I'm the inventor and lead programmer of Creatures, so I thought I'd be well
placed to answer part of your query. I'm not the best person to describe
how the game feels, I suppose, but I would like to assure you that it IS
genuine Artifical Life, and give you some details:
The creatures are driven by a biologically-inspired neural network, whose
wiring and dynamical properties are genetically determined. A biochemical
modelling system (also genetically configured) controls diffuse feedback
and reinforcement within the network, and also models the creatures' drives
and needs. In addition, the chemical model simulates physiological
structures such as digestive, reproductive and immune systems.
The network is genetically wired up to provide attention-direction and
learned action-selection functions, and includes episodic and relationship
memories, both short- and long-term. The detailed carrying out of a
creature's actions and the pre-processing of sensory data are both handled
by code, however. Of course, because the structure of the network is
genetically defined, and the genome is both mutable and variable-length,
there is no way of knowing what other neural structures could arise through
evolution, given sufficient generations.
Perhaps the key point in defence of the assertion that Creatures is genuine
A-life ("strong" A-life, even) is this: The program models biological
building blocks and the genome configures them into structures; the code
does NOT specify the behaviour of the creatures explicitely - this is an
emergent property of the system. We don't have genes for "bravery",
"aggressiveness" and so on, but for "chemo-receptors", "neural lobes", etc.
The creatures do what they do as a result of this biologically inspired and
biologically plausible model, not because I told them to do it.
What I aimed to achieve with Creatures was to create autonomous agents that
people were willing to believe were alive, not because I'd managed to fool
them into thinking this, but because I'd managed to convince them. The
creatures are not brilliant learners, nor especially intelligent, but they
do have a suitably life-like dynamical behaviour which emerges from the
structural and dynamical properties of the underlying model, and not as a
result of rules or "behaviour algorithms".
- - - - - - - - - - - - - - - - - - - - - -
Steve Grand
Director of Technology
Millennium Interactive Ltd.
stepheng@millennium.co.uk
- - - - - - - - - - - - - - - - - - - - - -