Sections  Introduction  Emergence  Complexity  Chaos Thoery  Cellular Automata  Life Forms  Neural Networks   Evolution  L systesm  Uses people.htmlBooks Links about this site Site Index  Front Page
Stewart Dean's Guide to Artificial Life

Life Forms

Currently artificial life study is much like entomology. Life forms simulated are simple and small with short life spans and simple world views. In Tom Ray's Tierra these creatures were very small, only a few instructions long. This did not stop them mimicking many of the behavioral patterns found in nature. These creatures could be seen as Kinetic cellular automata with location as one of their status values. There is a shifting scale of complexity from John Conway's life up to and beyond Karl Sim's evolving creatures or the polygons of Polyworld created by Larry Yeager.

A good example of seeing natural behaviour on the computer screen is Craig Reynolds' Boids. In 1986 Craig Reynolds wrote a program that he hoped would allow him to create a flock of creatures. These creatures existed in a 3D environment and where all identical, he believed that by using a simple set of rules he could make the creature flock like birds or fish. The three simple rules were:

  • Steer to avoid getting to close to neighbors.
  • Steer to keep on the average heading of the flock.
  • Steer to stay near the average position of the neighbour.
Rules about goal seeking and obstacle avoidance where also added and creatures now could navigate, as a flock, across and arena of columns. The creatures flowed around the columns and kept in a group, speeding up and slowing down to keep pace.

The program became known as Boids and created a lot of interest. Later on Craigs's rules where used to make a computer animation short called Breaking the Ice starring a bird and a fish called Stanley and Stella. This film was shown in 1987 at SIGGRAPH. The theory of boids has since been used in many other films including the stampede scene in the Lion King and for the Penguins in Batman Returns. More practical uses include the simulation of people as they enter a football ground allow for the correct arrangement of barriers.

As well as these creatures with preordained behaviour there are other creatures with behaviour which is of their own making. An environment is set up and functions are afforded to the growing creature. Then, via evolution, the creatures work out how to interact with their environment. Goals are often set as a motivation for the creatures to evolve as in Karl Sim's battling block creatures. Or the motivation for evolution may just be if you can't work out how to eat and breed you die as in the program Neoterics on the Macintosh.

Neoterics is a program created by Kevin Coble which is a simple approach to aLife. Bugs (blue triangles) are placed in a square arena of plants (varying intensity of green squares). This is a two dimensional arena with wrap around edges. The 'bugs' in the arena have sensors, a neural net and a library of possible actions. The states of each creature include energy level, which direction they are facing and physical location. About one hundred creatures are placed in the arena to start off with enough food to keep them more than happy. The bugs lose energy all the time. Different actions lead to different amount of energy loss. To keep going a bug has to keep eating his greens. Any energy spent in the arena is reconstituted as more greens randomly.

The usual history of a Neorotic system is first those creatures who have no clue quickly die out. Those unable to move quietly expire whilst those who can move but have not worked out how to eat are next. As the creatures get older so their energy demands grow till it becomes next to impossible to eat enough to move. This is where reproduction comes in. One of the banks of commands open to a creature is to reproduce. The options open to the creature are either reproduce asexually or find a mate. The first is, at first most common. Reproduction is highly demanding so often you see creatures give birth then die, birth leaves the parent with only one hundred units left. Once a creature reaches zero energy units it dies. Eventually sexual reproduction happens which is less demanding on the individual in terms of energy and, over time, gives certain creatures the advantage. This is a good program to leave running in the background whilst doing other tasks and visiting now and again to see how your creatures are doing. Clicking on a bug (if you can catch them, fun in its self) gives you the readout of a bug. Its possible to follow the generations of a creature and see how it evolves.

One of the originators in creature filled environments was Larry Yeager who created Polyworld, one of the stars of Stephen Levy's seminal Artificial Life. In this world creatures skimmed the surface of a world in search of food and mates. Creatures were born, died and where consumed either before or after death. As with many creatures in domains like these they have their own DNA which is mutated or altered through asexual reproduction of through the use of cross over via mating. These creatures showed all the signs of life yet Yeager was cautious. "So far", he has said, "What Polyworld has shown is that successful organisms in a biologically motivated and only somewhat complex environment have evolved adaptive strategies for living in this environment." When asked to describe these creatures he saw them as artificial life.

Another example I will give, revolves around multi-limbed creatures that battle for supremacy of a cube. Given the laws of physics, branching bodies and a set of sensors to interface them with the environment, creatures made out of blocks propel themselves around by swimming, jumping, rolling or walking. Their brains were neural nets created in much the same connection and branching method as their bodies. These creatures are the children of Karl Sim's work. They lived inside a Connection Machine, a computer which used parallel processors.

Using an arena with a cube in its center creatures are pitted against each other to gain possession of the cube. This is the carrot in the system, the obvious motivation. The more contact is kept with the cube the better a creature's score, the better its fitness score. Competitions were held with different species and in different combinations (all against one, all against all, tournament). Those who did the best were allowed to breed within their species. The results were creatures that, when seen in motion, move as if they are alive. Despite being made of intersecting blocks you know you have seen this type of motion before in nature. They are blocks possessed by some strange magic spell.

All these creatures are human initiated but soon find the same maths that nature uses to achieve their goals. In the words of Karl Sims

"[The creatures] suggest that it might be easier to evolve virtual entities exhibiting intelligent behaviour than it would be for humans to design and build them."

Lastly no artificial life site would not be complete without a mention of the 'game' Creature created by Cambridge UK based company Cyberlife. Creatures (now in it's second release - creatures 2) uses some ground breaking artificial life work to create virtual pets that are taught and evolve. Creatures has made artificial life a popular topic but it is a entertainment product first so the priority is on entertainment and involvement. For this reason Creatures does not provide an environment that is self organising - without constant the creatures you look after - called Norns - will dye very quickly. Also the number of Norns available in the world do not allow of complex social behaviours to be created and the small gene pool makes evolution a very risky business - the wrong mutation in one or more Norns can result in the end of Nornkind. In effect to keep the world going requires a lot of top down involvement and the world is too chaotic be conducive to a self organising system. For this reason it's not really ideal as an artificial life learning zone but is fun and it is well worth the frustration of trying to teach you Norns to eat, not to jump over cliffs and work the lifts.



Connection Machines

Connection machines were created by a company called "Thinking Machines". This company was set up by Danny Hillis one of the many of the people involved with aLife to have come out of MIT (Massachusetts Institute of Technology). His driving passion was the quest for artificial intelligence. He saw on of the main obstacles in true artificial intelligence the structure of conventional machines, that is computers based on Von Nuemann's original designs. All information is passed through a central processing unit that creates a bottle neck. He claimed because the more you tried to get through the bottle neck the more things got jammed up. In his own words:

"The more knowledge you gave them, the slower computers got. .... So we were in this paradox that if you tried to make computers smart, they got stupider."

Out of frustration came one of the most powerful artificial computers ever built which cost small fortunes but run thousands of processors next to each other. Since then more powerful machines have been built using the same principles, each using processors found in desktop computers. Unfortunately these machines are no longer built for life but for death. They are used to simulate the detonation of nuclear weapons.

Please send any corrections, comments or additions to: alife(at)stewdean(dot)com
 

 back a bit  next bit