Am I the only one that feels like full self-driving is going to be a long, long way off? Every day I drive my Model S (which I love) and I use autopilot on the highway (which is useful), but I see things every day that make me say "full self-driving... I just don't see it happening anytime soon." I see exit lanes on the highway in rush hour where exiting traffic unofficially splits and makes a single lane two lanes... one for each direction... is my FSD car going to make me the a-hole and sit perfectly in the middle? I drive through my neighborhood which has no lane markings on the road, nothing at all... no lines to separate one direction from another and not even lines highlighting the edge of the road, and I think to myself... how would FSD work in this situation? Scan the road and realize which half to stay on? What if a car is parked in the street (which happens all the time)? What if I get to a 4 way stop sign and there is an aggressive driver who doesn't wait their turn, or a timid driver who starts going after I start? What if we go at the same time? When I'm making a turn through an intersection where there are two turn lanes and I'm in the inner lane, but the folks in the outer lane don't go wide enough and the inner lane cars have to drive over the lines separating the road? Is my FSD car going to try to squeeze them over to make them be perfectly in their proper lane? What if I'm on the highway and need to exit but no one opens up space for me to get in? Personally I might try to force someone to let me over by inching very close to the gap between one car and the next hoping that they back off a bit... how would FSD do that if no one budged?
These are just a small amount of things I see on a daily basis. 90%, maybe even 95% of driving is easy... it's that 5-10% that's tricky where you have to bend/break the traffic rules a little bit to deal with something... how is FSD going to deal with all of that? Maybe if most cars were all FSD and they "talked" to one another it would make it a little easier, but obviously that's not the case.
Not just Tesla, but a lot of companies want to see FSD come into reality. For a lot of companies, it would mean they could get rid of a lot of employees (creating a lot of unemployment). The automotive industry see self driving as the big thing that will generate a bonanza of car sales (mostly to fleets) so they want to see it too.
However there are two huge hurdles to get over. First the software needs to deal with a plethora of edge conditions, some of which are very tricky. There are legions of programmers trying to solve those problems in many companies right now. Then there is convincing regulators, who are naturally conservative about change, especially radical change, to accept that enough edge conditions have been covered that FSD with no human at the controls is significantly better than what a human can do.
What Tesla's AP and the GM system on the Cadillac CT6 do is among the easier problems to solve. Self driving when you can track a car in front of you (stop and go traffic) and open highway are both relatively easy to deal with. Tesla's approach is different from GM's. GM's system relies on complete mapping of every inch of every road where the system is certified. Tesla's system tries to figure out what's going on around the car and make decisions based on that.
There are pluses and minuses to both approaches. GM's is more reliable where the data is good. Hackers could screw everything up by altering the data the cars rely on. I assume GM's cars download chunks of route data so the car doesn't need to be connected 24/7, but it does need to check in to get route information because roads can change.
Tesla's system deals with the road the way it is the day you drive it, but the sensors can be blinded by the sun or other weather conditions. There are also holes where AP decides to do the wrong thing.
Self flying aircraft have been around since the 90s. Commercial airliners fly from point A to B with the pilots just sitting and watching the computer work. They often take over for for take offs and landings, but the automated systems can handle that too in many conditions.
But the edge conditions with commercial aircraft are significantly fewer than with land transport. The only time aircraft tend to be close enough to collide is near airports and there are multiple layers of systems to keep them apart. If something does go wrong, all commercial pilots are significantly better trained than the average car driver and they run through many disaster scenarios.
I'm not sure FSD with no driver to take over will ever become approved universally approved. Commercial aircraft are much further along in the tech and they still require two pilots.
Car makers have been pushing hard to replace rear view mirrors with cameras for over a decade and so far they have only been approved in one jurisdiction. And using cameras instead of mirrors is far more minor than FSD.
It is possible that fully autonomous cars with no driver on controlled, limited routes might be approved, but being able to do FSD everywhere might not.
The future doesn't always look the way people predict. Back in the 50s there were lots of predictions of flying cars, and it never happened because it turned out to be more difficult to do in any kind of cost effective way.
