Tesla is expected to unveil a robotaxi tonight: 5 things to know

An idiosyncratic billionaire takes to the stage (with, perhaps, a humanoid robot by his side?) to unveil a futuristic technology that he promises will transform the world — a vision alternately celebrated, mocked and feared.

It’s a scene straight out of Hollywood. And that’s exactly where it’s going to happen tonight — on a Warner Brothers studio in Burbank, California.

Tesla is calling the event “We, Robot.” At 7 p.m. Pacific time, Tesla CEO Elon Musk is expected to unveil the company’s design for a dedicated robotaxi, a Tesla designed exclusively to ferry passengers without a driver — a feat the company’s semi-autonomous software has not yet demonstrated it can do.

“The future will be streamed live,” the company posted on X, the social media platform Musk owns.

Here are 5 things to know about Tesla’s big bet on autonomous driving.

Musk says robotaxis are key to Tesla’s future profits

Tesla makes money selling electric vehicles — in fact, its profit margins on its cars, which are consistently in the double digits, are enviable for an automaker. But Musk has his eye on the much fatter profit margins of the software industry.

In addition to selling cars, Tesla sells an expensive software package called “Full Self Driving (Supervised).” It can autonomously direct a Tesla on a wide range of roads, and navigate stoplights, stop signs and pedestrians without human input … most of the time. But it periodically requires a human to take over, which means it’s not truly autonomous.

Musk has always maintained that demand for the software will be much higher when it is fully autonomous — in part because that would allow people to make money off their personal vehicles by lending them out, like a driverless Uber or Lyft.

“The value of a fully electric autonomous fleet is generally gigantic — boggles the mind, really.” he told investors in 2021. “That will be one of the most valuable things that's ever done in the history of civilization.”

A custom-built vehicle just for robotaxi rides is a relatively new twist on Musk’s vision. Expect to hear more about it on Thursday.

He’s promised they’re coming next year — for years

Musk’s become infamous for his overly enthusiastic estimates of how soon Tesla robotaxis will arrive. He’s even taken to poking fun at himself for all the times he’s been wrong.

• Musk in 2019: “Next year for sure we will have over a million robotaxis on the road.” 
• Musk in 2020: “I think we could see robotaxis in operation … next year. Not in all markets, but in some.” 
• In 2021: “I’m highly confident the car will drive itself for the reliability in excess of a human this year.” 
• In 2022: I would be shocked if we do not achieve full self-driving safer-than-human this year.” 
• 2023: “Now, I know I'm the boy who cried [Full-Self-Driving], but man, I think — I think we'll be better than human by the end of this year.”
• And this summer: “Obviously, my predictions on this have been overly optimistic in the past. … Next year seems highly probable to me.” 

The problem is that while Tesla’s software can drive a vehicle without human help much of the time, it’s not yet reliable enough to drive unassisted all of the time.

Companies like Alphabet’s Waymo and GM’s Cruise, meanwhile, have already sent driverless taxis onto streets — although Cruise put human “safety drivers” back behind the wheel after a crash last year. The systems often have someone on call to assist remotely if a car gets stuck. But that’s a far cry from needing constant oversight: According to data it supplied the state of California, Waymo drove nearly 1.2 million fully driverless miles last year with a total of 14 “disengagements,” or times the software required manual control.

Even for those companies, robotaxis aren’t profitable yet. The auto market research giant J.D. Power recently surveyed people who have ridden in robotaxis and found that while passengers generally liked the experience, they don’t find the taxis practical. Until they’re cheaper and cover more ground, the pollsters concluded, “the service will remain a novelty transportation method.”

Tesla’s approach to autonomy is unusual — and controversial

So if robotaxis already exist, why doesn’t Tesla have this tech yet? There’s a big difference between how other companies — like Waymo, Cruise, the driverless trucking company Aurora and a host of startups — approach autonomy, and how Tesla is trying to do it.

Musk decided to build a system based only on relatively cheap cameras, with no other inputs; other companies also use radar and other pricey high-tech sensors. Musk also has embraced “end-to-end learning,” where the artificial intelligence “learns” how to drive from raw data; other companies add human-designed rules and guardrails to their AI systems.

Analyst George Gianarikas of Canaccord Genuity Group notes that Musk’s approach requires billions of dollars of upfront investment in AI, but much cheaper hardware on vehicles. That’s a combination that is expensive now, but would pay off if there were, say, millions of robotaxis on the road.

Musk is adamant that Tesla’s approach is superior. “Our entire road network is designed for biological neural nets” — that is, human brains — “and eyes, so naturally cameras and digital neural nets are the solution,” Musk told investors earlier this year. Tesla also has enormous amounts of driving data from its vehicles on the road today.

Other companies say this approach is not just wrong but dangerous. Aurora took the unusual step of preemptively emailing reporters ahead of Tesla’s event this week to share bullet points about exactly what they object to. Those included concerns about making sure a system is learning good driving behaviors — not bad ones, like running stop signs — and that there are systems of checks and balances.

Aurora's email paraphrases points CPO Sterling Anderson made in a recent webcast: the approach favored by Tesla, the summary notes, is a " 'train and pray' approach where you fix a problem by throwing more data at the system – we find this to be problematic in a safety critical industry where you need confidence and proof you’ve actually fixed."

Anderson used to work at Tesla, where he helped launch Tesla’s Autopilot software, its first partial-automation system, the Aurora email notes. Waymo just snagged a former Tesla exec for its team, too.

One wild card: What will regulators think?

The United States still has no federal laws governing self-driving, so a patchwork of state and city regulators set the boundaries of what companies can and cannot do.

Musk has always acknowledged that achieving full self-driving is not just a matter of technological innovation; if regulators aren’t convinced a robotaxi fleet is safe, it isn’t going anywhere.

That has implications for the physical design of vehicles. Cruise recently abandoned plans for a futuristic robotaxi vehicle with no steering wheel, returning to a more conventional design that a human could operate, primarily to reduce the risk of running afoul of regulators.

And governmental concerns could also affect software. Gianarikas says regulators who dig into the coding of a system built by “end-to-end” deep learning might not like what they find.

“You can imagine a scenario where [regulators] just kind of have this moment, like ‘What? You don't … have any hard-coded software rules?” he says. “‘How do you control it?’”

Still, Gianarikas notes that while there are plenty of reasons to be skeptical about a Tesla robotaxi fleet, Elon Musk has a track record of eventually proving skeptics wrong.

The reveal might include another kind of robot

The event’s name — “We, Robot” — is a nod to a classic Isaac Asimov short story collection exploring the ethical and psychological implications of building increasingly human-like robots. It’s also the title of a very vaguely related Will Smith action movie.

That hints at the possibility that, in addition to a robotaxi, the reveal could feature Optimus, the humanoid robot Tesla has been developing as an autonomous laborer capable of doing repetitive tasks.

“I think the long-term value of Optimus will exceed that of everything else at Tesla combined,” Musk told investors this summer. “A humanoid robot that can do pretty much anything you asked of it. … I think everyone on Earth is going to want one.”

Optimus, a black and white robot with a featureless face of smooth black glass, can walk — in a stilted, gliding sort of way. Tesla has shared videos of it sorting objects, standing on one leg and dancing.

Dan Ives, an analyst and a long-time Tesla bull, will be in attendance on Thursday night. He’s less interested in androids and more in whether Musk can demonstrate a fully autonomous vehicle that actually works.

“This needs to be a jaw-dropper type of event,” he said — hype and promises are not going to cut it.

“Billions of dollars spent on this,” said Ives. “This can’t just be, “Get the popcorn out.’”