Andres125sx wrote: ↑26 Oct 2018, 09:30
Actually, when you compare humans and AVs, we are flawed like it or not.
...
If your argument is "AVs should pass some exaustive QC to ensure his software has no bugs" then I´ll agree, but arguing AVs can have some minor flaw so they shouldn´t be on the road, when we humans have several mayor flaws, is unrealistic
You are severely underestimating the ability of the human brain to process and anticipate complex situations based on feedback, instinct and experience.
Even with our 'narrow FOV', we are able to assess fairly well what is happening around us, even when your attention is focused on something else. Sure, how well we are able to do that, depends on various factors. But even in a highly impaired or distracted state, we are able to prioritize quite well. Reading your post makes it sound like we are all some kind of flawed zombies on the road. We're not. As I said, most people, even those that are involved in accidents, have most probably driven hundred-thousands of kilometers in their life-span. It's inevitable that something at some point will happen, either by mistake or an unfortunate chain of events of being in the wrong place at the wrong time. Those experiences shape us. Most of us learn from it and become better drivers for it. Most importantly, we're accountable. Those that don't learn, drive irresponsible, get fines, perhaps go to jail etc.
The problem with 'machines' is that it's effectively a piece of software that doesn't care, doesn't know responsibility. And I question the ability of technology to progress to the point where a truly autonomous vehicle is feasible and safe.
Here some situations we encounter in every day life on the roads, yet conscious or subconscious do a good job of assessing the situations properly:
Driving through a residential area. A potential for playing kids that could run out onto the road at any time. Cars, regardless if manual or AV don't have instant stopping power. At speeds of 10m/s (36kmh) there's still heaps of opportunity to instantly kill a small child walking into your path. Now, an AV might do a better job at registering a subject walking into your path and braking with the perfect amount of force for optimal deceleration. True. But I question the cars ability to properly assess the situation around it. I haven't counted the number of instances where I've seen through the gaps of the cars a bunch of kids playing or their trajectory pointing towards the road even before the 'kid' jumped out between the cars. Or spotting the ball before rolling across the road moments before. No matter how good a 'machine' is, I sincerely doubt its (programmed) ability to properly analyze this information and assess the situation of what is about to unfold seconds into the future.
As a human, I have that experience in my subconscious of having been a kid myself once upon a time or simply having been in many other similar situations where I did see something like that unfold. This experience and instinct helps me to anticipate and react to something that hasn't even happened yet, even if it's just a subconscious cue of slowing down or applying the readiness to brake.
Another example. When I see this, I am not only processing two 'objects' walking towards the road. I see they are not looking, possibly distracted. I don't doubt that technology can easily distinguish 'people', but I am doubtful over the ability to properly assess situations and anticipate what is going to happen, because a machine isn't human and visual cues can be very subtle. Could a machine distinguish a handicapped person from a drunk? Differentiate a person who is focused on what is happening around them before i.e. crossing the road or being distracted, in an argument, on the phone, writing a text?
This is important and what makes us anticipate situations accordingly. We might see a situation unfold before our eyes without it even happening it. For example; A cyclist driving down the street, while a person is about to cross the road while being distracted by texting on the phone. Meanwhile, we see the cyclist about to turn his head to the back because someone has called after him. Seconds later, the two collide because the cyclist did not stop and the person walked into his path. Meanwhile you are in your car just meters behind following the cyclist. A person could quite easily see this happening and react to the potential danger before it even happened, while I question the ability of a machine to do the 'math' and anticipate correctly what is about to unfold. Now I am not saying that 'humans' will always get things right and anticipate situations, but even when they don't, there's a lot happening in your subconscious that you may not register but is still there. Instinct.
I look forward to seeing how the IT world will solve this, without having to study and program the knowledge and experience of 20+ years of human behavior. And we are just talking humans here. What about animals? Objects? Debris on the road. The knowledge of knowing there's an imperfection in the road surface at that particular spot because you have driven that road a thousand times in your life? How can we expect an AV to store all that information?