Re: Zen and the Art of Flying Saucer Maintenance

Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Sun, 13 Jul 1997 11:38:40 +0200 (MET DST)


There is a complexy threshold for man-made systems, which makes them
progressively expensive to operate. There also seems a pretty sharp
boundary where the system is suddenly _unable_ to operate at all. Many
aerospace designs and large software projects hover at that threshold.

This needs not to be so, however. Naturally evolved systems do not show such
marked brittleness, and their occasional failures can be attributed to
random mutation noise crucial for the darwinian optimization process to
function.

Systems which can tolerate defects are comparatively easy to do. As
next generations hardware is likely to feature many 10^9 active elements,
their production error and failure error rate must be counteracted by
redundancy to make them both affordable and operable. WSI FPGA and CAM
EHWs are obviously a step into the right direction. Langmuir-Blodgett
should give us 2 1/2 d-systems, whereas true 3d integration is only
achievable with crystal growth autoassembly.

Self-healing systems at microscopic module scale are only achievable by
means of molecular manufacturing, whether machine or solvate phase. Don't
expect them too soon.

'gene