Tesla Is Losing The Autonomous Vehicle Race

Elon’s “Full Self-Driving” system is far from what he promised.

For nearly a decade, Tesla has been touted as a technology company, not a car company. Why? Well, because all of their cars are equipped with cameras covering every angle around the car. They record every second the vehicle is in use and send the videos back to Tesla. Musk then uses this vast amount of data to train a self-driving AI. In theory, this should enable Tesla’s fully autonomous driving system to be significantly cheaper to install and far more robust than anyone else’s. This is half the reason Tesla is valued so highly, as this artificial intelligence could be worth billions. Yet, over the past year, the reality of Tesla’s approach has become apparent, and many rivals have surpassed them in this self-driving race. But how? And can Musk catch back up?

Let’s start at the beginning, when the polish originally started to come off Tesla’s AI. Way back in 2019, Musk promised a million fully autonomous Tesla robotaxis on the road by the end of 2020. Admittedly, Musk is known for over-promising on deadlines — just look at the Model 3, Cybertruck, Roadster, Semi, and even the Starship. But with all of these projects, we could see that he was still taking steps towards completion while the projects were delayed, including building the Model 3 factory, hiring Cybertruck factory technicians, and test-firing the Starship rockets. Yet, we have still to see Tesla take the necessary baby steps toward developing robotaxis. They haven’t even been granted the necessary permits. As such, Musk has now said their robotaxis will be available in 2024, but even this timeline seems far too optimistic.

So what has caused this? Well, the first hurdle might be that the DOJ is currently in the midst of investigating Tesla for false advertising by selling their driver assistance system (known as “Full Self-Driving”) as a finalised, fully autonomous system. There is even a risk Tesla could get slapped with manslaughter charges for spreading this misinformation (read more here). As you can imagine, getting these permits will be incredibly challenging with such an investigation looming over Tesla. But this might only be half the problem. You see, Tesla’s dogged determination to use a visual-only system may have prompted this mess. Let me explain.

Google’s Waymo and GM’s Cruise both use cameras and LiDAR to power their self-driving systems. LiDAR uses a scanning laser to create a 3D map of the world around the sensor. This provides it with a much more precise understanding of the events taking place and what is going on around them. It also means that they don’t need complex programmes to decode what is in the video feed. As a result, their AI has a significantly easier time figuring out how it should be driving. These companies also don’t have access to the vast data sets that Musk does, so instead, they train them on much more specific scenarios. By doing this, their AI can be refined far quicker and more accurately in less time than Tesla’s, due to their use of less but higher-quality data. In fact, many experts question whether Tesla’s approach actually has a viable path to full autonomy given that it is so complex in comparison.

But the proof is in the pudding here. You see, Waymo has been offering free driverless taxi rides in Phoenix since 2020. Both Cruise and Waymo also have Driverless Pilot permits from the California Public Utilities Commission. This allows them to pick up Californian passengers in a fully driverless vehicle at any time of day with a maximum speed of 65 mph. As of yet, Waymo can’t charge for rides, but because Cruise received a deployment permit for robotaxis in June, they can.

This is still a ways off from the evolution of fully-fledged nationwide robotaxis, but it is still miles ahead of where Tesla is, which begs the question: Why hasn’t Musk used LiDAR like this?

Well, for a long time, LiDAR systems were unreliable and extremely expensive, and as such, Musk saw them as unfit for commercial use. He also argued that humans are capable of driving with only visual data, so an autonomous car should be able to too.

Then, over the past few years, LiDAR technology has vastly improved. It is not only more robust and accurate, but it is also significantly less expensive than it was previously. Now, it is commercially feasible to deck out consumer cars with several high-grade LiDAR systems.

So can Tesla switch their MO and catch up? Well, Teslas owned by the company have been spotted using retrofitted LiDAR systems. So Musk is, at the very least, considering this approach. But he would have to effectively start from scratch, given that Tesla’s AI is not designed to use this type of data. In order for Tesla to attain the same progress, they would need to put considerable effort into developing a new AI (or at least heavily modifying their current one) and do a mass rollout of LiDAR systems in the new Teslas. Sadly, I can’t see this happening any time soon.

The only advantage Tesla might have over Waymo and Cruise is the fact that there are already hundreds of thousands of Teslas on the road with the hardware permitting them to be fully self-driving (at least theoretically). Therefore, even if Tesla is several years behind other companies, this might mean they can reach the mass market quicker. But don’t forget that GM owns Cruise, and Waymo has close links with the Mercedes-Benz Group, Nissan, Renault, Jeep, Peugeot, Opel, Vauxhall, Jaguar Land Rover, and Volvo, meaning that once these systems are refined, they will find their way into consumer products incredibly quickly.

So that is how Tesla is losing the EV race, and how it will be extremely difficult for them to recover. I would love to say that there is an easy solution for Tesla to catch up and gain the same level of technology and legal permits as Waymo or Cruise, but sadly, I can’t see a way to do so. Admittedly, Musk does have his new Dojo supercomputer, which may improve how Tesla trains its AI, but this alone probably won’t be enough. Either way, only time will tell. You never know. Musk might have something hidden up his sleeve.