Strong notions of artificial life, as discussed in [Sober, 1992] for example, claim that computers and computer programs may not merely simulate, but actually instantiate living systems. When considering the small, closed-world models typical of most of our work in artificial life, the strong claim is often dismissed with variants of the `confusion-of-levels' objection, pointing out that `a model of a hurricane won't get you wet' and so forth. Supporters of the strong claim might argue that life and hurricanes are qualitatively different, or that a hurricane model will drench a model of you, or that some simulations might be different from but every bit as useful as ``the real thing'' [Dennett, 1978], and the discussions continue.
Here I want only to recognize that such unsettled issues exist, then set them aside and veer erratically towards the real and the concrete, accepting uncertain philosophical footings below. I suspect that if you asked the millions of computer-using people today to give an example of `artificial life', a common answer would be `computer viruses'. Whatever the ontological status of a simulation of a forest fire, a computer virus is as real as the computer programs and data files it infects. The Melissa virus invaded over 80,000 reported computers in under two weeks in 1999 [Vatis, 1999]. Even with no `viral payload' whatsoever, Melissa clogged networks and incapacitated servers around the world; it had direct effects on the real people in the real world, it is not `merely' a model. Much as we may want to distance ourselves from the ethical and moral questions, do we really want to argue that the Melissa virus is not artificial life?
We might dismiss Melissa on the grounds that viruses shouldn't be considered alive anyway. The natural world supports some reasonably clear distinctions between hardy `living' systems that flourish in wide-ranging environments, versus mere `parasitic replicators' that require a `virus-friendly' environment, one willing not only to copy nearly any information at hand but then also to interpret the copied program regardless of what it does [Dawkins, 1991]. But in that sense there is a strong kinship between living cells and manufactured computers in that both cell and computer interiors present tremendously virus-friendly environments. Moreover, the actual physicality of a computer itself may support richer notions of life, compared to the apparent insubstantiality of a computer program and the resulting sometimes anemic quality of purely software candidates for artificial life.
Though we may prefer to work entirely with small, manageable programs--executable models that are `close to theory' in some sense--do we really want to argue that the rapidly expanding world of internetworked computer systems is not artificial life? Here, my aim is to run with the naive view to see what insights it may afford, beginning with the position that the connections between computation and life are genuine and the ramifications of that linkage are manifest not merely in small-scale closed-world computer models of natural world phenomena, but in the past, present, and likely future of manufactured computing devices on Earth.
There have been approaches to `real world' computing issues from explicitly alife perspectives--the enterprise I am here calling `real artificial life'--for example [Cohen, 1987,Spafford, 1992,Kephart, 1994,Ackley, 1996]. Either from the outset or over the course of their development, all of those efforts had a pronounced emphasis on computer security. Indeed, one of the largest areas of impact of the artificial life mindset upon current and emerging computing practices is in the area of `computer immune systems' for improving security and robustness, e.g. [Forrest et al., 1996,Kephart et al., 1995,Forrest et al., 1998]. Later we suggest why this has been so.
In this exploration we seek understanding of what is happening around us as the internet grows. We seek leverage from the idea that the stunning explosion of computing power on earth over the last few decades is not unprecedented, but has antecedents in the long development of life of earth. We seek better approaches to the design of manufactured computing, possibilities of a different relationship between us and our computers, and between computers and each other.
This endeavor turns the original charter of artificial life almost precisely on its head. Rather than seeking to understand natural life-as-it-is through the computational lens of artificial life-as-it-could-be [Langton, 1989], we seek to understand artificial computation-as-it-could-be through the living lens of natural computation-as-it-is. The endeavor can fail; there is no a priori assurance the connections between life and computation are bidirectional, or that any identified points of contact will be substantial and specific enough to be usefully predictive.
Still, given the tremendous current and future impacts of the computer and communications hardware and software that we choose to design, and the current paucity of a systematic basis underlying computational design for robust security and privacy, and the relentlessly myopic market-driven approach that most often dominates deployment decisions, it seems worth some struggle to uncover new perspectives.