7: Future Work
Due to the fact that the genotype data structures for all the species are identical, a file saved from one species' genotypes can actually be loaded into another species' genotypes in memory. This is an intriguing feature which was not originally intended, but discovered once when I accidentally loaded a file into the wrong species. The results of this are almost always uninteresting, because there is very little mapping from one species genome to another. But a logical mapping is conceivable. This leads one to imagine the potential for a system which allows this to be done in an organized fashion. It would allow a form of inter-species breeding - impossible in the biosphere but very common in the ideosphere.
An idea suggested by Resnick (94) comes to mind, which he calls "clip behaviors." Clip behaviors are analogous to clip art (commercially available images, icons, and symbols used in many desktop publishing systems for inclusion in designs). If the genotype/phenotype representations for behavior objects in the Character Evolution Tool were designed with mappings between them, such that species could be treated as classes and sub-classes, it would be feasible to load the genes for behaviors evolved from one species into another species' phenotypes, with predictable results. A feature of this sort would be a very powerful catalyst for the transferal of ideas from one design domain to another.
Family Trees
One can save and load genotype files for populations of evolved behavior objects. But as yet a scheme has not been developed in this system for relating the files according to familial context, for handling geneology. There is also no scheme for visualizing genealogical development. Further development in this regard would be a great contribution to the system, and it would empower the user with a rich environment with which to develop and manage developing, customized taxonomies of behavior objects.
Flesh and Bones
This thesis deals primarily with the creation of motion behavior. Line drawings serve well for this and so, in designing my articulated figures, I have not put flesh on the skeletons. But a complete animation studio would need a means for doing this as a next step in production. Once the motion behaviors had been developed for a character, an animator would have to then dress the character in flesh, clothes, what have you. The Character Evolution Tool does not address this need. But it can be seen as one component in an ensemble of animator's tools in which another component would be available for "dressing the motion".
Interacting Characters
A useful extension to this tool would be an environment in which multiple characters of different species could interact with each other. Most animated narratives do not feature one lone character in a simple environment, but have two or more characters which interact with each other. As yet there is no mechanism for evolving behaviors for interacting with other characters (of the same species or of different species). Adding this level to the Character Evolution Tool would bring an element of ecosystem into play - what would become evolvable would not only be individual behaviors, but interactions among individual behaviors.
Imagine a future extension of the interface to the Character Evolution Tool in which each animation window included more than one species of characters. One can see that evolution would take on a new level of complexity - it would become co-evolution. One reason it would be more complex is that the concept of fitness may no longer apply to one character alone, but could apply to another character in the same scene, or even to the mutual interactions between two or more characters. To handle the multiplicity of fitness targets, the user would need to have the freedom to select any one of the characters in the scene, or to select the whole interactive ecosystem, to reward fitness. In working on this level, it is conceivable that one can be evolving, not just characters, but interactions among characters.
A co-evolution system would probably require an entirely different interface than the one represented in the Character Evolution Tool. No more being trapped in their own little test tubes - each character would have to be free to move about and interact with other characters within a large environment, in order to be evaluated. This may pose problems, since interactive evolution typically involves evaluation of a collection of individuals in a population (presented in a grid) to guide evolution. In short, evolving an ecosystem would require a new interface design strategy, perhaps displaying something more akin to what a scene in the final animation would look like - the addition being the inclusion of the viewer (as user) interacting with an evolving cast of characters.
Glossary
