[JoshCole]

It seems to me that if an ant is ripped apart limb from limb after being found to be a part of a different ant colony this is not the perfect feed for an argument that ant colonies ought to have pheromones which never evaporate. The actual perfect argument would be proving that the word is stationary such that all past pheromone trails are still highly informative of the true utilities. Worse is that I find the entire structure to be suspect - so what if an at lays a new trail - this is not less collectivist for all that an individual ant laid it. The species was social and the nature of the collectivism is largely a product of the individual pheromone trails being interacted with locally.

I like this model because ants exhibit much of the mathematical structure of a reinforcement learning solution to the explore exploit dynamic in a non-stationary environment, but are simple enough that I can play the rollout in my head in real space and see the consequences.

[pitched]

> “If You Want To Run Fast, Run Alone. If You Want To Run Far, Run Together.”

With all dichotomies like this, we’re never actually aiming for the extremes of far or fast but somewhere in between. There must be some ants finding new trails and there must be most ants travelling them. A hybrid approach is always the goal. Arguments for purity are fantasy, but can be fun.