Copyright © 1998, Institute of Electrical and Electronics Engineers, Inc. All rights reserved. This article was published in IEEE Intelligent Systems magazine

Of mountains and molehills

Steve Grand

I’m writing this while perched on a hillside at 7000 feet on a beautiful crystalline October morning, looking out across a broad valley to the sunlit peaks of the Grand Teton range in the Rocky Mountains. I tell you this fact, not because it has the slightest relevance to the plot, but simply to make you feel jealous.

The real story takes place in a much more mundane location (or perhaps, if the Rockies are what you consider normal, a more fabulous and exotic one) in the eastern part of England, where the mountains are no taller than I am. That place is Cambridge or, to be more precise, Magdalene College, a mediaeval setting once home to Samuel Pepys, the famous diarist and contemporary of Isaac Newton. The occasion was a small but perfectly formed symposium, organised by my company, Cyberlife, and a group known as Biota.org, of which more later. The assembled crew included, amongst other equally smart people, the biologist Richard Dawkins, the "father" of the Artificial Life field Chris Langton, and the novelist Douglas Adams. Complementing this eclectic trio were other writers, founder A-lifers, AI researchers and virtual world experts, to name but a few.

A few issues ago I was bemoaning the fact (as I saw it then and see it now) that A-life was allowing itself to become locked-into a narrow ecological niche, centred largely on simulated evolution. I feared that this marked the beginning of the end for the discipline, or at least the end of the beginning. Somehow, A-life seemed to have lost its youthful exuberance and hubris. This conference (see http://www.cyberbiology.org) was our small contribution to turning back the tide. In contrast to the abstract, mathematical world of GAs and CAs, the theme for the meeting was "how do we go about creating complete, intelligent, sophisticated artificial life forms?" The dual aims were to re-emphasise the importance of the term "organism" in the study of life, and to shift the focus away from pure science and towards technology.

In the event, the meeting turned out to be more about philosophy. I’d suggested differentiating our little movement from the "classical" A-life world by coining a new term, Cyberbiology. It wasn’t my intention to start a whole new field—that would demand a great deal more scientific foundation than I am able to offer. Anyway, I wouldn’t have dared to be so presumptuous with Chris Langton in the audience. However, I did want to suggest a shift of viewpoint, to a more "cybernetic", systems approach to Biology, AI and life in general. To cut a long story very short, I wanted to suggest that it was time for us to throw off the shackles imposed by three hundred years of highly successful Newtonian thought. Instead, we should take a less materialist view of the universe, in which everything is regarded as software, and the world is studied and classified, not in terms of "things", but in terms of the interplay of cause and effect. A view in which atoms, mind and society are seen as essentially the same kind of non-stuff, everything is irreducibly connected to everything else, and it is not the things themselves but the relationships between them that matter (see Intelligent Connections, Nov/Dec 1997). "My name is Steve Grand, and I am a Holistic".

The relevance of this wishy-washy sentimentalism to the matter in hand was largely the assertion that such a cybernetic viewpoint tells us much of what we need to know about the appropriate toolkit and methodology for the construction of intelligent life forms. When the universe is viewed as a set of inter-related phenomena, whose continued existence can be explained in terms of various forms of self-maintenance through feedback, we see a small set of common building blocks turning up time and time again. The way a cloud persists, the way an intelligent organism exploits its ability to predict future events, the way societies form and reform, all involve the same basic operators, albeit arranged in different ways and instantiated in different structures. A synapse, an enzyme reaction, a transistor and an import tax are all examples of the same thing—a modulator. This is just a re-statement of the principles of Cybernetics, as formulated by Weiner. However, one reason for tacking the suffix "-biology" onto the prefix "cyber-" was that I think Biology has failed to take the importance of the cybernetic view into account. A more important one is that Biology contains clear physical instantiations of these basic cybernetic elements (modulators, integrators, etc.), and I am convinced that the best route to the development of genuinely intelligent systems is by understanding the ways in which such elements can be brought together into circuits.

