some still underestimate how hard this problem (FSD) is to solve. And under appreciating how much progress has been made/rate it continues to be made. Consider this: VAG (Volkswagen Auto Group) with the Taycan and Audi Etron...arent even TRYING to truly solve this problem. Those cars (costing MORE than a comparable Model S) have smart cruise on highway, but zero city self driving capabilities now nor in the near future. And if I recall, GM's Smart Cruise is highway only and not even ALL highways so um, yeah
There’s a difference between “we can do it” and “we can do it by next Tuesday and it’s gonna be simple” and I know your smart enough to know that, silly.
Something I misunderstood about the rewrite initially is that I thought it was a one-time event rather than a long, gradual process. Turns out a lot of the software is still in the process of being rewritten. Software development is like Zeno's paradox where you get closer and closer to finishing a project without ever actually crossing the finish line.
Not just that, but i’m smart enough to know you are making up stuff - “we can do it by next Tuesday and it’s gonna be simple” isn’t indicative of their actual plan.
I don't want Tesla to roll it out until FSD beta can also find parking and park for me. That would be the ultimate FSD satisfaction: to have the car drive all the way to the destination and park without any interventions.
You need to branch out your search. Don't look at the traditional automakers for tech, look at the EV startups. 42Marks does one of the best independent testing of ADAS in china and they did one for Navigate on Autopilot, Navigated Guided Pilot and Navigate on NIO Pilot. The results show these companies have caught up and over took Tesla's NOA in china in alot of scenarios. If they were an American company, it would be the same also. Watch Video using closed caption:
Do they run city NOA beta ? Would be interesting to watch. China has a fantastic pool of data scientists and cheap labelers.
By the end of 2021, 2 companies (Geely's Lync & Co - Zero and Allibaba's Zhiji Auto - IM) will both likely also have point to point (door to door) systems, with Geely most likely because its using SuperVision and Zhiji saying its system is pending regulation.
Thats great info, but doesnt refute my original statement, which also had a scope of vehicles approved for US roads (my reference to GM, VAG, Tesla). But right now, today, even with the deficiencies in FSD, I can honestly state that a daily trip 5 miles away from my home to the local coffee shop (no turns, several curves, multiple stoplights) is quite enjoyable and relaxing with FSD enabled and without having to touch the steering wheel except to turn out of my driveway and into the parking lot of the shop. Im not sure any other carmaker can provide that service to me today, in the USA.
That is not a good indication of GM's progress on FSD. Super Cruise is just a L2 ADAS system. GM outsourced their FSD development to a child company called Cruise Inc that is developing their FSD. And Cruise has L4 driverless robotaxis in San Francisco. In fact by most metrics, Cruise is #2 in the FSD race, only behind Waymo. So GM does have really good FSD in development. I would assume that once Cruise's FSD reaches a certain level that GM could start putting it on consumer cars.
Everyone knows when dealing with quantum leaps that there are no gradual steps. We just lurch to the next electron orbital energy state. So we’ll move to the next open principal quantum number, possibly with a new azimuthal quantum number. That’s what the version numbers mean I think. Another key thing is you cannot determine exactly where we are in this process, because otherwise we would not know how much momentum we have towards accomplishing the goal! Yet: If we determine we have a lot of momentum, we won’t know whether we are close, because we won’t know where we are! So the quantum leap descriptor is highly appropriate.
I think they don't want to expand in Winter countries because the system is not ready for snow roads covered.
Evolution is another good analogy. Sometimes, evolution will take you into a dead-end, cul-de-sac. A "local minima" (or maxima, depends on how you are measuring). You can't evolve your way out of that, it's a dead-end. But there is no way to know its a dead-end until you get there, AND you have awareness of evolution. That branch, then gets stuck at the end. It can't 'devolve' back the way it came, only way out, is if the environment changes around it. If the environment doesn't change, only way forward, it to start the journey again, from a different start-point, and make different choices, based upon knowledge gained in the first journey ('God View, Directed Evolution'). Different Species, Software re-write. The new species will almost certainly be better, based on the knowledge gained. BUT will it be enough to jump to the next level? Can't tell until you get there, the first time. Been in the software development industry for many decades. Thankfully, NO software system is ever 'FINISHED', (deliberately in capitals). It can always be improved, added to, made more reliable, faster etc. etc. It gets 'done' when it's good-enough to be useful/profitable, and the unwanted effects (bugs) it exhibits don't out-weigh it's usefulness.