I 100% firmly believe the singularity is a myth. It will never come to pass.
me
A singularity is not required to learn how to drive.
How does this in any way argue against anything I said?
Presumably you were just mentioning that apropos of nothing, not having anything to do with the conversation about self driving cars? If so, how was I supposed to know that?
You have high expectations from computers.
I'm just acknowledging the relationship between computing capability and time.
I have a love/hate relationship with computers.
I don't think so. I think you have a love/hate relationship with programming. Perhaps more specifically, programmers. I think you have a love/hate relationship with people.
Honestly, you're up against a mountain of experiences over a lifetime.
I shouldn't be. Your lifetime of experiences should easily show you how non-linear the development of computing technology is. Think back 10, 20, 30, even 40 years and consider what computers absolutely could not do then, and what they can do now. Your lifetime of experience should be the thing that leads you to the conclusion that computers are working their way into every facet of human development, very rapidly.
I will believe an AV can navigate any and every road by my home, year-round, in all weather, without incident, when I see it (not from inside the car). I was under the impression that such a feat is currently not even in the scope of AV development (and that's fair). Ergo, AVs are more realistic for urban use. Which was the gist of what I said in the first place.
It would be trivial to do with today's technology. A sophisticated approach would be to create an ultra precise navigation environment (perhaps using double or triple differenced GPS signals), in which a sufficiently outfitted vehicle (such as one with snow tires or tank treads) could navigate any weather conditions would be able to do so completely blind (I mean with no cameras) using the GPS signal alone.
A less sophisticated approach would be to spend many hours training vehicles driving around your home in all kinds of weather conditions to build up the machine learning needed to get around without a driver.
I think what you meant was to do it with one arm behind your back - meaning that you want a self driving car to show up with no prior knowledge of the area and do it in real time by imaging alone. And while that might not be necessary, there's no reason to think that computers would be incapable of accomplishing the task at some point. Certainly it's not in the near term for self-driving capability, and my guess would be that such a problem would be one of the last things that self driving algorithms would manage. However, that'll be where it's headed.
At first, automation might limit itself to urban areas with well defined roads and signage. It'll stop and wait for a driver (perhaps a remote driver) when it reaches conditions that it considers to be too challenging. But it can learn from that driver, and with enough experience, an environment that used to be too challenging has had many examples of a driver navigating through it, and then the driver is no longer needed to intervene.
That kind of behavior might be exactly what tackles some of the dirt and gravel roads that get used in rural areas. You may even end up showing an algorithm how to do it yourself.
I should point out here that computers have some inherent advantages in these kinds of conditions. Computers can manage some pretty complex math based on imagery that human brains cannot. A couple of distant trees can triangulate where a road usually is (but is buried by snow), and for a human this might be guesswork. A computer can figure it out quite precisely.
It might be challenging now, but it's easy to see why a computer can end up being actually far better at the job than you at some point.
And yes, last time I checked, the human brain is still mysterious to us. We're nowhere near decoding it yet. Forgive me for a little rhetorical hyperbole with "unimaginably".
The human brain is a bag of meat and chemicals. It is a biological computer, honed by trial and error in much the same way as machine learning works for computers. Our brains have not worked out every last nugget of understanding about our brains, but our brains are quite imaginative, even imaginative enough to imagine the complexity of our brains. Remember, the more complicated our brains, the more complicated a brain we can imagine.
I think it's a little odd that you pick driving (presumably on dirt and snow) as being something that is unreachable for computers. There is nothing about that problem that seems insurmountable. It's definitely not the most difficult thing we do with our brains, and computers have shown quite a lot of promise in the area of driving
specifically. There are other problems, like writing a movie or novel, or creating a joke, or creating an emotionally moving painting, that seem far more insurmountable for computers. Yet I can't tell you I'm supremely confident that any of it is out of reach - not because I have high standards for computing, but because I recognize the limitations of the human meat brain, and because the history of computers suggests that relatively short timespans bring about wild changes in capability.