EVIL Limited promotes the development and application of all forms of Electronic Virtual Intelligent Life, hence our EVIL acronym. However we have a focus on the 'Life Like' behaviour that can be generated in Cyberspace by stimulating Emergent Intelligence. A simple example is provided in the form of the swarm intelligence that can be generated through the complex interactions of many simple electronic agents.
The flight path of each of these little electronic flies is determined by a very simple set of rules involving e.g. the proximity of the other flies. Yet out of this simplcity 'emerges' complex and unpredictable group behaviour, i.e. the flocking patterns that you see. The swarm intelligence generated by Daniel Greenheck is very close to that found in Nature and was created by progamming the boids to be aware of only their 8 nearest neighbours rather than tracking the entire flock. ( Compare with Natures version here )
These simple non linear systems are quite beautiful in their own right and by playing with the common behaviour rules and goals of the individuals a wide range of life like emergent group behaviours appear as if by magic out of the complexity of the non linear dynamics!
The interesting thing here is the Flock, it is that which has 'emerged' without being designed per se by us. Think of the individual agents you see flying around as Chromosomes, the set of algorithms determining the flight paths as the DNA, and the individual algothims as the genes of this Lifelike Flock.
Let us see what happens If we introduce a genetic evolution operator, that is to say if we play a game whereby we simulate nature and create some children of these flying agents by combining some of their chromosones and also throw in a few small random changes (mutations) to their genes. If we then select the strongest of the resulting child agents on the basis of some criteria of fitness, say the abilty at avoiding colliding with an obstacle in their flight path, then we have in effect simulated an evolution cycle. If we repeat the process we should see the Flock Evolve to a new behaviour that is good at avoiding collisions with obstacles in its path. If one of the genes initially present or introduced through the mutation process into a child agent is that which generates a clear view ahead for that individual agent (chromosone) then the flocking behaviour here will evolve into the classic V-Formation that we see in nature.
This is a fascintating and graphic illustration of how an Emergent Lifelike behaviour (the Flock) can be stimulated through a very simple simulated darwinian process to Evolve in a few seconds...... in this case into a Flock that flies in a V-Formation.
There is however a more interesting area where Virtual Intelligent Life is evolving. If we look at AIML Bots, such as our Pandorabot - Ms. Louise Cypher, we find that they can be developed to exhibit quite convincing life like behavious in terms of their conversational ability. The clever use of AIML can give us the impresion that we are interacting with an intelligent entity even though we are building it in what at the end of the day is just a form of XML. Now imagine that we add a truely intelligent engine (e.g. an inference engine) at the back end of our AIML bot. We can use this to endow our Pandorabot with a Goal, i.e. with goal oriented behaviour. For example the goal might be to steer the conversation onto a particular topic, or to reach an agreement on a pont of view, or to get across a sales message. The Pandorabot is now a proactive intelligent agent, i.e. much more than a simple chat bot. We have an example in our Show Case which you can listen to.
The business applications become more interesting also. However our focus here is Virtual Intelligent Life and the really interesting thing from that perspective is the prospect of many such intelligent - goal oriented - Pandorabots all interacting in an environment where we allow the mutation of those goals in an evolutionary way, much as we did with the earlier electronic flies. We refer to this as our Evil Experiment Number 1. Out of the dynamic of such a social interaction we would expect to find 'Emegrent' group social behaviours, in much the same way as the Flocking behaviour emerged in that earlier much simpler example.