I think it's still an open question whether a computer can simulate a mind. Godels incompleteness theorem is often cited as an obstacle in doing this - but there is disagreement on how that actually relates to this.rbarreira wrote:Are you saying that your brain does not follow the laws of physics? If it does follow them then it's very hard to argue that your consciousness / intelligence / anything else happening in your brain is not reproducible.Daniel Shawul wrote:Turns out article I read is quite old. Rat's and Cat's brain have been solved.
the neurons maybe, but not consciousness.
cmputers are deterministic, a humans brain is not.
at least mine is not or at least i believe its not.
irrespective of my writing style.
which i learned from Blik(C) btw on ICC
As for being deterministic or not, that doesn't matter, not all computers are deterministic anyway.
That does not mean an AI is not possible that is comparable to mans. But the brain is NOT a digital computer, so it would be incredible inefficient trying to simulate one with a digital computer even if it were possible. Perhaps a human-like intelligence will be possible in the distant future that operates by different principles. Sort of like how a horse and an automobile both can transport objects but by significantly different means.
I think the practical issue with Godels's theorem, correct me if I'm wrong because I'm not an expert on this, is that the human brain is not capable of understanding itself and how it works - it cannot describe itself and neither can any formal system. We can understand some high level aspects of ourselves but not the low level details necessary to re-create ourselves. I don't know if that precludes building one based on a "simple" copy without actually understanding what we are copying or not.
One thing is pretty clear right now. We don't know how to build a smart computer and no matter how much CPU power we were given the issue isn't the power, it's how to write the code. It would not surprise me if your desktop computer was enough to do some impressive things in AI if only we knew how.
So this is not like chess, where the faster the computer the better it will play. Think of it more like chess where mini-max and alpha-beta are not known - trying to make it play strong without even knowing how to do a search. Such a program would be a horrible mess of more and more rules and having 1000x more rules might make it play 2 ELO stronger.
Don