Tesla Robotaxis. I was wrong.

I've been doubtful that Tesla (or anybody) would ever make an AI which would be able to safely drive a car. Well, I'm eating my words.

From The Driven.

After more than a decade of development on its revolutionary vision based autonomous driving software, Tesla will finally reveal its much anticipated Robotaxi on August 8, 2024.
The Robotaxi unveil will mark the convergence of Tesla’s latest Full Self Driving software and revolutionary 3rd generation vehicle manufacturing and usher in a new era of “Transport as a Service” (TAAS) with massive ramifications for the 70 million unit per annum global fossil car industry.

The announcement comes as the online Tesla community is abuzz with Full Self Driving (FSD) Beta software testers raving about the latest FSD version 12.3.3 update, with drivers reporting zero interventions during long drives in complex city traffic.

Former Tesla employee and YouTuber Farzad Mesbahi discussed the latest software update with James Douma, who’s one of tens of thousands of Tesla drivers in the US who’ve been testing the Beta software over the past two years.

“It’s a pretty remarkable departure in behaviour from V11,” said Douma. “It just works, you just don’t have interventions anymore.”

Douma, who’s been testing the latest update for the last two weeks, says he’s completed hours of city driving without manually overriding the software.

“The first thing I did was spend 3 hours driving all over the part of LA I live in, just random pin drops.” said Douma.

“And I didn’t have any interventions, it was rock solid.”

Another FSD tester and online Tesla blogger Omar Qazi, AKA @WholeMarsBlog, has also been testing FSD beta and has posted some stunning videos of version 12.3.3 in action around San Francisco. 

 

Unlike other companies who’ve attempted to use LiDAR to solve autonomous driving, Tesla’s strategy from the beginning was to use a vision-based system of camera’s and artificial intelligence.

The theory being that humans naturally use vision to drive and navigate the world so why shouldn’t machines? Our road networks are all designed for vision with lines and signs which can be easy read by cameras and AI.

The software is so advanced that it can differentiate between sedans, utes, trucks and buses as well as motorbikes, scooters and bicycles. It can accurately identify pedestrians, traffic cones, wheelie bins and even dogs and place them in 3D space with astonishing precision.

Unlike the purely object based LiDAR system, the cameras can also identify and read traffic signage such as stop signs, traffic lights, speed limits, road works and even the arrows and symbols painted onto road surfaces. For an in-depth look of Tesla’s FSD software development see The Rise of the Machines: Tesla drives 50km autonomously through heavy LA traffic.

Unlike other companies who’ve attempted to use LiDAR to solve autonomous driving, Tesla’s strategy from the beginning was to use a vision-based system of camera’s and artificial intelligence.

The theory being that humans naturally use vision to drive and navigate the world so why shouldn’t machines? Our road networks are all designed for vision with lines and signs which can be easy read by cameras and AI.

The software is so advanced that it can differentiate between sedans, utes, trucks and buses as well as motorbikes, scooters and bicycles. It can accurately identify pedestrians, traffic cones, wheelie bins and even dogs and place them in 3D space with astonishing precision.

Unlike the purely object based LiDAR system, the cameras can also identify and read traffic signage such as stop signs, traffic lights, speed limits, road works and even the arrows and symbols painted onto road surfaces. For an in-depth look of Tesla’s FSD software development see The Rise of the Machines: Tesla drives 50km autonomously through heavy LA traffic.

If Tesla delivers on its August 8th commitment to showcase the self-driving Tesla Robotaxi, it will mark yet another correct prediction made by technology futurist Tony Seba.

Seba, who was interviewed on The Driven podcast last year, predicted in his 2014 book Clean Disruption that lithium-ion batteries would reach $50/kWh by 2027.

That was a forecast that many people said was crazy. However, it now seems Seba’s prediction was too conservative as Chinese battery maker (and Tesla supplier) CATL is likely to reach the milestone by mid-2024.

Despite being considered one of the boldest technology forecasters in the world, Seba has also underestimated the speed of development of battery longevity. In 2017 he predicted the first million-mile battery by 2030 however last week CATL announced a new EV battery with a 1.5 million km warranty, effectively beating Seba’s prediction by 5 years.

On autonomous vehicles Seba had some fascinating insights which he shared during his interview with The Driven.

“The day that we get level four, autonomous technology ready and approved by regulators, when that converges with on-demand, and electric transportation we will get what we call transportation as a service [TAAS].” Seba told The Driven.

“Some call it Robotaxi. Essentially, when that happens the cost per mile of transportation is going to drop by anywhere from 10 to 20 times.”

“So for most people who can barely pay their bills, it won’t make any sense to own a car,” said Seba.

“Do I spend $50,000 over the next five years to own a car? Or do I pay $100 a month for a subscription to transportation as a service?”

Seba says ICE vehicles get around 140,000 miles (225,000 km) over their lifetime. An EV with a 1.5 million km battery will get almost 7 times that amount. This means that EVs will last at least 6-7 times longer than ICE vehicles meaning the global car market will likely drop by over 75% because people won’t need to replace cars as often.

“People are going to be buying vehicles a lot less often. So with that, essentially cut the global vehicle market by a factor of four or five.”

TAAS combined with the million-mile battery will mean new vehicle sales will drop even further as people opt for super cheap electric robotaxi transport instead of spending tens of thousands on private vehicles.

“Either way, it’s pretty much over for internal combustion engine.” says Seba.

I doubt that robotaxis will be the money-spinner Musk says.  If they become that profitable, everyone will buy a Model 3 to make money, and the charges they will be able to levy will go down.  (BTW, I don't think Tesla will be allowed to run a robotaxi monopoly --- but that doesn't mean they won't be able to charge a lot for FSD)  But that only implies that TaaS will take off.  Seba is right.  Why pay a fortune for a car which sits in your driveway or at in a car park for most of its life?  Taxis are expensive because they have to have a human driver and because they're ICEVs.  Robotaxis will be cheap.

Anyway, now I'm convinced.   I was wrong.