>>13086863>>13086894Our computers are not too "weak" to do so. They are too weak to "brute force" it, and likely always will be due to the speed of light limiting single core performance. OP's crop briefly talks about parallelization, and indeed the human brain is an incredibly large number of neurons working in parallel. We need to create an AI that actually works and is parallel, i.e. multithreadable, otherwise our computers won't be able to run it.
Multithreading is a big deal in videogames, but it's something very hard to implement in practice. Let me give an example. I have a list of a billion numbers, and I want to find the biggest one. The straightforward way would be to tell a single thread to search through it and give me an answer. I could multithread it with X cores, however, by splitting the list into X equal smaller lists, have each thread find the biggest of their sublist, and then we have a list of X numbers that we can easily find the biggest out of. This is a very easy example, but this would likely end up being slower than a single thread. Why? Since we would need to pass all those numbers to all those different cores, and that is incredibly slow compared to the lightning fast ability of a single core to do some operations like X > Y checks.
Thus comes the issue with our recreation of the human brain. It is almost certain that a single core will be too slow to recreate the brain. And single cores have hit their limit due to the speed of light today. So we have to parallelize/multithread our AI. That's hard, considering we don't have much of a clue on how to create a single threaded AI to begin with. And we will likely need to create a specialized computer that behaves more like a graphics card, too, since the way threads currently work is not designed to facilitate the way neurons work, i.e., fast transfer of info to a handful of other nearby neurons. And even that won't mean it will be fast enough to pass lists of millions of numbers, of course.
