You are starting to push on entropy; as the material changes form, the mass is unevenly disturbed and less of certain compoudns remain over others. Large amounts would infact give some stability, but mostly it has to do with how this complexity is implimented. If you wanted an organism that used sulfur to breathe, it could be developed for sulfuric conditions; to make it breathe air or some easy compound you doul make it defaulted. I don't think we need the chemistry to be simulated so much, as the result.
I dunno, I rather think that it's the simulation of the actual chemistry that is interesting instead of the result. With an abstract artificial chemistry that was as open ended as real chemistry is, you might see some things evolve that you did not specifically code for. Like self replicating molecules that coat themselves in a bubble of fatty lipids. If you just code the end result of real chemistry, you have a finite number of reactions that can occur, and you can never an ability develop that you don't specifically code for.
If the Sulfur-Dependent creature went into an environment which produced less sulfur, its species would suffer as a result.
Think of it like this: right now it is almost impossible in the program to develop a simple plant-herbivore-carnivore model. Why would we want to add another feature that just increases the difficulties of different species living in the same sim? Where's the fun of having a bot that doesn't do well in a sim because it depends on chemical X and chemical X isn't in the sim.
Or to look at it from a different angle, chemical dependencies only become an issue when they're preventing a bot from being successful (because there's not enough). Even if you were to try to fix this issue by
rewarding bots that get more of the chemical instead of punishing bots that don't have enough, you still essentially have a system where not having enough of X limits your growth compared with having enough of X. Same principle applies in reverse if X is bad to have.
Three years ago I was strongly convinced that adding an artificial biochemistry that the bots could manipulate would help add complexity to the sim. (I had a fairly clever system in mind, if I do say so myself). However, now I'm not so sure. It isn't really complexity if you just have two species that co-exist because neither is capable of feeding from the other. My thinking now is to expand the concept of body into three distinct substances: chloroplasts which make nrg from sunlight, sugars/fats which just store nrg for later use, and muscle which gives more power to actions; entirely ignore the whole question of the chemistry involved, and arrive at a simplified, abstract system that describes the typical food chain. If bots can specialize in extracting energy from animal muscle or veggy chloroplasts, we might see some actual complexity and food webs evolve.
It's more of a top-down approach than an artificial biochemistry, but I think it woudl help focus evolution towards producing a food web (which is one of my holy grails for the project).
