f6916c06-f052-4d01-9813-e0c5fe3eff39
Yahoo Finance. “Tesla whistleblower raises doubts over safety of self-driving cars.” https://finance.yahoo.com/news/tesla-whistleblower-raises-doubts-over-103607298.html
Outlook India. “Former Tesla Employee, Now A Whistleblower, Exposes Alleged Safety Issues in Self-Driving Technology.” https://www.outlookindia.com/international/us/former-tesla-employee-now-a-whistleblower-exposes-alleged-safety-issues-in-self-driving-technology-news-334912
CNBC. “Tesla faces NHTSA investigation of ‘Full Self-Driving’ after fatal collision.” 2024. https://www.cnbc.com/2024/10/18/tesla-faces-nhtsa-investigation-of-full-self-driving-after-fatal-collision.html
Not a Tesla App. “Tesla’s FSD In Europe Faces More Regulatory Delays.” https://www.notateslaapp.com/news/2586/teslas-fsd-in-europe-faces-more-regulatory-delays
CleanTechnica. “Tesla FSD Banned In UK, Trial Ends In China After One Week.” 2025. https://cleantechnica.com/2025/03/27/tesla-fsd-banned-in-uk-trial-ends-in-china-after-one-week/
Time. “Waymo’s Self-Driving Future Is Here.” https://time.com/collections/time100-companies-2025/7289599/waymo/
Yahoo Finance. “What Cruise’s self-driving end means for Tesla and Waymo: Morning Brief.” https://finance.yahoo.com/news/what-cruises-self-driving-end-means-for-tesla-and-waymo-morning-brief-110053822.html
Academic Block. “Quebec Bridge Collapse.” https://www.academicblock.com/technology/engineering-disasters/quebec-bridge-collapse
The Autopian. “Elon Musk Predicts Level 4 Or 5 Full Self-Driving ‘Later This Year’ For the Tenth Year In A Row.” https://www.theautopian.com/elon-musk-predicts-level-4-or-5-full-self-driving-later-this-year-for-the-tenth-year-in-a-row/
Wikipedia. “List of predictions for autonomous Tesla vehicles by Elon Musk.” https://en.wikipedia.org/wiki/List_of_predictions_for_autonomous_Tesla_vehicles_by_Elon_Musk
ResearchGate. “SURVEY ON SENSOR FUSION FOR AUTONOMOUS DRIVING: TECHNIQUES AND CHALLENGES.” https://www.researchgate.net/publication/370491935_SURVEY_ON_SENSOR_FUSION_FOR_AUTONOMOUS_DRIVING_TECHNIQUES_AND_CHALLENGES