Reviews

Digital Digital Burgess Conference Follow-up:
General Conference Report
Conference was held August 29-September 1 1997, Banff Alberta, Canada.

Also see Links to Digital Burgess Speaker Sites and Other Information

Digital Burgess was a cross-discipline meeting of computer scientists, paleontologists, and artists to look at artificial life’s colonization of virtual space. Held in the rarefied heights of the Rocky mountains - at the Banff Center for the Arts - the conference aimed to look at parallels between the Cambrian explosion of new lifeforms, 550 million years ago, and the current development of digital lifeforms. Back in the Pre-Cambrian age some mysterious effect caused the single cell soup of life in the oceans to diversify into an extraordinary variety of multi-cellular organisms:10,000 new species, the basis for all life alive today, and for many more creatures now extinct. The members of the Biota organization - organizers of the conference - saw similarities between current experiments into Artificial-life, digital genetic-style manipulations, and the creation of 3D lifeforms on the Internet, and those experiments of evolution that appeared in such abundance in the Cambrian era. The conference brought together experts on the interpretation of fossils, designers of simulated organisms and eco-systems, and artists experimenting with the evolution of genotypes (a DNA or other set of instructions) expressed through the phenotype (physical form) of 2D and 3D images, and 3D animation.

You may wonder why meet in a remote place like Banff in Alberta Canada? The answer lies in its proximity to the world’s greatest site for fossils of soft-bodied Cambrian organisms – the Walcott quarry of the Burgess Shale. This is the source of discoveries referred to as ‘weird wonders’ because they don’t fit within the classification systems for organisms alive today. As preparation for the more sessile stages of the conference the delegates undertook a 16 mile hike to the quarry. This pilgrimage gave us all a feeling for the meaning of survival of the fittest.

The conference itself was bracketed by sessions open to the public. At the opening session Desmond Collins, Chief Scientist for the Burgess Shale at the Royal Ontario Museum, gave everyone some background on the significance of the Burgess Shale and the type of discoveries made there. He has spent 15 seasons camped out at the quarry, pulling fossils from the shale and studying them. New finds keep occurring, the latest ones being two creatures with the nicknames Mickey Mouse and Squid on a stick - official classification waiting on finding if they fit anywhere among the known species on Earth. Aside from their diversity and evolutionary history the Burgess creatures are significant because they show us that the dinosaurs weren’t the only mass extinction to happen on Earth. What happened to destroy so many species is of concern to those of us still living on the planet - will it happen again?

Although fossils are riveting to those who understand their significance, there is still the issue of how to involve the public in the excitement of fossils. Paul Johnston a curator from the nearby Royal Tyrrell Museum discussed his plans for making the Burgess organisms easier to see and understand. He proposes making man-sized models of them, adding color, in line with that seen in tropical reef sea-creatures, then setting them in a recreation of the Cambrian ocean world. This attempt to communicate clearly was echoed by the opening to the main conference sessions. Deliberately kept small, only fifty places were available, to ensure uninhibited communication - the conference opened with a plea for jargon free clarity from paleontologist Dr. William Riedel. Urging his own group, the Paleontologists to remember they were not just talking to colleagues he also appealed to the computer scientists to avoid assuming everyone had a comprehensive knowledge of their field. Bringing together the results of work in paleontology and computer science is one of the benefits of this type of conference.

Then it was into Paleontology. Stephan Bengtson tried to teach us how to look into the past, reminding us we only have the tiny keyhole provided by fossils. One of the notable features of the past is mass extinctions. Ninety nine percent of organisms from the far past are extinct, so the forms we see today are but a fraction of the original diversity of species. When we look at the far past it can be hard to interpret what we find if we just expect it to be the same as things around today. Once we reset our expectations we find that the closer we get to the Cambrian era the more species exist. If this diversity comes from the sheer space on an empty planet then we can draw parallels with the unoccupied Internet and expect a similar phenomenal growth of new types of organisms there once the beginnings of life-like organization take hold. The Cambrian seas had abundant single-celled organisms to use as food, does the Internet have enough spare CPU cycles and memory to provide an equivalent digital food supply and space to exist?

