Eugen Leitl <eu...@leitl.org> writes: > On Thu, Jul 19, 2012 at 02:28:18PM +0200, John Nilsson wrote: >> More work relative to an approach where full specification and controll is >> feasible. I was thinking that in a not to distant future we'll want to >> build systems of such complexity that we need to let go of such dreams. >> >> It could be enough with one system. How do you evolve a system that has >> emerged from som initial condition directed by user input. Even with only >> one instance of it running you might have no way to recreate it so you must >> patch it, and given sufficient complexity you might have no way to know how >> a binary diff should be created. > > It seems a great idea for evolutionary computation (GA/GP) but an > awful idea for human engineering.
Perhaps not. The idea would be that we would design our systems not with hard boundaries and functionalities, but like living organisms are designed. Notice that all the program of an organism is contained in each of its cells. Cells behave differently depending on the environment (organ, tissue) there are in. If you migrate a cell to a different organ, it may start behaving differently. It can because it has all the programs. On the other hand, if you move a library to another place, it will just break because it doesn't have the programs needed in that other place. -- __Pascal Bourguignon__ http://www.informatimago.com/ A bad day in () is better than a good day in {}. _______________________________________________ fonc mailing list fonc@vpri.org http://vpri.org/mailman/listinfo/fonc
