abbott@aerospace.aero.org (Russell J. Abbott) (09/26/90)
I would appreciate references to the state of the art of oracle machines. (I've hardly even thought about their existence for 25 years.) If you are interested, the reason for my interest has to do with the field of "Artificial Life" (AL). (I apologize for the name, but it seems to have caught on.) One aspect of AL is the interaction among entities, their environment, and the information that organizes the ways these entities interact with their environment. A typical example of such information is the genetic code. An AL issue is how such information evolves. A simple model of all this is: entities are computers, the information that directs them consists of programs, and the environment is whatever the computers interact with. A problem is: what are the primitive interaction operations between the computers and the environment? Consider the emergence/evolution of (programs for) sight. What sort of interface between entities and their environments allows for "learning" how to see? What primitives must exist? It seems to me that primitives are the wrong notion. Instead of primitives, I would prefer to suppose the existence of oracles in the environment. Entities stumble upon and learn to use these oracles. (The "learning" occurs one way or another, e.g., thought evolution. Genetic algorithms are typically frequently as a model.) In the case of sight, the oracle is something that converts light to some signal that the entity can deal with. An even more clear cut example is that of some migratory birds that navigate through the use of magnetic materials that they incorporate into their bodies. Sensitivity to magnetic forces need not be a primitive operation of all entities. All one needs is to find a magnetic force oracle and learn to use it. Sight can be modeled similarly as learning to use a light sensitive oracle. It doesn't matter whether the light sensitive material is biological or not. So I'm interest in finding out whether "oracle theory" is at all useful for such a model. Thanks for your help (and any comments). -- Russ Abbott@itro3.aero.org