There is one technology that Elon Musk has a vendetta against. Lidar. According to Musk, “Anyone relying on lidar is doomed. Doomed. Expensive sensors that are unnecessary.” As such, Tesla’s FSD (Full Self Driving) doesn’t use these laser-based 3D mapping sensors; in fact, it only uses cameras to understand the world around it. This goes against almost the entire self-driving industry, which widely believes lidar to be essential for any fully autonomous self-driving cars. As such, there is a huge question of whether Tesla could ever create a safe self-driving car with its current approach, which is profoundly worrying considering Musk is betting the entire company’s future on unlocking this technology before anyone else. But, it seems Musk may have realised the error of his ways, as Tesla just purchased a whopping $2 million worth of lidars from Luminar. This shift in ethos might seem minor, but it could save Tesla. Let me explain.
When I say Musk has bet the company entirely on self-driving, I mean it! Tesla has scrapped its highly anticipated Model 2 in favour of a robotaxi on the same platform. It’s also halted gigacasting development, which could have made all Tesla models even cheaper to manufacture. Musk has even fired core talent, such as the head of battery development, head of vehicle development, and head of supercharger operations. Musk has even disbanded the teams involved in these projects! This seems to be a cost-saving move to funnel as much funds as possible into developing Tesla’s FSD AI into a legally fully autonomous system.
In other words, Musk is sacrificing all of Tesla’s advantages in the EV world in a bid to bring self-driving cars to fruition.
This is a problem, as FSD is far from reaching this goal!
You see, back in 2021, Musk went against his own engineers’ advice and removed the radar and ultrasonic sensors from all Teslas, leaving FSD only camera data to understand the world around them. These engineers actually went to the press and leaked that this move made FSD far less safe and that the public should not be using the product. Recent data released by Tesla confirmed this, suggesting that FSD causes at least twice as many deaths per mile driven as a human driver, and that is even with a human monitoring FSD! This figure would be way higher if FSD was left to its own devices and used as a self-driving car.
With no other sensors, FSD has a huge Achilles heel. There is no way to double-check or ratify misreadings, which often happens with computer vision (i.e. an AI that can categorise what is in a video feed). If the cameras are comprised through sudden bad lighting, mud or rain, the system can drive completely blind. So even if the FSD AI has billions of dollars poured into its development and becomes a far better driver, this built-in fault can easily stop it from meeting the safety standards to become a fully autonomous system.
This is why almost every other self-driving program uses lidar. This technology uses range-finder lasers to create a real-time, detailed 3D point cloud map of the sensor’s surroundings. This high-fidelity enables self-driving cars to identify objects like other motorists, lanes, pedestrians, cyclists, roadsigns and junctions far more consistently than any other technology. It isn’t perfect, as it can be compromised by rain and can’t read a road sign. But if you combine lidar with camera feeds, radar, and ultrasonic sensors, then this suite gives a self-driving car enough data to read the road correctly and consistently and provides the system with enough redundancy to manage complex conditions, which could comprise different sensor types.
Now, Luminar’s lidars cost around $1,000 each. So Tesla’s $2 million order equates to around 2,000 lidar units! Tesla has ordered from Luminar before, but it was in far smaller quantities, as they only used them to aid in developing and verifying the computer vision behind FSD. So why has Tesla ordered so many?
Well, Tesla is expected to unveil its Model 2-based robotaxi later this year. Musk has even proposed to Chinese authorities to test this robotaxi on their roads, to which they were very receptive. With FSD in its current state, such a robotaxi would be a disaster. It would almost certainly crash and have incidents far too often. All eyes are already on Tesla, and investors are very tentative about the shift to focus the company’s entire future on self-driving technology. Even the slightest hint that this robotaxi is dangerous could have a catastrophic impact. Musk simply can’t take any chances. As such, I think this lidar order is for the first robotaxis, and this is the first sign that FSD will adopt lidar technology over the next few years.
This should make FSD and this robotaxi far more reliable and safe. As such, it makes sense for Tesla to rapidly adopt lidar sensors, as it makes Musk’s huge self-driving bet far more likely to stick. It’s still far from dead certain to pay off though; we don’t even know if current AI technology can reach the consistency and safety required for self-driving cars, as AI training and development are starting to hit diminishing returns. There are also other issues, such as figuring out how to retrofit these lidars to Teslas that have been fitted with FSD, which Tesla has promised will eventually be self-driving. Tesla also doesn’t have a vast amount of real-world lidar driving data, and they will need a colossal amount to train the FSD AI to use this new type of data.
However, Tesla has yet to confirm any lidar plans. For all we know, these could be for an entirely different project. It makes sense for them to be for FSD and the robotaxi, but Musk is acting incredibly erratic at the moment, so this doesn’t mean much. But, for the sake of Tesla’s future, I hope Musk isn’t so stubborn as to ignore the obvious advantages of lidar because it could make or break Tesla’s future.
Thanks for reading! Content like this doesn’t happen without your support. So, if you want to see more like this, don’t forget to Subscribe and follow me on BlueSky or X and help get the word out by hitting the share button below.
Sources: Reuters, Bloomberg, Forbes, The Verge, Spectrum IEEE, Luminar, Planet Earth & Beyond