say we find a program on the net, and we want to how well it supports life without looking at its source code. so we run the program, and something lifelike appeared on the screen. there are several possibilities:
maybe it's just conway's life started with a particular configuration that produces a particularly long, varied, lifelike run, but any change to the seed ruins the plant.
or maybe the program simply plays a time-lapse video recording of real cells.
or maybe we just got lucky: it just generates random shapes, and we happened to be looking when the army of typing monkeys did something intelligent.
the real question is: how large a variety of interesting behavior can the program produce, and how often does it produce interesting behavior in that space? see here. i call this property vitality.
eg in boids a school/flock emerges from a couple of plausible local behavior rules. in conway's life, a squirmy behavior emerges, but only one behavior. after looking at either of these systems for a while, one gets bored. interesting and novel events don't happen often enough.
how wide a range of different behaviors can the system create? tierra's GA creates a host of strategies, but do the same basic strategies appear every time you run the program? i don't know, i haven't tried it. how alive is a system one must run many times (or wait a very, very long time) before the truly interesting and exotic appear?
the point of Stephen Gould's book Wonderful Life is that in the evolutionary history of life on Earth, every time you run the program, you would get wildly differing results. that is: real life is very vital (see here).
Gell-Mann talks about depth and crypticity.
the other question (how likely are duds) is a classic cosmological/philisophical problem, see The Anthropic Principle.