Digital Digital Burgess Conference Follow-up:
Lethaia Foundation Review
Conference was held August 29-September 1 1997, Banff Alberta, Canada.
Lethaia, Vol. 30, pp. 335-336. Oslo, 1997 12 15
New alliances for palaeontology
What do artificial life, computer art, and palaeontology have in common? Well, a meeting was recently (August 29th to September 1st, 1997) held at the Banff Centre for the Arts in the Canadian Rocky Mountains to find out. Entitled 'Digital Burgess', it was the brainchild of Bruce Damer of the Digitalspace Corporation, and the idea was to bring together computer scientists, artists, palaeontologists and biologists around the theme of the Cambrian Explosion as a source of inspiration for interaction.
Why the Cambrian Explosion? This event has aroused great interest also outside palaeontological circles, largely thanks to Gould's (1989) 'Wonderful Life. The Burgess Shale and the Nature of History'. The Burgess Shale, although now only one in a series of spectacular Cambrian lagerstätten, has taken on an almost mythical significance, comparable to that of archaeology's Rosetta Stone.
In addition to serving as metaphor and proxy for the Cambrian Explosion, the Burgess Shale gave a hands-on experience to the meeting participants - Banff is only an hour's drive from Field, British Columbia. Desmond Collins of the Royal Ontario Museum, long-time investigator of the Burgess Shale, introduced the group to the biota of the Shale, and on Sunday, August 30th, he led a party in the long climb to Walcott's Quarry near Mt. Field. For many, this was the first exposure to fossils in the field.
The following two days saw presentations and discussions regarding possible ways that evolutionary biology and palaeontology can interact with computer-based reasearch into artificial life to achieve better models of evolution. A second pervasive theme was how visualization techniques can be used in research and to present palaeontology to non-palaeontologists.
It was a meeting of disparity. Bill Riedel, who was responsible for making palaeontology more than an icon at 'Digital Burgess', identified a spectrum of approaches - from the fact-fettered to the airily speculative. He somewhat mischievously requested that contributors identify their place on that scale. Some came out on the side of the hardcore reductionists and some rather as unrepentent dreamers, but the question, not unexpectedly, turned out to have many dimensions, and the position of the participating scientists, computer modellers, and artists in the multidimensional spectrum was not always obvious.
Beginning the sessions, the palaeontologists and biologists presented aspects of their fields, particularly with regard to to the interplay between data and theory, and the possible interfaces with the computer sciences. This was followed by presentations from computer scientists on artificial evolution, containing much of direct and potential interest to anybody interested in 'real' evolution. The systems that are being created in the computers may still be regarded as mimicking life rather than as being alive, but the distinction may become more and more difficult to recognize as they begin to exhibit emergent properties independent of instructions from the programmers.
Tom Ray presented his Network Tierra, a digital biotope of interconnected computers in which virtual creatures evolve (using processor time as energy resource) through selection upon random 'mutations'. The biota has started to exhibit complex ecological interactions, host-parasite relationships, arms races, etc.
Karl Sims demonstrated the power of evolutionary algorithms in computation with his image-generating system, in which the viewer can select desirable varieties among a set of related images. The selected image (or pair of images) then becomes the parent of a new set. After only a small number of generations, the images thus evolved can be stunning in their visual qualities and variations. Even more amazing was Sims's demonstration of virtual lego-like objects with basic physical properties and the ability to assume any proportions, to join, to torque at the joints, and to evolve neural cirquitry. These were selected, using optimization algorithms, for their ability to swim, run, jump and follow moving objects. The resulting anatomies are sometimes weird but often remarkably well-adapted to their digital environment, and to the observer they are amazingly life-like (one of the recurring forms was a gracefully undulating 'water snake').
One wondered what such digital evolution could lead to, given more biology-like material to act on. A hint of an answer was given by Demetri Terzopoulos, who modelled fish in ways that were partly conventional (skin upon wireframes), but rather than programming their movements in detail he let their behaviour 'evolve' in the virtual environment. Sharks would attempt to swim and be selected for their efficiency; as a result they would soon evolve effective and amazingly life-like movements, in spite of the fact that they had a stylized anatomy and were 'swimming' in a simple virtual medium lacking the hydrodynamic complexities of real water.
Other contributions concerning evolution in computers included Larry Yaeger's presentation of Polyworld (a system of ecologically interacting simple 'creatures'), Przemyslaw Prusinkiewicz's and Christian Jacob's demonstration of the use of L-systems (Lindenmayer Systems) to simulate plant development and evolution, Chris Winter's and Paul Marrow's account of how British Telecom is attempting to apply biological knowledge to the development of tomorrow's computer applications, and Rajashi Das's discussion of emergent computation in systems made up of simple components limited to local interaction.
The presence of emergent properties (i.e. properties not put into the system from the beginning) in computer systems is obviously of great potential interest to those who model life-like systems. But for the natural scientist who wants to use such models to understand real life, it may be disturbing that computer systems given to evolve by themselves very soon become opaque even to their creator. If we don't understand how the model works, what use is it as a model of reality? Maybe the way to look at it is that even a perfectly understood and working model does not guarantee compliance with reality: the proof is rather in the performance. A computer-based 'biota', even if following largely unknown mechanisms, may be useful in several ways to evolutionary biologists: it lends itself to experimentation and tinkering in a way that the natural biota does not, and it serves as an 'extraterrestrial' biota against which we may test what aspects of evolution of life are dependent on the particular constraints of Earth's environments, DNA, etc. We may develop tools with which to understand the internal workings of digital 'life', much as we use molecular biology today for understanding real life. And if properties of the system are emergent and cannot be explained by lines of code, it is basically no different from the fact that nucleotides (or quarks) cannot explain the existence of cricket players - or reductionists.
Palaeontologists and evolutionary biologists appear to have good reasons for paying attention to what is now stirring within computers, and computer scientists may want to learn more about evolving living and fossil organisms in order to apply this knowledge to their modelling. Even so, one must not underestimate the diffuculties in taking the interaction beyond the initial sense of wonder. But already today, most palaeontologists could profit greatly from computer modellers and artists when it comes to reconstructing and visualizing extinct organisms. The artistic contributions to 'Digital Burgess' (also by people who did not refer to themselves as artists) were impressive, but what captured my imagination most was one of the few non-digital contributions: Joel Hagen's 'extraterrestrial' skeletons, modelled in old-fashioned clay. These 'xenopalaeontological' objects could serve as a fascinating ironic accent in a natural history museum.
Information about the conference, with links to web pages of several contributors, can be found at http://www.biota.org/conf97/index.html.
Gould, S.J. 1989: Wonderful Life. The Burgess Shale and the Nature of History. 347 pp. Norton, New York, N.Y.
Stefan Bengtson [firstname.lastname@example.org], Department of Palaeozoology, Swedish Museum of Natural History, Box 50007, S-104 05 Stockholm, Sweden; 28th November, 1997.
© Lethaia Foundation, 1997