In its aggregation of the driving data from vehicles for training FSD, does Tesla incorporate data from drivers who do not have FSD or Autopilot engaged? This could provide valuable input for machine learning in particularly difficult driving situations.
I watched this video of a tester trying Beta 10 in a turn under Seattle monorail pillars:
. The beta tester tried the turn a few times, but each time he panicked and took control when it appeared that the car was heading toward a pillar. The driver has repeatedly driven this route manually. I expected the car to learn when the driver took took contro, so it would do better the next. But it appeared that the car made the same mistake each time.
Tesla would have to ask a user for permission to collect this data, but I am sure many drivers would be delighted to contribute to the cause to improve FSD.
I watched this video of a tester trying Beta 10 in a turn under Seattle monorail pillars:
Tesla would have to ask a user for permission to collect this data, but I am sure many drivers would be delighted to contribute to the cause to improve FSD.