I was pretty disappointed that I couldn't attend the GECCO (Genetic and Evolutionary Computation Conference) in 2007, since it was in London. In the wake of Katrina, budgetary concerns led to a shutdown of all overseas travel funding, and the status of that funding is still unclear. Even with financial assistance, I would have had a lot of fees out of pocket, so it just wasn't feasible.

I've never been to London, or anywhere in Europe for that matter, and I also would have liked to have seen this discussion of evolution and complexity with Richard Dawkins, Lewis Wolpert, and Steve Jones. Thankfully it's online, though I don't think you can stream it. You have to download it with a torrent client, though I've gotten permission from the moderator to upload it to Google Video, so as soon as I do that I'll update this post with a link. [Update: Uploaded to Google Video and linked here.]



Anyway, one very interesting exchange was sparked by a question at around 33:45 in the video, having to do with historical, physical, and computational constraints and their relative importance to evolution. Lewis Wolpert asserted that even with an enormous amount of selection pressure and time, you couldn't evolve a mouse with feathered wings.

Steve Jones then follows up, agreeing, talking about historical constraints, and says "That's why pigs can't fly. Pigs can't fly because they descend from ancestors without wings. It's as simple as that."

But wait a minute, I thought. Birds also descended from ancestors without wings. That seems like a pretty absurd thing for two guys who supposedly really know their biology to be saying. All I can reckon is that Jones meant to say something else.

At that point, Dawkins thankfully followed up with a question. He asked why it wouldn't be theoretically possible to apply selection pressure that reverted the phenotype of the mouse to that of the common ancestor, and then apply pressure to favor wings and flight, basically to work your way back down the tree, then up a different path. Wolpert immediately whipped out the strawman, giving some anecdote about a conversation with a physicist in which he asked if it were possible for the man to wake up the next morning as a platypus. I thought this made Wolpert look doubly silly. He kept going on about how difficult it would be to get the right bones onto the back of the mouse. And Jones didn't budge either.

I wish they'd had more time to hash it out a bit more, because it's a topic that interests me greatly, the relative importance of simplification, as opposed to complexification, in evolutionary dynamics. Philip and I wrote a paper on it for the first GECCO (2004) I presented at: A Comparative Analysis of Simplification and Complexification in the Evolution of Neural Network Topologies.

There are plenty of cases in biology where traits have either simplified (leaving vestigial remnants, like the bones in the fins of whales or useless eyes in cavedwellers) or disappeared altogether. There is no reason that I know of why, in theory, pressure could not be applied to a population of dogs so that they would eventually become a population of bacteria. In practice, the amount of time and the exact types of pressure would most likely be unknowable, but I know of no sound argument why the dynamics that shove an evolving gene pool in one direction are not reversible. And if you can believe that a human and a bacteria have a common ancestor, diverging and evolving by tiny steps over millions of years, then why could you not believe that all those tiny steps could be taken in the other direction, given the appropriate pressure?

Is there some evidence that more complex organisms have some sort of inertia? Is there evidence that it is much easier to build up complexity via evolution than to simplify? Even if complexity were subject to some kind of snowball effect, one could argue that it would take longer to revert to a common ancestor, but it still seems unwise to suggest that it is impossible.