I may go into that in more detail another time, but the interesting and gratifying thing in relation to this symposium was that many of the same ideas turned up in other speakers’ presentations too. All the speakers had been invited because I felt some common bond with them, and I hoped they would have a lot in common with each other. Happily this turned out to be even more the case than I could have wished, and there was an exciting feeling of accord throughout the event. Chris Langton gave the summing-up speech at the end, and it was extremely encouraging to hear him liken the atmosphere to that at the very first A-life conference.

Aside from the cybernetic emphasis, the focus on whole creatures and the willingness of the participants to discuss normally "taboo" subjects such as consciousness, the other unusual feature of the meeting was the preponderance of people interested in virtual worlds. This was partly because our co-organiser, the Biota.org group, is a special interest group of the Contact Consortium, which is heavily involved with avatar-based virtual worlds. I consider this to be a significant fact. The early pioneers of A-life came from the mainstream sciences, which are not renowned for their emphasis on visualisation—plotting curves in MatLab is the closest most scientists ever get to something as disreputable as computer graphics. Only a few of the A-life First Generation used non-trivial graphics to illuminate their work, for example Craig Reynolds, Demetri Terzopoulos and Larry Yaeger, the latter two of whom were at this conference. Without any background in graphical realisation and embodiment in 3D cyberspace, it is perhaps no wonder that most A-life work has been disembodied, abstract and mathematical. At the same time, much computer graphics work is image-conscious in the extreme, and therefore fails to recognise that just because something looks realistic doesn’t mean it will behave so. The graphics community have tended to work backwards from the visual appearance to the behaviour, which is thus sham and superficial. On the other hand, the A-life community, who understand the concept of emergent behaviour very well, fail to carry the thing through and render that behaviour visible and embodied.

This conference (and to a lesser extent its predecessor, held in Banff last year) marked the confluence of these two modes of thought. If they fuse successfully, we can look forward, not only to a whole-organism, structural, embodied and situated approach to AI, but also to a range of new application areas. Shared virtual worlds have not yet caught on. They tend to comprise a bunch of assorted humans (in the form of their virtual personae or avatars), standing round and looking embarrassed because they can’t think of anything to say to each other. This awkwardness stems from the fact that most virtual worlds are awfully sparse and sterile places, with nothing to do and little to see. There are several reasons for this sterility, but a major one is the lack of any artificial life forms to populate the world. Human beings quite reasonably only ever want to play the part of hero in their escapist virtual lives. Yet virtual worlds, just like the real one, cannot thrive in the absence of bit-part players such as waiters, guards and AI researchers. It is not unreasonable to predict that the future of the Internet lies in the generation of continuous, Euclidean cyberspace. Most of the human brain is designed for handling 3D concepts, and current "informational space" paradigms like the Web fail to tap into all that power. If the successor to the Web is virtual reality, then someone is going to have to build all the synthetic creatures and people that will live in it. Perhaps this is where the first real applications of Artificial Life will occur.

In any case, my friends and I have done our best to rekindle the fading flame of Artificial Life, drawing the threads back together towards whole living things and giving these hope of somewhere interesting to live. Whether this molehill will ever grow into a mountain I don’t know. It depends on whether we can sustain the commitment of the group, ground the rhetoric in theory and make commercial progress in both artificial creatures and virtual worlds. But anyway, those three days in Cambridge were stimulating, uplifting and fun, whatever the outcome.

And so today I’m in Wyoming, at a think tank conference that’s looking at the question of "now we have all this genomic data, what on earth do we do with it?" Again it’s a highly interdisciplinary conference, albeit heavily focused on Molecular Biology. Again I am hearing murmurs of agreement whenever I talk about holism and "learning by putting things together", as opposed to "learning by taking things apart". I’ve even heard things that sound heretically like cybernetic ideas. Perhaps changes are in the wind. Perhaps three hundred years of classical Physics is about to give way to a new Millennium in which Biology is queen; perhaps even Cyberbiology will have its place. Something to muse on anyway. However, you’ll have to excuse me now, I’m afraid—I’m supposed to be at a presentation. On the other hand I might just sit here a bit longer and stare some more at those mountains.

Copyright © 2004 Cyberlife Research Ltd.
Last modified: 06/04/04