Tesla Suspiciously Delays Robotaxi Unveiling
What's Going On?
Are you excited about Tesla’s August 8th “Robotaxi Day”? Well, tough. Bloomberg has just reported that it will be delayed until October! Apparently, this is “to allow teams working on the project more time to build additional vehicle prototypes,” at least, according to the people familiar with the decision who informed Bloomberg. But such reasoning makes very little sense when you think about it. Think back to any of Tesla’s launches; they aren’t a demonstration of the final product but more an announcement of their plans with an initial prototype. As such, Tesla doesn’t even need a functioning robotaxi to do the launch, just a rough prototype to show what the vehicle will look like and how it will work. But even if they did, they have been working for years on the “Model 2”, which this robotaxi is based on, so surely putting together a rough prototype should be easy. What’s more, Musk has claimed their self-driving technology is better and safer than that of a human driver, so surely making this prototype a fully autonomous demonstration unit would be easy, too. Right? Well, there is something damn suspicious about this delay.=
So, why delay Robotaxi Day?
Well, Tesla was set for a huge valuation increase once they announced solid robotaxi plans. After all, a fully scaled, fully autonomous taxi service could greatly add to Tesla’s annual revenue. As such, having Robotaxi Day Scheduled so close to Musk’s recent pay package vote significantly impacted many shareholders’ decisions. So, some have suggested Musk scheduled Robotaxi Day on August 8th, fully knowing they couldn’t meet that deadline and would have to delay, just to ensure there was a media buzz and interest in Tesla shares leading up to his vote, making him likely to win his insane $56 billion pay package.
Musk has demonstrated multiple times he is happy to manipulate the narrative and skirt the laws for his gain, so I have no doubt this could be the reason for the delayed Robotaxi Day. But it doesn’t explain why Tesla can’t deliver, at the very least, a concept car and their plans for it on Robotaxi Day.
So why can’t they make and show off this vehicle in time? Well, Tesla’s “Full Self Driving” (FSD) system is far from ready for fully autonomous operations, as it has an inherent fatal flaw. What’s more, fixing this flaw will dramatically change the design, roll-out and operation of the robotaxi. Let me explain.
As I have covered many times now, Tesla’s FSD only uses camera feeds to understand the world around it, while pretty much all other self-driving programs use a suite of different sensors like LIDAR, 4D radar and ultrasonic sensors. This is the fatal flaw. You see, the computer vision AI that identifies the road, road markings, signs, traffic cones, other cars, pedestrians, cyclists and wildlife, like all AI, can never be completely accurate. What’s more, there are severe limitations of the cameras, such as not working well in low light or being whited out by bright light or reflections, meaning even if the AI was completely accurate, the system would still encounter dangerous errors. For example, there are multiple videos online of Teslas driving using FSD that attempt to run over cyclists, driving into rivers or driving straight into the side of an overturned lorry because the cameras and computer vision failed.
Self-driving cars with these other sensors have a way of checking and mitigating the errors of their computer vision system. This enables them to be far more reliable. These systems require more training as they deal with a lot more data. But it is worth it for a safer, more reliable system.
As such, as FSD still only relies on video feeds, it is far from ready for full autonomy. It’s not just me saying that, either.
Electrek journalist Fred Lambert, who is by no means a Tesla sceptic, recently covered how Musk had FSD optimised on routes he and FSD influencers drove rather than the most popular routes. He points out how this has caused a dramatic discrepancy between the videos of FSD from Tesla and FSD influencers and his and his friend’s experience of the system in their Teslas. I, personally, feel this decision by Musk is unethical, deeply dangerous and potentially fraudulent, as it markets the system as more capable than it actually is. But Lambert makes no such accusations. However, Lambert was backed up in the comments, with one user posting, “This is why I was very thankful for the free month of FSD. It showed many of us Tesla owners that FSD is nowhere near what Musk has been touting for years. Not saying that every user had a bad experience, but it proved to me that it’s nowhere near worth $8,000 or $99/mo.”
But FSD was once selling for twice that! Musk had to slash the cost in half to get enough customers through the door. The reason why is clear when you look at the data. With reasonable assumptions, the current FSD system, which is supervised by a human, appears to have fatal accidents at twice the rate of a human driver! This is backed up by usage data, as customers who have paid thousands for FSD only use the system 15% of the time. Indeed, many customers have stated that they don’t trust the system, so they don’t use it.
If customers aren’t willing to use a system that they have spent a small fortune on because they distrust it, even though they are supervising it and can take over at any time, how are they going to trust a robotaxi using the same system that doesn’t have any option for them to take control? Simple, they won’t!
As such, I have a little theory.
Musk has repeatedly stated that the robotaxi will be “Cybertruck-like.” I also believe Musk is now painfully aware that they will need to adopt Lidar and other sensors in order to create a useable, trustworthy robotaxi. However, correctly integrating these sensors into a car design, particularly one made of cold-rolled steel like the Cybertruck, is incredibly challenging. Moreover, the AI will need training on highly detailed Lidar and radar maps of the streets it will operate on. Something Tesla has yet to do.
So, Tesla can’t be seen launching a car without Lidar and other sensors, as it is publicly known, particularly among institutional investors, that FSD’s vision-only system is not good enough for a robotaxi. But, they can’t even demonstrate a stationary concept robotaxi with Lidar, as they have yet to figure out how to integrate such elements into the design. As such, a concept vehicle would be wholly inaccurate to the final product. Moreover, they haven’t done the requisite Lidar mapping work, which could take many months, if not years, to complete. As such, if they launched a Lidar-equipped robotaxi, investors would question how close Tesla actually is to bringing such a robotaxi to market, as their competitors have far more mature Lidar systems and datasets. To top it off, Musk has promised that every Tesla made since 2016 will one day be fully autonomous, but none of them have these sensors, meaning that if FSD has to evolve to use Lidar, Tesla would have to spend millions, maybe billions of dollars retrofitting millions of Teslas with these sensors. Investors know this, so if Tesla’s Robotaxi uses Lidar, they could rightfully predict hard times ahead for the company as these costs rack up.
As such, Tesla is caught between a rock and a hard place when it comes to the Robotaxi. So, it’s no wonder Musk is putting it off.
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 help get the word out by hitting the share button below.
Sources: Electrek, BI, Will Lockett, Planet Earth & Beyond, Electrek, Electrek, Bloomberg