You misunderstand my point... larger ratio of base pairs to behavior means that the behavior is likely more robust to noise, and the original evolution of the behavior would have been smoother (i.e. proceeded in less of a step-wise manner, because each step would be a much smaller fraction of the required change than previously thought).
Traits embedded into a dense vector space encoded as DNA base pairs is a pretty radical deviation from the one-hot vector approach we were taught in school. Wouldn't it be lovely if we found that vector addition worked similarly in this domain?
