Excerpted from "This is my Vision of Life" (edge.org) by Richard Dawkins
In The Blind Watchmaker, I wanted to get across the idea that cumulative selection can give rise to immense complexity and dramatic changes. So I wrote a computer program for the Macintosh, which presented on the screen a range of phenotypes which were built by an algorithm which I called its embryology, which was actually a tree-growing algorithm. And the shape of the tree was governed by genes. There were nine genes I think in the first version, and so what the user saw on the screen was a "parent", as I called them, in the middle, and eight [actually fourteen, misremembered as eight in the interview] other biomorphs around it were the offspring. They were built by genes which were nine numbers. The genes could mutate by either having a small amount added to their value or a small amount subtracted from their value. So all the nine biomorphs looked a bit different, were obviously descended from the same parent, but they were a little bit different. And you could choose with a mouse which one to breed from, it glided to the center of the screen, produced fourteen offspring and so on. It went on and on through generation after generation. You could breed anything you like. It was a most extraordinary experience to breed massively different shapes from the original by gradual degrees, and they came out looking like insects, and flowers and all sorts of things.
I'm pleased to note that although I’d thought I’d lost these biomorphs, because modern Macs don’t run the software that old Macs do ... a wonderful man called Alan Canon in Kentucky wrote to me and said he wanted to revive them. So I sent him all my old Pascal code, which would no longer run, and he’s now hard at work producing phoenix from the ashes—my old programs—and I’m simply delighted by this.
I then went to the Artificial Life Conference, organized by Chris Langton, and I gave a talk called "The Evolution of Evolvability", which I think was the first time the phrase had ever been used, and it's being used quite a lot.
The original biomorph program had nine genes. I then later enlarged it to 16 genes. I added genes that did things like segmentation, that had biomorphs that were arranged serially along the body like a centipede which has lots of different segments, or like a lobster which has lots of segments, but each segment can be a little bit different. I had genes that had symmetries of various kinds. So I increased the number of genes from nine to sixteen and the repertoire of biomorphs that became possible to breed then dramatically increased. It was still limited, but nevertheless it increased. And it occurred to me that this was a good metaphor for radical changes in embryology that happened at certain important times in evolution. For example, I just mentioned segmentation. The very first segmented animal had some kind of major mutation, which gave it two segments instead of one, I'm guessing. It may have been three. It can't have had just one and a half segments. There must have been at least two. It duplicated everything about the body. If you look at the body of an earthworm or a centipede, it's like a train, like a truck. Each truck is similar to the neighboring trucks and may be identical.
Before the origin of segmentation in the ancestors of earthworms, or the ancestors of centipedes, the ancestors of vertebrates, animals must have evolved as just one single segment, and they would have evolved in the same sort of way as my biomorphs did when they had only nine genes. Then the first segmented animal was born. It must have been radically different from its parents. This must have been a major mutation. And as soon as the first segmented animal was born with two segments, the same as each other, probably ... it wasn't a difficult thing to do in one sense because all the embryological machinery to make one segment was already there. And so to double it would have been obviously a major step. Nevertheless, all the machinery is there. It's not like inventing a whole new organ, like an eye. That cannot happen. It's got to happen by gradual cumulative selection, which is the main message of The Blind Watchmaker. But once you've got the machinery to make an eye, or to make a vertebra, or to make a heart or anything like that, you could make two because the machinery is already there. That's what segmentation is.
And so when segmentation was invented by some kind of macro mutation, a whole new flowering of evolution became possible and vertebrates, arthropods, annelids, all exploit this new embryological trick of segmentation. And I illustrated this with my biomorphs because when I added the segmentation gene for the macro mutation, which I actually had to program in, when I added it, it meant that a whole new flowering of morphology could appear on the screen. You could evolve much more exciting animals because segmentation was there. Similarly with the genes for symmetry. I had genes doing kind of mirror image morphology in two different planes. And immediately I started being able to breed things like flowers, butterflies, beautiful creatures.
The evolution of evolvability, then, is an evolutionary change which makes a radical alteration in embryology, and that opens up floodgates of further evolution which were not possible before. Segmentation is one example, sex may be another one. Torsion in mollusks may be another one. These are major changes, which I think are rare. They may happen once every 100 million years, but there's kind of normal evolution which goes on by the normal cumulative, slow, gradual process that we mostly teach about. But every now and again I suspect there's a major jump, a macro mutation which opens up new floodgates, and segmentation would be the best example. I was really led to think about this by the addition of seven more genes to my original nine gene biomorph, and that's what I talked about at Chris Langton's Artificial Life Conference, and I called it "The Evolution of Evolvability".
I incorporated these ideas of evolution evolvability in Climbing Mount Improbable, which is a bit similar to The Blind Watchmaker, but has a lot more in it. And by then I'd added a whole lot more genes, in this case introducing colors, and we now have color biomorphs. And perhaps rather more interestingly, I teamed up with Ted Kaehler. He was one of Apple's star programmers. I met him at the Artificial Life Conference. And after that we collaborated on a new project which I called "Arthromorphs", which was somewhat similar to biomorphs, but with a totally different kind of embryology, and much more based upon segmentation, and much more based upon especially arthropod segmentation. And the arthromorph program didn't require the programmer, namely me, to introduce the new watershed changes, the new macro mutations which led to new flowerings of evolution. It happened internally, it happened in the computer. They really were macro mutations. That was a big step in my use of computers in both understanding and teaching about evolution.