- Tesla’s Full Self-Driving is starting to grow up, Tesla CEO Elon Musk claimed Wednesday.
- The automaker’s driver-assistance software is becoming more robust, and Musk claims it’s almost good enough to take the training wheels off.
- By the end of the year, Tesla believes it will be able to remove the safety driver from its car.
Tesla says its Full Self-Driving software is finally growing up. But much like a teenager with a learner’s permit, the software isn’t quite ready to be left unsupervised—yet.
That’s set to change by the end of the year, or so Elon Musk claimed on Wednesday.
During Tesla’s quarterly earnings, the CEO promised that its Robotaxi fleet in Austin will have its safety drivers removed by the end of the year, barring any roadblocks from regulators. It also plans to expand its service to between 8 and 10 metro areas.
Photo by: Tesla
“It depends on various regulatory approvals,” said Musk. “You can actually see most of our regulatory applications online, because they’re public information. We expect to be operating in Nevada, Florida and Arizona by the end of the year.”
Tesla is currently operating in Austin, Texas and the Bay Area near San Francisco, California. It also seemingly plans to launch in Illinois and Colorado, as suggested by job postings last week.
So far, Tesla’s Robotaxi fleet has logged 250,000 miles in Texas and over a million miles in California. FSD Supervised—the version that anyone with a Tesla can get today today for $8,000—has amassed a whopping 6 billion miles driven to-date, which is an ocean of data that Tesla has been training on. That being said, it’s still not enough for Musk to go full-gas on calling for the removal of safety drivers.
The CEO says that it “makes sense” to keep the safety driver in the front seat for at least a little while.
“I think even if the regulators weren’t making us do it […] it make sense to have a safety driver or occupant in the car,” said Musk during the call. He added that Tesla would use safety drivers for “maybe 3 months” when it enters a new market to make sure there are not area-specific challenges, then pull out the drivers and let the cars putter around on their own.
Musk also made a rather out-of-pocket conservative remark about the current customer-facing software: early releases aren’t for everyone.
“We need time to smooth the rough edges,” he admitted.
He cautioned that brand new major releases of FSD prioritize tech advancements and safety over comfort. The means waiting for an incremental release like FSD F14.2.1 or later (the most advanced version available today is 14.1.3) before updating if comfort is a big deal. But if you want the latest and greatest today, you owners can still enabled “Advanced” software updates in their cars, giving them access to the more experimental versions.
New and cool, or stable and safe. Everyone gets to choose.
Speaking of new, Musk says that Tesla is also planning to bring reasoning to a future release of its new FSD version 14 software. This will enable the car to make more complicated decisions based on the world around it, like deciding where it should look for parking after it drops a driver off at the front door of a store. If the parking lot is crowded, for example, the car might choose to look for a spot in the back of the lot rather than right up front.
The real test is whether or not folks can text and drive—or at least that’s what Musk says. The car isn’t there yet, but the driver assistance software seems to be advancing rapidly. Maybe not quite the HAL 9000 of 2025, but it’s getting there.
For now, Musk says he remains confident that FSD will eventually be safer than a human driver without anybody behind the wheel. Surprisingly, he didn’t put an exact timeline on this metric, but the context around it paints the bigger picture: the geofenced rollout could soon begin. Until then, Tesla’s customer version of FSD will require alert human supervsion at all times.