Roy Plotnick urged us to think about what would happen if we reran evolutionary history. Would it be like a tape where you always get the same ending, a play with a fixed script or, as he argues, is it more like improvisational theater? One of his areas of study is what drives extinctions? Internal or external events? He looked at the question of should we be prepared for the Internet equivalent of a meteor impact affecting the evolution of lifeforms, or is there some ecological pattern inherent in life that makes extinctions inevitable? Using Cellular Automata he models the effect of forest fires on an ecology. He found that if he prevented small fires from happening then a major fire would inevitably occur. He extrapolates from this that the long term distribution of a species is robust - life survives extinctions over and over again. So, if we create life on the Internet we can expect to find the same effects there. We should let Net organisms self-regulate their evolution and not try to rigidly control their behavior with regulations and safety measures. Disaster is part of life and if we have truly created life on the Internet then it should be able to survive, what form survives is of course unknown. Just as on Earth we don’t know if life will continue as humans or insects.

Bruce Runnegar showed us what digital techniques can do for Paleontologists. He told us "Paleontologists study creatures of the imagination. The digital tools are things Paleontologists have needed for generations." The tools he’s referring to are 3D modelers to recreate the form of long dead organisms. Fitting creatures into their place in evolutionary history is easier when you can see 3D form and derive function from it. He showed us a series of fossils now understood better through virtual modeling. From the crystal sphere of Braarudosphera to the bundle of straws that is Ernietta, expanding the flattened fossil into three dimensions then animating it in a virtual world helps graft orphans (plesions - extinct side branches) back onto the tree of life. Dr. Runnegar says that when looking for a biological relationship between a current and ancient organism building a 3D model, to show identity of structure and similarities of developmental sequence, is tremendously useful.

Ricardo Colasanti, an ecologist now working for Cyberlife, showed us his approach to finding the rules that life uses to organize behavior. Using a simple 2D display and a modular system for modeling the structure and function of plants he set up simple rules, ran his model, and checked the resultant growth patterns against his observations of nature. A perfect match showed he’d found another of Nature’s secrets. Expect to see the results of his work appear in the Gaia tool for modeling life.

Richard Gordon, a theoretical biologist, presented details of his CyberWorm project. This ambitious four dimensional modeling scheme aims to completely simulate a Nematode worm, from its genes on up to its physical form and behavior. If he captures every aspect of the worm in software will he have created life? We may have to wait somewhere between ten and thirty years to find out - this is a long term project. This apparently pure research project could have useful spin-offs in the understanding of complex systems. This is one of the reasons computer scientists have become interested in the work of evolutionary biologists and paleontologists.

Two representatives from British Telecom (BT) came to present their spin on why evolution of software is important. Chris Winter took us on a high speed tour of the size of the problem: BT needs ten percent more software a year but can’t get the programmers to produce it and is running into a complexity wall. BT has no idea how to engineer the networking software its going to need in the twenty first century - so its going to try and evolve it. They’ve already evaluated some evolutionary techniques, GA, GP and Tierra. GA gave poor diversity of results, its creativity at producing novel algorithms was awful, and it only worked for small scale solutions. GP produced better diversity in its mutated programs but still wasn’t creative enough to be useful and was small scale. Tierra was excellent at creating diversity and creating small useful programs but was again a small scale solution. None of these approaches could produce the ten million lines of code that BT want.

Yet looking at the gap between human DNA (100 MB) and the complete UNIX OS (500 MB) and what they can each achieve BT are certain that evolvable computers must be the answer. They hire people like Paul Marrow, an evolutionary biologist, to study the conditions necessary for large scale evolution. He told us about his work to find a balance between stability and instability, the point where a system can be destabilized so it can mutate. Then he gave us a list of properties of biological systems as an aid to appreciate the complexity of the problem. Do you know a piece of software with these properties?

