If you are going to talk about statistics then please cite the accident statistics of people using things like FSD vs. not...
From the horse's mouth it appears using things like FSD is considerably safer than not[0]. While an independent statistic would be nice, I couldn't find one. Would appreciate seeing your source
I used your metrics -- a few (let's say 3) interventions over 600 miles.
Tesla doesn't publish the relevant stats, probably because they're so outrageously good and strong and perfect that the public can't handle how amazing FSD works in reality.
Edit in response to your edit: You're showing me Autopilot data which 1) is a different system and 2) is only useful on highways, so even if it did nothing at all you'd expect to see a massive reduction in frequency of accidents when it was "using Autopilot." Given that Autopilot is mostly ADS features, you'd also expect a large reduction. The better comparison point is "United States recent luxury model highway miles." Distance sensors, lane assist, auto-braking, etc all very obviously improve safety.
Tesla doesn't release raw data, hard to read anything into their claims either way. I do suspect though that if the raw data was beyond reproach then there would be a zip file with CSVs of it in them for third parties/academics to go over.
I've used FSD as a subscription from time to time. I can't think of an intervention (and there have been many) that was to avoid an accident. They were all things like "that's closer to those cones than I'd like to be," "okay, apparently missing lane markers that have yet to be repainted is understandably hard," or "that was a trailer speed limit sign -- we're not towing so now we're going way below the posted limit."
The first two examples, in most scenarios, would yield an accident if unrecovered. The lattermost is also quite dangerous on highways (because driving way below posted limits causes accidents).
A lot of my interventions when trying it out were things like it going too close to curbs for my comfort, so more violations of my own defensive driving standards rather than potential accidents. But! It did have some near traffic violations that would have happened if I hadn’t intervened.
But the distance between my own driving standards and accident is large.
Do you have an exact read on where the wheels on the other side of the car are relative to a curb when you're turning? I don't, so I give it a healthy error margin, where I'm certain it won't. I'm reasonably confident it wouldn't have made contact without my help, but that's not good enough for my driving standards. That mismatch was consistent enough that I ended up not using FSD anymore, it was more stressful than not using it at all.
Also, tire-puncturing junk tends to build up next to curbs, so there are other reasons for my margin preferences than just avoiding accidents.
There's a pretty big spectrum between perfect driving and a collision. FSD could drive on the wrong side of the road for miles with no accident because it's an empty road, or oncoming traffic swerved to avoid you.
That report is pretty much worthless and fails any sort of statistical rigor. It makes many apples-to-oranges comparison and doesn't control for variables such as road type (highway/city), geography, time of day, age of cars and safety features.
Tesla refuses to provide raw data like the other self driving companies do to avoid scrutiny.
From the horse's mouth it appears using things like FSD is considerably safer than not[0]. While an independent statistic would be nice, I couldn't find one. Would appreciate seeing your source
[0]: https://www.tesla.com/VehicleSafetyReport