Originally posted on TeslaOracle.com
Now it has been more than a week when Tesla released the Full Self-Driving (FSD) Beta to select customers deemed safe and cautious by the automaker.
Initially, the beta testers released short video clips of the FSD testing on city streets but this was done in a hurry. With each passing day, extensive tests are surfacing on social media platforms like Twitter and YouTube.
The latest version of the FSD Beta software is 2020.40.8.12 which has the latest improvements done within the last week since the release of the first version.
The team of FSD beta testers is pushing their Teslas in various road conditions and situations to test the capabilities of the latest release. Roundabouts, speed bumps, intersections, traffic lights, day and night testing are to name a few of these scenarios.
Let’s watch how the latest version of FSD Beta performs in different scenarios in videos posted by owners and social media influencers.
This testing by few select drivers is also giving a chance of learning to Tesla Neural Net and this totally re-written Tesla self-driving software AI. Multiple beta testers are reporting that by driving the car multiple times at the same spot, it actually gets better at that spot.
And this learning curve is not just for a single car, this learning data is sent back to the Tesla Dojo neural net learning program and the entire fleet gets the update.
Like the roundabout in the following video where @tesla_raj’s Tesla Model 3 learned from its mistakes and after the driver taught it how to do it correctly. Roundabouts seem to be the trickiest of all the challenges for Tesla Autopilot FSD.
It took only a day for Tesla Autopilot to learn a tricky intersection that it was failing at before — kudos to you Elon Musk you mad man and Andrej Karpathy the AI guru.