More cool work out of Jordan Pollack’s lab at Brandeis. They’ve build a nice force simulator and design system for Legos called EvoCAD that lets you evolve Lego structures that satisfy various conditions. Pablo Funes, who recently finished his PhD, used this to evolve spans of over 2 meters, tree-like structures, and counter-balanced cranes (cool video of it lifting 1/2 kg).
As pointed out by Think Artificial in their discussion of this work
…genetic algorithms result in … naturally messy constructs. This is actually one of the reasons that genetic algorithms are popular within artificial creativity research. Evolution results in various “unanticipatedâ€? or “surprisingâ€? constructs which can be perceived as creative.
Evolution, whether biological or “artificial” like that reported here, is an inherently messy business. It’s always forced to work with what it has, and is further hampered by the lack of any long-term plan or design. An incredibly powerful process, but unquestionably messy.
Thanks to Genetic Argonaut for the initial pointer.