  • A self-replicating genetic system where a complex genotype is mapped onto an even more complex phenotype.
  • A development process that can be driven by co-evolution or competition with other species.
  • A modular nature which is robust in the face of mutation.
  • A tendency to use networks internally and externally.
  • A response to the environment based on adaptation to challenges.
  • A rich suite of behaviors including learning, communication, conflict, cooperation and altruism.

Chris Winter then bounced back to talk about how BT is applying these ideas. They’re looking for new hardware for computation, and its not digital but analog approaches that they’re researching. A couple of the many technologies they’re evaluating are EPACs (Electronic Programmable Analog Circuits) and FPGAs (Field Programmable Gate Arrays). As a small example of the output from the research Chris showed us a shoal of VRML fish that analyze and display the relative affinities between documents in a database. Using self-organizing biological principles the fish exhibit behavior that gives immediate visual cues to the relationship between the data they represent. Imagine spotting new trend by the size of the purple fish shoal on your screensaver increasing! This is not copying biology but learning from it.

Tom Ray warned us that evolution in a digital medium is not necessarily a model of evolution on Earth. Life on Earth operates with carbon chemistry, this is not the only way. There could be other trees of life, on other planets and in digital space. Tom knows evolution can work in the medium of digital computation. He works with digital organisms and studies the processes of their evolution. For him "cyberspace is a real space with real processes". His digital ecosystem is called Tierra. Within Tierra Tom tries to simulate the effects of the Cambrian explosion by using evolutionary techniques to generate complexity. He is hoping that his system will generate complex organisms akin to those seen in the Burgess shale. The latest research with Tierra is looking at the possibilities with networked ecosystems. The organisms can move from one server to another feeding on CPU cycles and living in RAM. Where they choose to live depends on the relative abundance of these resources. Tierra has reproduction and evolution.

Polyworld has reproduction, evolution and behavior. Its creator, Larry Yaeger, is concerned with quantifying life: how do you tell if its alive? He has a list of criteria: Life is a pattern in space-time rather than an object, Life is capable of self-reproduction, Life has a metabolism, has functional interaction with its environment, the list goes on and on. He is working on the evolution of neural systems within an ecology starting from his own electronic primordial soup. His organisms find themselves in a space where there is food and lots of other organisms. He gives them some basic behaviors: eating, moving about, mating, fighting then leaves them to evolve their own neural architecture. New species with distinct behaviors emerge from this forcing ground: Joggers that just want to run around and mate, Dervishes that spin frantically to encounter the maximum amount of food and mates, Indolent cannibals that lie around and eat other organisms like themselves, and many more. Larry is hoping to discover the level of system complexity necessary to evolve intelligence. He muses on questions such as can there be intelligence without life? And can there be an information based measure of life? For now his definition is Life is a word to describe systems with a high level of behavioral complexity.

Demetri Terzopoulos, a professor at Toronto University and Intel’s graphic research group leader, showed his range of artificial animals. Working to achieve realism in appearance and behavior for his creatures and their environment Demetri is going for the comprehensive zoological model covering physics, locomotion, perception, behavior and learning. The ultimate aim is to make these creatures into lifelike autonomous agents. The fish demo was pretty convincing: starting out as clumsy movers the fish, be they sharks or rays, learnt to swim all by themselves. This meant a lot of computation, matching the driving forces of the elastic elements in their muscle fibers against the displacement of virtual fluid. Behavior is built up from low level routines which can be put together to result in typical fish behavior like schooling and mating. These virtual animals then become lab rats to study problems in learning, cognition, and perception. Paleologists in the audience were excited at the idea of being able to examine how the fossils they find moved and behaved when they were alive. But as Demetri’s research moves into areas of cognition, causality, and planning the line between his simulations of development and evolution and the real thing will become harder to draw. It could give a whole new meaning to ‘Intel inside’.

