Tesla Again Says, No Robotaxi (Yet)

Desperately scrambling to catch up to the technology being used by dozens of other automotive companies such as Google's Waymo and GM's Cruise, Elon Musk has been making vaporware proclamations for years, the latest of which occurred this week at a private event in Hollywood. Touted as the unveiling of Tesla's self-driving robotaxi, it was simply another evening of braying about a concept vehicle that does not exist. There was no demonstration that the technology could work, how it could work, or even how a prototype model might drive.

Musk has claimed his robotaxi will run without any controls like a steering wheel or brake and accelerator pedals. True, companies like Navya have run autonomous shuttles around the world for years without steering wheels or pedals. But Tesla has no experience with the necessary technology for such feats. For example, the company has no teleoperations system or infrastructure that would be needed to support such vehicles. Moreover, the CEO remains stubbornly ignorant about what it takes to make autonomous vehicles work. Consequently, Tesla's driver assistance programs have been killing owners (over a dozen according to NHTSA) vs. all other car makers making hands-free driver assistance systems, which have so far killed none.

Musk insisted again that existing Teslas, which rely on simple, low-resolution cameras, will be able to run autonomous driving software when (or if) it arrives. But lacking sufficient perception systems and technology such as lidar and far infrared sensors, Tesla's vehicles will never have the smarts to safely operate on their own.

Among some of the other ideas that Musk floated again this week:

- Tesla owners will rent out their autonomous vehicles to strangers to use: Some people will do that, but such rentals didn't work out for owners and companies that have already tried this. And one has to wonder how many times you'll clean out our car from the effluvia (and, let's be honest, vomit) before you decide it ain't worth it. It's more likely to go the Airbnb route, with absentee owners of self-driving cars renting out vehicles they will never use (and someone else will clean). We don't call that revolutionary. We call that a "taxi company."

- A humanoid robot that will help you unload the groceries from your car: Wait, I thought it would be autonomous. So won't you just send your car out to the store so folks at the superstore can simply fill up your Tesla and then it will arrive at home on its own? Great, but isn't it still easier now to order on Amazon and have them deliver it rather than go to all the trouble of sending your autonomous car and getting your $40,000 Rosie the Robot to bring it in the house?

- Tesla will do autonomy on the cheap: Tesla still primarily relies on cheap video cameras--something so dangerous even its partner, Mobileye, dumped Tesla. Some have argued that it's all humans need--vision--to operate a car, so why can't computers do the same? What this fails to understand is that human vision is, as they say, many orders of magnitude of higher resolution than even today's 8K cameras. So it's like comparing rocks and apples. And of course, our whole bodies act as sensors, including balance, motion, and hearing, when driving, plus we have computing power (our brains) that exceeds that of the largest language model or deep learning network.

Bottom line: Tesla needs to add sufficient perception systems, which all Tesla vehicles currently lack. That's why the company's driver assistance programs still run into large stationary objects and become confused in tunnels. And since Tesla EVs cannot currently manage even the simplest of pre-determined, traffic-free routes, such as the company's own short tunnels in Las Vegas, it's extremely unlikely they will be able to do so any time in the future.