At this point, the test group is getting acquainted with the technology and apparently "teaching" it. Think of a 15-year-old driver getting their hours in so they can eventually get their license. It's a nail-biter for the parents, and it's not really safe, but it's necessary. FSD is similar in many ways. Perhaps these cars should be required to have a variation of the "student driver" sign.
In addition, Tesla hasn't really revealed what the process is here. Is the system actually learning as it drives? Are these testers really supposed to "teach" it? What happens when they report something to Tesla? What is Tesla doing with the reams of data it's collecting from this process? We would be lying if we said we could answer these questions honestly and with all the facts.
With that said, these recent, short video clips from Jame Locke attempt to give us some understanding. Above, Locke shows his Tesla learning to recognize and slow down for speed bumps in a residential neighborhood. Below, we have two demonstrations of the Tesla dealing with roundabouts. There are two roundabout attempts, in order.
You can see how cautious the system is in the previous three videos above. However, adding to what we said above, we'd like to know if this "learning" experience immediately makes its way to all cars. In other words, when Locke "teaches" his car to observe and slow down for speed bumps or to successfully navigate a roundabout, does that information forward to all FSD Beta cars and make them successful at the two operations?
If these cars are actually learning and then sending the new information to the fleet, continuous improvements are inevitable. This will prove especially true once tens of thousands of people are using the feature. However, what happens if people teach it wrong? What happens if the car gets into or causes a collision?
We're happy to see progress, but we still have so many questions. We'd love to read about your takeaways in the comment section below.