I guess Microsoft or Apple always knew 5 years in advance what computer processing power they would need to run their systems that were still in development and never had to re-evaluate or ever make a mistake?
Besides, what would you prefer: that they admit their mistakes and get it right in the end or never achieve FSD at all because they refuse to admit they were wrong or stubbornly persist in doing something that the computer can't do?
And I am not letting them get away with anything but I do want them to succeed.
What Apple and Microsoft were doing was evolving existing operating systems to do known tasks. The problem with FSD is it's an ever growing problem. It looks like a moderately difficult problem from the start, but once you get into it you run into more edge conditions than a knife factory.
I've done a fair bit of real time programming. Including some of the automation of commercial aircraft (I wasn't working on the code that went into the aircraft's systems, I was working on simulation systems to test those systems).
No computer actually reacts immediately, but it can react faster than a human if designed right, which is close enough to real time in most cases. When designing a system, you need to take into account all the variables. In an aircraft system, there are a lot of variables, but there are far fewer than with cars. With aircraft you need to be concerned about colliding with other planes near airports, mechanical failures, and human error. Colliding with other aircraft is dealt with in multiple layers, the first being ground controllers and secondly a system called TCAS where planes that get too close to one another communicate with each other and decide how to avoid one another. TCAS works well, but it depends on all the aircraft involved having the system. Something that won't exist with the first generation of self driving cars.
With cars, you have the above problems and added on top the possibility of collision is there everywhere, the humans involved have much less training than pilots, and the only upside is mechanical failure is not always as severe a problem with a car. Cars can pull over to the side of the road. But added on top of those problems, there are all sorts of road hazards like debris on the road, a broken windshield blinding optical sensors, people doing mischief like dropping bags of paint from overpasses, stuff falling off the truck in front of you, snow blocking sensors, sun glare blinding sensors, road construction changing the lanes, road lines that have faded or light conditions that make mirage road lines, and the list goes on and on.
The road environment we have today is designed for humans with human perception systems and human brains. Experienced drivers tend to be better than new drivers because they have learned to instantaneously parse irrelevant things and focus on the most important things. When necessary, humans can also move the sensors if the normal location isn't working right. We also have visors and sunglasses to help keep glare from blinding sensors at critical moments. On a car, the sensors are fixed.
Back when I first learned about AP1 and how Elon was predicting full self driving, knowing what I know about developing real time systems, my first thought was "good luck". More computing power will probably help with some of the problems they are encountering, but ultimately I'm not sure the problem can be cracked without changes to the transportation infrastructure.
As I see it, yes, FSD feels like a long way off now because all we have to judge is the current AP which is not even designed to be self-driving. So yeah, right now, it looks slow. But FSD development will move at an exponential rate. And as soon as we see big progress on AP, people will start re-evaluating FSD and predicting it will happen sooner. I predict that when V9 comes out, we will get a much better AP and people will start saying "hmm, maybe FSD is not so impossible after all" and then the AP3 chip will come out and some FSD features like stop sign recognition will come out and then people will go "hmm, this is starting to look promising, maybe FSD is only a couple of years off".
I'm not sure of that. We'll see.
