 | |  | |  | |  |
Read the following discussion:
www.kurzweilai.net/meme/frame.html?main=memelist.html?m=4%23688
Can you see how obscessed both humans are with discussing human this and human that?
They are likely both right, in a generous sense of the term. The most efficient way of
simulating a human brain will use specialized hardware and not a general computer. We can
already see that modern supercomputers are using multiple processors. If each processor
simulated a group of neurons, we could perhaps come the closest to a working model using
computers. But a better approach will probably call for specific hardware to simulate neurons
connected to thousands of other neurons and thereby gain the advantages of networking.
In his talk, Gelertner says: "Now, the reason I believe consciousness will never be built out of
software is that where software is executing, by definition we can separate out, peel off a
portable layer that can run in a logically identical way on any computing platform—for
example, on a human mind."
What he is objecting to is the separation of the software from the hardware. He has a point in
that the most efficient implementation of an intelligent machine is likely to be more like a
brain than like a computer with a single CPU. As the intelligent machine evolves to look more
like a brain, it is likely to be less universal and less easily reprogrammed than a single CPU
computer.
DPM
