So from the latest excerpt some new information was gained:
By mid-April 2023, it was time for Musk to try the new neural network planner. He sat in the driver’s seat next to Ashok Elluswamy, Tesla’s director of Autopilot software. Three members of the Autopilot team got in the back. As they prepared to leave the parking lot at Tesla’s Palo Alto office complex, Musk selected a location on the map for the car to go and took his hands off the wheel.
When the car turned onto the main road, the first scary challenge arose: a bicyclist was heading their way. On its own, the car yielded, just as a human would have done.
For 25 minutes, the car drove on fast roads and neighborhood streets, handling complex turns and avoiding cyclists, pedestrians and pets. Musk never touched the wheel. Only a couple of times did he intervene by tapping the accelerator when he thought the car was being overly cautious, such as when it was too deferential at a four-way stop sign. At one point the car conducted a maneuver that he thought was better than he would have done. “Oh, wow,” he said, “even my human neural network failed here, but the car did the right thing.” He was so pleased that he started whistling Mozart’s “A Little Night Music” serenade in G major.
So FSD v12 was good enough to be be driven by Musk by mid april. Since then they have been improving the system until we got to see the demo in aug with one intervention in a 40min drive. I guess this is a bit disappointing since we thought the big leap happened more recently, but also shows that Tesla has been putting a lot of work into this and it was not such a risky demo as it seemed.
During the discussion, Musk latched on to a key fact the team had discovered: The neural network did not work well until it had been trained on at least a million video clips. This gave Tesla a big advantage over other car and AI companies. It had a fleet of almost 2 million Teslas around the world collecting video clips every day. “We are uniquely positioned to do this,” Elluswamy said at the meeting.
Here it sounds as if Elon only liked it because Tesla had an advantage this way. More like it's because he saw the potential of the system by extrapolating in his head what it would be able to do with even more data and compute. And it's not a million video clips that's needed, it's a million video clips of challenging situations where experts drivers have been driving expertly. They have all the failure cases from V1 to V11 and all the data they used to fix them to start with, that's their main advantage here.
My guess is that around a year ago or something a side project of end2end started showing primise they got more resources around the end of last year and by april they were confident enough to let Musk try it. Musk was sold and said that from now on, that would be their main priority. Then somewhere between april and the demo it was clear that V12 would overtake V11 on miles/intervention and around now it actually is better than V11 at driving but needs more validation before it goes wide(ie it may not require intervention as often, but it may have more catastrophic failures).
Overall it makes me more bullish for FSD but little bet bearish on the timeline until customers can try it.
