A lot of the AI debate here is misguided.
1) The discussion on AI here is similar to how people imagined we would be flying on jet packs to work in the year 2000.
2) Hard AI, is about as achievable as warp gates or faster than light travel currently. We had/have to figure out how to make something understand the concept of time for example, and time doesn't mean a clock counter keeps going up.
Any and all discussions which assume that AI will take over, have conversations, act in even remotely anthromorphic ways, come from a position of speculatory hope. In all my readings, AI researches instead hope to be able to create even one of the subsets of our ability to think.
We've just been cracked enough processing and coding power to model a single neuron in the past 365 days. A full brain is some decades away. And that would get us only to sub human intelligence.
1) This is true to some extent, we are speculating about a technology that hasn't been fully explored yet, we don't know what technologies will arise in the future, or how all this will pan out.
2) A) Comparing it to faster than light travel is a bit ridiculous, that's something we have absolutely no clue how it could ever be done, this is something already on our way to doing.
B) I have absolutely no clue what you're saying about understanding time, do you mean it has to have the ability to perceive time? You do know that the human brain can't actually keep track of time without external clues right?
3) Being able to talk and "act in even remotely anthromorphic way" is sort of really necessary if you want it to do anything pragmatic apart from just sit in a simulation pretending to be a turtle or something.
4) As for the neuron stuff, firstly we're already pretty close to simulating
a rat brain, and the people behind it say they think they can do a human one in ten years (I think that's probably a little too optimistic, but whatever, it's closer than you seem to think).
Secondly it's not necessary to simulate a neuron to achieve intelligence, at least not fully. Simulating a human brain on a supercomputer like this is honestly kind of a hack solution, we can do something much, much more streamlined than this by abstracting the processes of neurons more, or making use of microprocessors to do a lot of heafty parallel processing. Or heck take steps away from basing our work off standard processors, a neuron is essentially just a rather complex logic gate, change the way we work with processors and I'd bet we could do something much better.
I'd also like to point out that while a lot of people have been saying that AI's aren't going to be a thing and that enhancing human capabilities will be the wave of the future, nobody seems to be coming forward with ways to actually do this, which sort of suggests that we're a lot closer to creating a super-intelligence based off of electrical computing than we are based off of anything biological.