Steve Grand, famous for his Creatures software came to help us find the true meaning of artificial life. He rejects the lists of criteria for life. Instead, taking us through a series of entertaining thought experiments, he came up with a way to define the possibility of life. A first level simulation won’t do to create life. A first order simulation is not an instance of the thing it simulates. A model of quantum mechanics isn’t quantum mechanics but atoms and molecules built using the model can be real in its terms. You have to build one simulation upon another and at some nth layer you may create life. This also applies to artificial intelligence and consciousness. This theory so impressed the other speakers that everyone that spoke after Steve referred to his theory as a solid way to understand what is really being achieved in a simulation of life. Steve’s Norn creatures are third order simulations, being based on his artificial genetics based on his artificial chemistry - but whether they are alive or not Steve can’t say.

Bruce Damer, the main organizer of the conference, talked about the philosophical implications of a Cambrian Explosion of life in digital space. He took us on a tour of existing work on the Internet such as the Biota Nerve Garden installation of a VRML island covered with L-system plants. He also looked at the idea of Alphaworld and its cohorts as an emergent phenomena in Cyberspace. He believes that in order to get a Cambrian explosion on the Internet you just leave people in a rich environment. It takes a 3D environment to be rich enough to provoke the development of new organisms. This is more of a creative ferment than an evolutionary one. However Bruce says "If digital media are complex enough to harbor life, life will push its way in there". And then there’s his long term belief that the ultimate aim of life is to leave the planet’s gravity well: "to get out of a gravity well life should be a digital/analog signal". So he leaves us with the question: "Can we provide the matrix for life to continue evolving? - That’s worthwhile."

Karl Sims presented his theories and demonstrations of ‘survival of the prettiest’ - using aesthetic selections to guide evolution in 2D, 3D or even of motion algorithms. Although his demonstrations once needed the computational power of a CM2 supercomputer they now run on SGI workstations and in about three months, the 3D creatures should be available on the Web. A VRML / Java powered version will use an adaptation of Biota’s nerve garden token ecology to create a stimulus response coupling to replace the physics calculations that are currently employed as a framework for organisms’ behavior. You can download videos of the original creations from the Biota Web site now and marvel at the ingenious algorithms that evolved to move boxy creations along -- it’s like watching kitchen cabinets dance. The best mover wins the right to reproduce.

There was lots more art. Darrel Anderson showed Grobot, his tool for kids to make their own explorations in evolutionary art. Steven Rooke brought along a video of his work based on Karl Sims ideas. Przemyslaw Prusinkiewicz and Christian Jacob showed us their amazing work with L-systems to grow complex and realistic plants from algorithms. To follow up on all this work, which really needs a lot of pictures to explain it, use the links in the sidebar.

At the end of the final open session Tom Ray gave us his vision of the future - a digital nature with a life comparable in richness to that of organic life. He wants us to be ready for this explosion of digital life and proposes that ten percent of Internet Cyberspace be set aside as a nature reserve for the products of digital evolution. He says "We could let the spontaneous processes of evolution produce digital organisms then look for applications for them." Like the rainforest this reservoir of organisms could be both a beautiful place to visit and a storehouse of software treasures to develop further. Tom says: "The idea of evolving super complex software is appealing – we are at the limits of complexity of what we can build for ourselves." Just watch out for the cybergoats eating your software flowers. We’re at the beginning of the Earth’s virtual Cambrian explosion of digital biota -- who knows what will develop next?

Links to Digital Burgess Speaker Sites and Other Information

See the Conference Photo Albums




Click here to see images of the conference


See the Burgess Shale Panoramas created by Todd Goldenbaum
Click on these images to get a larger still GIF version



Click here to see the Burgess Shale Panorama
(note you must download the RealSpace Viewer Plug-in first!




Conference Home


Biota.org