Saturday, 1 December 2007


Two interesting seminars this week. The first was by John Houghton, ex- of the IPCC, and was a public lecture on climate change for a UK organisation known as Christians in Science. It was part of a series that this organisation has been running in Southampton, and the third that I've attended.

Anyway, strictly speaking, it wasn't interesting for the scientific content. That was tailored for a general audience, so only really covered material that I'm very familiar with. That said, he was able to reinforce points on the consequences of increasing aridity, and on the net transfer of wealth from the developing to the developed world despite aid and trade.

The interesting content of the talk lay in its discussion of climate change through the prism of religion. Bar a disparaging remark directed at unnamed scientists (i.e. Dawkins) who suggest an incompatibility between science and religion, the thrust of Houghton's religious message lay in stewardship of the Earth. While none of what he said was new to me, it was interesting to see an authoritative defense of the Earth from this angle. Were I religious, I would think the arguments he presented would be very persuasive. To the extent that one wonders what certain ostensibly religious leaders of the Free World would say in response. Dissembling is my first thought.

The second seminar was NOCS's Friday seminar. Unexpectedly, the speaker, Richard Watson, was one that I requested a long time ago. I can't entirely remember why I requested him, but I suspect that his work was being misused by ID evil doers. Anyway, he's a computer scientist working in a group that applies lessons from Nature to computer science problems, though his work also travels the other way, suggesting how Nature may work in return.

The focus of his work is the development of algorithms for solving optimisation problems (e.g. locating function maxima in multi-dimensional space). His thesis is that conventional evolutionary algorithms for doing this are both ineffectual and inefficient. Gradual evolution based on sampling the immediate "environment" is good only for finding local maxima, with the global maximum often impossible to reach.

His innovation mirrors symbiogenesis from biology, where separate organisms combine in symbiotic evolutionary relationships. By allowing computational elements to combine their "traits", rather than simply recombine or mutate them, Watson finds that the resulting algorithm is considerably more effectual and efficient. Watson went on to present results from earlier in the week where he further tweaked his algorithm so that computational elements, rather than combining randomly, chose combinations with other elements that they "encountered" (not quite sure how this was defined) most frequently. This seemed to be a quantum leap, with the already efficient algorithm increasing massively in efficiency (= time to solution).

Watson draws parallels between his algorithm and the evolutionary leaps involved in events like the evolution of cells, prokaryotes, eukaryotes, multicellular eukaryotes, sexual reproduction: events that are not easily explained by gradual evolution. Since several of these events involve symbiogenesis, the parallel was convincing to me. He argues that the natural analogue of his algorithm provides an alternative to standard Darwinism, particularly where "irreducible complexity" may seem a hurdle. This may be why he attracted the attention of IDers (and me). His argument is not that this mechanism replaces Darwinism, but it does offer a distinct alternative mode for evolutionary innovation. As it involves the coming together of evolutionary lineages, it is not something that is perceived as common (though horizontal gene transfer is now a much more widely accepted process in prokaryotes), and this may reflect it being an infrequent process in the wider world.

That said, after the seminar DIR made the observation that it may have some relationship with punctuated equilibrium - certainly his simulation results were very suggestive of this. I was more interested in whether, in an "evolution of evolvability" sense, there would be features of extant organisms that would be suggestive of this. I was thinking of, for example, adaptations to encourage HGT, but it's not something I know much of, and it's not something that Watson has thought much about either (after all, he's a computer scientist, not a biologist).

Anyway, for a rather specialised evolutionary topic, Watson gave a very good seminar. Perhaps over heavy on background, but that's ideal for students. And he's certainly given me something to think about regarding evolution. Something that, particularly pleasingly, won't come as much comfort to evil IDers.

No comments: