Evolutionary Design or How I Learned to Automate Innovation

As human-computer interaction becomes richer, we're asking computers to take on tasks normally reserved to those with biochemical substrates for thought.  There has been a movement under foot to see whether or  not computers can innovate... can be creative in ways similar to humans.  Its called evolutionary design.

In evolutionary design computers create thousands of variants representing design solutions until a best-fit solution to a design problem is found.  Its a brute force trial and error approach but done at speeds that human teams can't replicate. In essence, evolutionary design takes humans out of the design equation.  It starts with humans but with enough iterations on design problems eventually one could conceiveably factor humans out of the equation completely.  Two initial "parent" designs are provided as starting points to a design modelling program.  The program then takes design elements from each parent design and recombines them in unique ways.  Take a pair of sneakers for instance -- the modelling program could take the ventilation pattern from one sneaker design and the sole waffle pattern of the other parent and blend them into a unique offspring design. In fact, thousands of offspring designs can be produced that combines features of each parent to come up with new more evolved designs.  The program then selects the offspring it thinks are worth a detailed look at and the process is repeated countless times to see if breeding with other promising offspring comes up with even better results.

Question... is design recombination though a form of creative thought?  Would I ever come up with something original that in some way or another is not related to the parent design but is disruptive (in a good way) in its own right? Perhaps.  Perhaps offspring can evolve to the point where the designs are dramatically different from the original parent designs but I think evolutionary design can be leveraged to drive incremental improvements.  But if you  are looking for the next best thing since sliced bread, is this gonna get you there???

The hope here of course is that after thousands of "generations," the modeler will create some design combinations that might never have occurred to a human.

Traditionally  evolutionary design was found primarily in the aviation and auto industries but now is moving into a wide range of other industries  and device designs--  USB memory sticks, yacht keels, high bandwidth optical fiber, Wi-Fi antennae, cochlear implants... and the list goes on. What's making evolutionary design available to a wide range of industries? Why Moore's law of course.  Although evolutionary design has been around for years, cheap processing power hasn't. Its needed since it can take up to 20 million generations before a usable design is created. And what used to take months of design simulations now happens in a matter of days.

Its even being used to design around patents. Stanford design students were hired to design a Wi-Fi antenna for a company that wanted to avoid paying Cisco fees for their antenna  design. They used evolutionary design to come up with a wi-fi antenna that didn't use any features covered in the Cisco patent and purportedely is more efficient as well.

All of this begs the question, just where is the line between humans and machine intelligence when it comes to design?  Are humans needed to produce designs?