Cycling aside, to reiterate something I said in another thread, I would temper expectations about FSD being useful in the UK any time soon. Yes, it may be legally allowed, but I doubt Tesla's implementation is going to be functional over here in the near future.
My understanding of machine learning (as an IT consultant, but not an ML specialist) is that you present a neural network with a massive set of input data, and a set of 'correctness' criteria. It then undergoes a process similar to natural selection, where it tries various configurations of that neural net and retains those configurations that score higher on the correctness metric, and discards those that score lower.
You end up with a system that no-one understands - you can't look at a neural net and 'read the code' like you can with something programmed by humans. It's almost like a chaotic or complex system, in that semantically different inputs will yield wildly differently results. So getting FSD to work over here won't be as simple as flipping the X-axis, because UK junctions don't work like US ones.
I've not seen any evidence that Tesla are prioritising learning non-North American road systems. If anything at all, there's evidence to the contrary in that EAP works better (anecdotally) on the continent than in the UK, and after having recently completed a 3,000 mile European road trip, I'm part of the anecdata.
I'd love FSD to be a useful thing. I'd hate for people to spend money on it, expecting a result any time soon.
