Digital Evolution

The Genomics Forum and local listings mag The Skinny have got together to sponsor an evening discussion of what computer games can tell us about evolution at the Drill Hall in Dalmeny Street, Edinburgh.

About seventy people came along to this event. Professor Steve Sturdy introduced me and Alastair Elfick and posed some questions: could games and simulations teach us anything about biological evolution? Could something like life emerge (or be created) in electronic information systems? And are virtual worlds such as Second Life a site of social evolution?

Dr Alastair Elfick spoke first. He explained that although he is called a synthetic biologist, he is actually an engineer and not a biologist. 'Synthetic biology' is genetic engineering Mk 2. It's an attempt to apply engineering principles to the design of genes and organisms - unlike the practice of genetic engineering, which he likened to throwing concrete and steel into a river and if you could walk across the result, calling it a bridge. The aim of synthetic biology is to produce 'biobricks': predictable, off the shelf components that can be slotted together for a defined purpose. Evolution is a problem for synthetic biology: it's something you don't want to happen in your intelligently designed systems!

(This is a very compressed and scrappy version of what Alastair Elfick said. What follows is a somewhat tidied-up and expanded version of what I said.) Following on from what Alastair said: I got my first degree in zoology and went on to do research in biomechanics in a Mechanical Engineering Dept, so I was a biologist and not an engineer. After that I went into IT, where programming was at that time very much like Alastair has described genetic engineering: moving messy pieces of code about and seeing if they worked. But just as synthetic biology is making genetic engineering more rigorous, so structured programming forced us to clean up our act. It was while I was working in IT that I wrote my first novel, The Star Fraction. Some of how I handled the idea of digital evolution in that novel came from reading The Selfish Gene and The Blind Watchmaker. These books filled me with zeal to spread their message. I really thought Richard Dawkins wasn't getting enough publicity.

What I learned from Dawkins' books, or rather re-learned because I'd studied evolution at university, was the power of the Darwinian logic. That logic is very simple. It rests on three postulates: replication, heritable variation, and competition. Replication: something can make copies of itself which then go on to make further copies. Heritable variation: sometimes there are mistakes in copying, and these mistakes themselves get copied. Competition: the materials available to make further copies are not in infinite supply, and some of the variant copies have a better chance of getting hold of the available material than other variants. And that's it! Out of these three conditions you will get natural selection. The upshot of all this is that given some kind of self-replicating molecule or other relatively stable system, it's only a matter of time before you get, as Neal Stephenson put it, 'organisms spamming their environment with imperfect copies of themselves.' As a friend of mine said, that sounds a lot like parenthood. Well, yes, it sounds like parenthood because it is parenthood.

While writing or thinking about my first novel I also read William Poundstone's The Recursive Universe, a fascinating book that introduced me to Conway's Game of Life - not a computer game as such, but a vivid and visible demonstration of how complexity can emerge from the repeated application of a few very simple rules. The Game has had a lot of mathematical thought and computer resources thrown at it, and it has been shown that in principle the universe of Life could generate and sustain self-replicating structures. Mashing this idea together with the 'Watchmaker' program described in Dawkins' book got me thinking about life in computer systems, and by creatively misapplying the metaphors of viruses and bugs I came up with the notions in The Star Fraction of an emergent AI running wild in the world's computer nets and evolving under the savage selection pressure of the programs sent to hunt it down.

[In the discussion, Charles Stross pointed out that while computer viruses aren't a likely starting point for runaway evolution, there is a very common procedure that does seem to select for machine intelligence: the anti-spam measure that tries to block spam by making any incoming message do something only a human being can do. He points out that elsewhere in computer science there's a name for this: it's a Turing Test.]

I'll finish with an amusing or disturbing thought. It's been estimated that spam and anti-spam now constitutes a significant percentage of Internet traffic, much of which is invisible to everyday users, like dark matter is invisible to astronomers. Now what this makes me think about is how the universe might appear to any intelligences that had evolved or were created in a simulated world inside a computer system. It's possible that they would come to understand the virtual 'laws of physics' of the world they lived in, but might never have any way to learn the real underlying laws, which would be the rules of the simulation's physics engine. The disturbing thought is, what distinguishes their situation from ours?