On top of that, if an AI created a device that no science can adequately explain, there's ten different reasons why this special device would not be crammed into consumer electronics.
Not quite convinced, consider the work that's been done on genetic algorithms and evolved circuits - eg Thompson http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.50.9691&rep=rep1&type=pdf . There's a possible line of development there.
Thanks for the link, JimC! Good reading!
My former coworkers at Nokia told me that story. When they reached the punchline in the end, and told the bit about how removing totally disconnect parts of the circuitry made it malfunction, I recall being impressed at first and a bit skeptical later. After all, it would have been very much in character for them trying to put a fast one by a math guy largely ignorant about EE. Not unlike other rookie tricks: "remember to lube the muffler bearings" or "you go fetch the keys to the trebuchet shooting range".
See also
http://mitpress.mit.edu/sites/default/files/titles/alife/0262290758chap70.pdf Using Meta-Genetic Algorithms to tune parameters of Genetic Algorithms to find lowest energy Molecular Conformers
ZE Brain, MA Addicoat
Proceedings of the 12th International Conference on the Synthesis and Simulation of Living Systems
I'd given the simplified poster for a lay audience unfamiliar with the problem domain before.
https://www.researchgate.net/publication/275953662_Meta-Genetic_Algorithms_Molecules_and_Supercomputers_PosterMeta-Genetic Algorithms, Molecules, and Supercomputers
MA Addicoat, ZE Brain
Poster, SC10 Conference on High Performance Computing, Networking, Storage and Analysis
My PhD topic.
Genetic and Evolutionary algorithms have already been used in such areas as antenna and rocket nozzle design, where theory is still imprecise.
As long as we can determine that solution A is better than solution B, and there's a way of encoding A and B that has a close mapping to them during evolutionary generation of solutions, we really don't need to know exactly why they work, we can optimise anyway. Knowing what the optimum is then gives us clues into the mechanisms involved.