Having read about molecular biology and so forth, I would love to be 
able to simulate this kind of thing in a computer. With real organisms, 
you have to guess what happened from the evidence that is left today. 
With a computer simulation, you could potentially record the entire 
history of the whole system.

There's a couple of problems with that though.

One problem is chemistry. In real life, the amino acid sequence of a 
protein dictates the way it folds up to make a functional molecule, and 
the folded shape dictates what the molecule actually does. But 
currently, predicting the way that one single small protein will fold is 
a computational problem so crushing that it brings state-of-the-art 
supercomputers to their knees. The largest computer cluster in the 
world, folding@home, exists to simulate *simplified* protein folding 
experiments.

If it takes that much computer power just to calculate the shape of one 
single protein, then simulating an entire living cell, consisting of 
many hundred trillion protein and other molecules is obviously vastly 
infeasible. Simulating a multicellular organism such as a tiny ant is 
laughably impossible.

The next problem is the size of a typical genome. According to 
Wikipedia, the various genomes that have been sequenced vary from 10 MB 
to 10 GB of data. Now, evolution usually happens in populations of 
thousands if not millions of individuals. If each individual has a 10 GB 
genome and you've got a million individuals, that's 10 petabytes of 
data, right there. And if you want to keep a complete record of the 
genome of every single organism that has ever lived... we're easily 
talking about exabytes of data here.

And then of course, there's the sheer implausibility of /creating/ life 
in the first place. You might have to wait a *long* time for your 
randomly mixing chemicals to do anything even remotely interesting.

Come to think of it, evolution is very, very slow anyway. To get 
anything interesting, you would have to have vast environments of varied 
types to inhabit, and you would have to wait a very long time.

In all, simulating life as it exists on Earth is infeasible. All the 
computers on Earth couldn't handle a few microbes, never mind anything 
that could be called an ecosystem.

Most artificial life simulations don't do this, of course. They simulate 
a small data set, apply some kind of arbitrary "fitness" function to it, 
and keep all the individuals with fitness above some arbitrary cutoff 
value. They then do some kind of duplication / mutation process, and repeat.

The thing is, usually the way the fitness function is set up, there is 
exactly one possible solution, and the system is so simple that there's 
only a handful of ways to achieve that solution. And the whole machinery 
of the system is very simple. Usually only one tiny component of it can 
change.

In real living systems, on the other hand, *everything* can potentially 
be modified. For example, an organism can evolve a brand new amino acid. 
Not just theoretically; this has actually happened. But a computer 
program that just copies a set of codes according to hard-coded rules 
can't ever do something like that.

Real systems have other interesting properties. Organisms can actually 
alter their environment. (E.g., Earth's atmosphere didn't originally 
contain oxygen. Plants invented that.) And organisms don't just die if 
their fitness is less than X, otherwise live forever. They live or die 
depending on how favourable the conditions are. There are predators and 
prey. There is competition for space, food and materials. All this 
richness that the computer simulations almost never include.

On the other hand, I'm still at a loss for how to include all this 
interesting stuff without the simulation slowing down to the point where 
it's slower than *actual* evolution...

Post a reply to this message