A former Tesla Autopilot Engineer hinted at how much the company’s Full Self-Driving (FSD) suite could improve in the upcoming over-the-air (OTA) updates on the Third Row Podcast. Eshak Mir joined the members of Third Row Podcast in its recent episode. The former member of Tesla’sTesla’s Autopilot team shared some details about the path that FSD was on and where it could lead in the future.
During his interview with the Third Row Podcast crew, Mir extensively discussed Tesla’s Autopilot rewrite and how it would affect the company’s progress with FSD in the long run. “With Autopilot, we are definitely—we were—working as hard as we can towards making that full FSD a dream to come true—pretty much—for everyone. And we were working extra hours—pretty much—to do as much as we can, and the technology is just evolving itself,” Mir said.
“Like we all know about the Autopilot rewrites. That’s what we were working on, and that was really showing that—as we were working on that—it was improving by like [1000%]…It was a game-changer…And working on that showed us that—you know—we can do so much better,” he clarified further.
He explained that the new Autopilot rewrite mostly eliminated all the little mistakes that Tesla had made with its driver-assist system in the past. Mir explained that the Autopilot rewrite helped Tesla combine the video feeds from its car’s eight cameras into one 3D layout, which made it easier for data to be processed.
Based on Mir’s description, the Autopilot rewrite allows Tesla’s computer to process images even more like the human eye now. The human eye usually processes pieces of data and sends it to the brain, which then puts everything together, so the human views the entire picture.
With the rewrite, Teslas will be able to process the information from all eight cameras and piece them together, resulting in a 360-degree picture.
“What the new rewrite did [was help lay out] all those cameras cameras’ [images] into one view and then made like a 360-view that helped you helped you map—where in the last program we were only able to do one to two—this one 1,000 to 2,000 [frames].”
He also shared that the Autopilot rewrite started with 3D labeling, which improved the accuracy of labeling each frame. “The videos playback to check the accuracy of the labeling. It’s still frames. However, it’s done in 3D, but the 3D just helps us able more. It’s still pictures stitched together."
Mir also shared that Tesla’s Hardware 3.0 was instrumental in the Autopilot rewrite and that Hardware 1.0 or 2.0 didn’t have the capabilities to process that much data. He strongly recommended that Tesla owners consider upgrading to Hardware 3.0, so the transition for upcoming software updates is smoother.
Recent news about Tesla’s Autopilot system indicate that the company has a few surprises up its sleeves. For example, there was the revelation that Tesla Autopilot would soon be able to stop for red lights and stop signs automatically. Based on an EV enthusiast who received early access to the feature, Tesla could roll out the new Autopilot capability in the next OTA update.
Mir said that Tesla owners may not always be seeing Autopilot updates in their vehicles, but the company is continuously improving the system. He agreed with Sofiaan Fraval when the Third Row Podcast member said, “Yeah, so just keep Wi-Fi on, right?”
In February, Elon Musk tweeted an open invitation for people with the skills to join Tesla’s AI team. The announcement and the new Tesla AI webpage proved that the company would be laser-focused on developing FSD in 2020. FSD, after all, is instrumental to Elon Musk's overall goal for Tesla, especially with the company's goal of deploying a fleet of autonomous vehicles that can earn for their owners through a ride-hailing service.
Featured Image Credit: TeslaFollow @PurplePanda88