Before January 2020, Tesla will have gathered more than 2 billion miles worth of real-world driving data for Autopilot. If MIT researcher Lex Fridman’s estimates are correct, Tesla would have doubled the amount of data it has gathered in less than a year. The company has made strides over the years in the world of autonomous vehicles. Every year that passes seems to bring humanity exponentially closer to full self-driving cars.
Tesla has not officially announced how much data it has gathered from Autopilot. However, Tesla enthusiast Fridman crunched the numbers and came up with some viable estimates based on his personal counter. Fridman doesn’t mince numbers. He is an MIT researcher who specializes in human-centered AI, deep learning, autonomous vehicles, and personal robotics. In other words, he understands Musk-speak, especially when its about artificial intelligence and the future of robotics.
Based on Fridman’s estimates, Tesla Autopilot miles could be hitting 2 billion real-world miles by November 11, 2019. Based on that projection, he estimated that Tesla would hit over 2.19 billion miles worth of data on Autopilot by 2020.
It took 4 years for Tesla to gather its first billion miles of data, but less than a year to double it. If Fridman’s calculations are correct, then Tesla’s full self-driving suite is growing more efficient and could be ready sooner than expected.
Contrary to Fridman’s numbers, ARK Invest’s Cathie Wood estimated that the Tesla Autopilot has already gathered 10 to 12 billion miles of real-world data based on ARK analyst and former Nvidia employee James Wang’s observations. Wang declared that Tesla’s AI chips are at least four years ahead of all its competitors in the autonomous market.
However, it seems Fridman’s calculations were accurate since Elon Musk liked a link to a page on his site where he discussed his estimates. Musk usually corrects wrong information, so his like was sort of like a confirmation stamp to the Tesla community.
Although, in ARK’s defense, Wang may have counted data gathered while Shadow Mode was enabled in his estimations. In Shadow Mode, Autopilot is off, but the cameras are still on and can still collect data on the road. Tesla’s full self-driving feature relies on the vehicles’ eight cameras, 12 ultrasonic sensors, radar, and GPS to gather data for the Neural Net. The whole system works much like the human senses and the brain.
Either way, though, a billion or two billion is a significant number of real-world driving data for Autopilot. It is leagues away from other self-driving companies, like Waymo—which incidentally has only collected a 10 million mile worth of real-world data as of June 2019.
With a billion worth of real-world data, Tesla is poised to lead the autonomous transportation service business in the future. In a 2019 informational article, ARK Invest estimated autonomous taxi revenues could reach US$8 trillion by 2035.
Currently, the investment firm placed a US$2 trillion-value on the Autonomous Taxi opportunity. In a CNBC interview, Cathie Wood forecasted gross margins well into the 80s for Tesla once it establishes its autonomous fleet. She also predicted TSLA stocks would reach $6K per share, partly because of its AI chip and Autopilot’s real-world data.