Surprising examples of non-human optimization

I am very much in­ter­ested in ex­am­ples of non-hu­man op­ti­miza­tion pro­cesses pro­duc­ing work­ing, but sur­pris­ing solu­tions. What is most fas­ci­nat­ing is how they show hu­man ap­proach is of­ten not the only one and much more alien solu­tions can be found, which hu­mans are just not ca­pa­ble of con­ceiv­ing. It is very prob­a­ble, that more and more such solu­tions will arise and will slowly make big part of tech­nol­ogy un­un­der­stand­able by hu­mans.

I pre­sent fol­low­ing ex­am­ples and ask for link­ing more in com­ments:

1. Nick Bostrom de­scribes efforts in evolv­ing cir­cuits that would pro­duce os­cillo­scope and fre­quency dis­crim­i­na­tor, that yielded very un­ortho­dox de­signs:
http://​​home­page.ntl­​​r.stow1/​​jb/​​pub­li­ca­tions/​​Bird_CEC2002.pdf (IV. B. Os­cilla­tor Ex­per­i­ments; also C. and D. in that sec­tion)

2. Al­gorithms learns to play NES games with some eerie strate­gies:
https://​​​​qXXZLoq2zFc?t=361 (de­scrip­tion by Vsause)
http://​​hack­a­​​2013/​​04/​​14/​​teach­ing-a-com­puter-to-play-mario-seem­ingly-through-voodoo/​​ (more info)

3. Eurisko find­ing un­ex­pected way of win­ning Trav­el­ler TCS strat­edy game: