X-Message-Number: 25138 Date: Tue, 23 Nov 2004 05:53:31 -0500 From: Thomas Donaldson <> Subject: CryoNet #25130 - #25137 FOR Yvan Bozzonetti: The problem with FPGAs is that they're far too small and lack all the connections needed. If we were to try to implement a brain with FPGAs, we'd set up one for each (overlapping) set of neurons which might form connections depending on the information they receive. The number of ACTUAL connections will always be limited, but the number of POSSIBLE connections will be much larger; and note that we'll need at least one for every neuron, and probably more than one because neurons communicate by other than electrical means, too. Moreover the range to which a neuron can grow a connection is fuzzier and generally larger than the range to which it has actually grown connections. This means that if you want your FPGAs to construct a brain, they'll have to contain potential connections between each neuron at a much larger range. (One way they're too small). And remember that an FPGA isn't the neuron itself, it's aimed at producing the CONNECTIONS of a neuron. Again, it's essential to the working of our neurons that their connections are changeable and none of them must persist. If I understand how FPGAs work, we can program them once but not more than once. So that we'd want changeable FPGAs. This is why I do not believe we now have the technology to imitate a HUMAN brain. As I said before, given especially that neural connections are NOT fixed, it looks to me that growing connections makes a lot more sense than beginning with a device containing all possible connections and then pruning them down. And it makes more and more sense the bigger the brain becomes. Besides, just what's your prejudice against devices that act biologically, anyway? A variable FPGA might turn out to be a quite valuable device, even if it's not made with silicon. Best wishes and long long life for all, Thomas Donaldson Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=25138