The things we could do

[greatblondelf]

We could reproduce the "prisoner's dilemma" algorithm competition, defining the basic "building blocks" of a "playing" algorithm by things like grudge-holding, natural cooperativeness, responsiveness, and memory. We could define a set of building blocks by a long string of alleles representing that algorithm. Then we could randomly pit them against one another, and define "fitness" as the actual point-win-rate of an algorithm over, say, 6 random matches. Then we could select for, and recombine, the genetic patterns to look for emergent properties of the new system, and see which algorithms tend to win. Then we could.....well, write a paper. The really interesting bit is that, since competition with the population is the metric of fitness, that metric time-evolves along with the genotype population.

Anyone wanna do this some weekend? I'm really getting in the mood to do some kind of "one-off" paper like this...For extra credit, my bet is that the population settles down to one of several "attractors" in phase space, which I'll define as the average expression for each gene in the population. I hereby call these "genetic attractors," and their strength and location in the phase space should be dictated by the measurement method for fitness of an arbitrary genotype.

Okay, I'm done now. I need to publish some things on the order of NOW, so...as soon as I'm done with the American Nuclear Society, it's on.