The post-singularity computing power available to transhumanity, and the development of realistic simulations, have made the use of evolutionary algorithms as design tools cheap, easy and practical. Most products will go through, at minimum, a period of "refinement", using algorithms to simulate and evolve simulations of the product towards various metrics of quality. Many others are evolved even more than they are designed. This can seem like magic, but can also cause problems, most notably, not knowing how your own product works. Evolutionary inspectors go through the records of evolutionary processes generation by generation, breaking down exactly what developed and why.
Why did the successful line develop as it did? What was the most successful generation before overspecialization? Will the product perform as well in the real world as it did in simulation, or did the algorithm find flaws in the simulation to exploit? Are their adaptations that could become spin-offs, capable of enhancing other products? Did the product develop weaknesses that could be addressed by a designer? How is a simple adaptation able to achieve inexplicable results?
No comments:
Post a Comment