Spatial Distribution

By replaying the `movies' recorded during the run (Section 5.1.2) we can see how the spatial distribution of the programs, and of various specific aspects of the programs and the environment, changed throughout the run.

As the run settled down, the spatial distribution of the programs became more regular, with many programs placed such that all of their immediately neighbouring grid positions were empty (compare the positions of programs in Figure 5.25, taken at time slice 5,000, with those in Figure 5.26, taken at 910,000).

   

Figure 5.25: Standard Run: Cell Energy Distribution, Time Slice 5,000. (White squares are empty.)

Figure 5.26: Standard Run: Cell Energy Distribution, Time Slice 910,000. (White squares are empty.)

Figure 5.27: Standard Run: Cell Age Distribution, Time Slice 900,000. (White squares are empty.)

The fact that the programs were evolving to collect more energy from the environment is demonstrated by the much higher average energy levels of programs in Figure 5.26 compared to those in Figure 5.25.

As the population tended to be dominated by a single genotype throughout the run, the movie of spatial distribution of program length is uninteresting. The distribution of ages of programs across the grid was fairly even, as shown in Figure 5.27.

No multicellular (parallel) programs emerged during the run. Also, because the system parameter neighbouring_genomes_readable was set to no, neither did any programs evolve which executed foreign code. After the move instruction included in the original genotype 348AAAA had been lost in genotype 366AADW, only an occasional variant rediscovered it subsequently, but this was never again taken up by most of the population.