The Challenge of Producing Human-Competitive Results by Means of Genetic and Evolutionary Computation

Abstract

Human-competitive results include those equivalent tonew scientific results published in peer-reviewedscientific journals, solutions to long-standing orindisputably difficult problems, patented inventions,and results that tie or beat human contestants inregulated competitions. We argue that the pursuit ofhuman competitive results is not only a worthy goal initself, but a useful compass for guiding the futuregrowth of the field. We say this for reasons ofutility, objectivity, complexity, and interminability.We believe that the continuing generation of evermoreimportant human-competitive results relies on progressin three areas of research: multiobjectiveoptimisation, parallel computing, and the developmentand perfection of competent genetic and evolutionarysearch methods. Addressing the characteristics ofhuman-competitive problems is one way to expand thetheoretical underpinnings of the field of genetic andevolutionary computation.