Reviews

Digital Digital Burgess Conference Reviews:
Roy E. Plotnick
Conference was held August 29-September 1 1997, Banff Alberta, Canada
.

The Digital Burgess conference was almost certainly the most unusual meeting ever held concerning a paleontological subject. During it, a small coterie of paleontologists and other natural scientists interacted with a much larger group of computer scientists and digital artists to communicate ideas, explore areas of mutual interest, and lay the groundwork for possible future interactions. As befits an effort of this nature, there were many areas of contention, confusion, and lack of communication. By identifying what these areas are, this will hopefully lead to serious efforts by all involved to continue the dialog and learn to speak a common language. Nevertheless, I believe that nearly all attendees benefitted greatly for the exposure to novel ideas and were left excited and energized by what they were exposed to.

Clear distinctions between the natural scientists and many of the other attendees were in the goals of computer simulations. To quote the ecologist Levins (1966), scientists view models as "a reconstruction of nature for the purpose of study." Because of the inherent complexity of natural systems and processes, scientific modeling involves the reduction of this complexity while retaining essential aspects. It would not be beneficial, even if possible, to build a computer model that exactly reproduces a bacteria unless the process of building the model yields insights into how real bacteria work. Similarly, building a simulated fish that swims, as Demetri Terzopoulas described, would be of limited interest to most zoologists unless the simulation was grounded in real fish anatomy and realistic, detailed fluid mechanics and biomechanics. The potential for doing so is clearly nascent, however, in his models and those of Karl Sims.

In contrast, many of the artists and computer scientists seemed to invert this approach, by using biological evolution as model for how the digital world will change in the future. Conference organizer Bruce Damer, for example, discussed "biological models as appropriate models for Cyberspace." In order to accomplish this modeling, however, evolutionary processes were reduced to very basic neo-Darwinian concepts. In nearly all cases, the processes of mutation, recombination, and selection (both "natural" and "artificial") were the only ones considered. Similarly, simulated "ecologies" were characterized almost totally by processes of competition and predation.

This dichotomy between computer systems as models of biological evolution and biological systems as models for computer system evolution was best seen in the fascinating presentation of Tom Ray, who stated that "just as evolution on other planets is not a model for life on earth, nor is natural evolution in the digital medium." Tom's Tierra world creatures are not bacteria, but are there own type of organism, living in a medium and under rules that only metaphorically resemble that of the "real" world. Nevertheless, both Tierra and Larry Yaeger's Polyworld have the potential for recognizing emergent properties that might be common to biological, digital, and "alien" complex systems.

How can computer scientists and paleontologists (and the broader group of evolutionary biologists) aid each other? I, for one, would love to have some hand-me down computers and some free help in programming! More seriously, as mentioned earlier, the potential to produce realistic, moving models of fossil organisms based on realistic biology and physics is tremendously exciting. As I mentioned in my own presentation, the ability to use powerful computer models would allow evolutionary scientists to play "what if" games with the history of life and help determine what aspects of evolution are contingent and which are predictable. The work with L-systems of Przemyslaw Prusinkiewicz shows a possible next direction for theoretical morphology.

Conversely, computer scientists should benefit from the insights into evolutionary processes that natural scientists have (the interaction between Chris Winter and Paul Morrow was an outstanding example of this). For example, the "environment" in which the computer programs evolved was static, not dynamic. I wonder if Tom Ray's Network Tierra would evolve differently if individual CPU's frequently crashed, disconnected, or otherwise changed. Can we evolve a computer system that has the flexibility to deal with a fluctuating, often unstable environment? This strikes as a desirable goal as computer systems become ever larger and more complex. I would also be interested in seeing how concepts of hierarchy and feedback might be incorporated, as well as the possibility of non-optimum, but satisfactory solutions.

Review by Roy E. Plotnick

Department of Earth and Environmental Sciences
University of Illinois at Chicago
845 W. Taylor St.
Chicago, IL 60607
plotnick@uic.edu
web page: http://www.uic.edu/depts/geos/plotnick.htm





Conference Home


Biota.org