592 Mark A. Bedau
Hillis [1992] demonstrated that co-evolution can spur evolutionary progress, and
co-evolutionary arms races might help drive continual evolutionary progression by
continually changing the environment for evolution. But the original and most
modified versions of Tierra involve some form of co-evolution and yet the envi-
ronment eventually becomes essentially stable, so there is probably more to the
story. Further progress on open-ended evolution would be aided by quantitative
comparisons across different artificial and natural evolving systems. Bedau and
Packard and their collaborators have taken a step in that direction by defining and
studying evolutionary activity statistics. Comparing data from different artificial
and natural evolving systems suggests that there are qualitatively different classes
of evolutionary dynamics, and no known artificial system generates the kind of
evolutionary dynamics exhibited by the biosphere [Bedauet al., 1997; 1998]. We
are apparently still missing critical insights about the mechanisms by which evolu-
tion continually creates the new kinds of environments that continually elicit new
kinds of adaptations.
3 PHILOSOPHICAL IMPLICATIONS OF ARTIFICIAL LIFEThe scientific and engineering of artificial life has rich implications for a number
of broad philosophical issues. This section illustrates these implications for a few
philosophical issues.
Philosophy and artificial life are natural intellectual partners, for three reasons.
By creating wholly new kinds of life-like phenomena, artificial life continually forces
us to reexamine and reassess what it is to be alive, adaptive, intelligent, creative,
etc. In addition, both philosophy and artificial life seek to understand phenomena
at a level of generality that ignores contingencies and reveals essential natures.
Finally, artificial life’s computational methodology is a direct and natural ex-
tension of philosophy’s traditional methodology ofa priorithought experiment.
Aiming to capture the simple essence of vital processes, artificial life abstractsaway
as many details of living systems as possible. The resulting artificial life models
are thought experiments that are explored by actually syntehsizing instances of
the models. Like the traditional armchair thought experiments, artificial life simu-
lations attempt to answer “What if X?” questions, but the premises they pose are
too complicated to be understood except by synthesizing them. These synthetic
methods are often computational (in soft artificial life), but they sometimes in-
volve constructing novel hardware (in hard artificial life) or even construcing novel
systems by biochemical means (in wet artificial life). In each case, the motivation
is the same: the behavior of the system cannot be determined except through
direct experience. These constructive thought experiments bring a new kind of
clarity and constructive evidence to philosophy.