Monday, June 2, 2008

Can Life Evolve in Simulated Worlds?

Many cute little characters like this "Norn" are the stars of a computer game called "Creatures". In the game, the little characters hatch from eggs, then proceed to learn, forget, eat, get old, get bored, become curious, etc, etc, all seemingly without the player's interference. Could these little creatures really be called alive? Steve Grand, the head programmer of the game believed the answer to be "Yes", and he wrote a whole book to explain why.
The book is called "Creation: Life and How to Make it". On the whole I recommend it, though the cockiness of the title rears its head in the writing as well. My opinion is that the author's philosophies about what life is, and his critiques of other such philosophies are a little hasty, however there are one or two very valuable ideas I came away with.
The first is the idea that while simulations (of life) never become the real thing (actual life), a simulated world or substrate is able to give rise to things that have every right to be called real. For example, one could write a computer program to simulate the behaviour of atoms in some physical space. Under no interpretation would those atoms ever be called real atoms. However, should they begin to cluster into molecules, then perhaps those molecules could be called "real". I think the argument gets more compelling if one takes it up to an even higher level, to proteins, cells, etc. How would a cell "know" that it's constituent atoms were simulated? Does it make a difference?
Regulars to the blog take note: this argument made me think of my objection to things like the "Big Dog" robot displaying real intelligence. Its behaviour is directly programmed, rather than arising naturally from some lower level ideas of gravity, orientation, etc (and perhaps essentially, a desire not to fall!).
Of course, Grand claims in his book that his game provides detailed rules for a virtual world, and for both neural and "digestive" systems for his creatures, and from this substrate springs what he claims is bona-fide life. This leads me to the second thing I liked about the book; the details about the extremely clever design required to program learning (and forgetting!) creatures using very finite computer memory. I admit, after scoffing at his philosophy, his programming ideas about how to simulate lifelike neurons and chemistry in his creatures were nearly inspiring.
For two disappointing final notes, Grand admits he doesn't believe his creatures are conscious; just alive (he says consciousness is tied up with models of the world which his creatures do not have). Secondly, I downloaded a free mini-version of the game called Docking Station, and found that despite all the intelligent simulating I knew was going on, the game was gaudy and somewhat boring. However, you may be interested to know that at the printing of Grand's book, the population of both Creatures, and Creatures players worldwide numbered several million each.


Rachael said...

Thanks for this Chris. I'm not sure what to make of it just yet, but I'm glad you decided to post about "Creatures".

Ashley said...

Are people still playing this? I remember tinkering with it a few years ago. It was cool but maybe a little overhyped, just as you said, Chris.

I wonder if Grand and Molyneux are acquainted.. :)