This title may make it seem like I’m some anti-intellectual creationist who wouldn’t care about evolution unless it somehow affected my everyday life. That couldn’t be further from the truth: I’m a proud supporter of Darwin’s theory of evoultion.
That said, I also like to help fight ignorance. It would be awesome if everyone were as thoughtful as Dopers are, in the sense that they would care about understanding the universe. However, I have to admit that most people hate learning and their only reason for admiring science is the cool shit that it can produce.
This would mean that an effective tactic against creationism would be to show practical applications of evolutionary theory. The common response from scientists is that they must understand evolutionary theory in order to fight anti-biotic resistant bacteria. Unfortunately, creationists can counter with the claim that they believe in “microevolution, but not macroevolution.” The line between the two is arbitrary, but whatever.
Could anyone point out any practical applications of the lessons learned through, say, phylogenetic trees?
I’m also interested in the problems that can be solved through evolutionary computing. I’ve tried learning about it on the internet, but it is difficult without a technical background that I don’t have. The kind of thing I’m looking for is a completely unexpected result that evolutionary algorithms have produced (such as page 11 of this PDF WARNING: PDF).
My masters thesis used evolutionary algorithms to create efficient tours of theme parks (i.e., to wait less in line at Walt Disney World). It’s been commercialized, and appears in the most popular guidebook on that destination.
I’m not sure whether having an EA work on this qualifies as “unexpected” - I certainly always thought it had a shot - but it seems to surprise some folks.
I’ve seen some work on using genetic programming to produce a ranking function that would be used by a search engine to rank web pages relevant to a query. The results were pretty good IIRC.
Welcome to the boards Len. I think I’ve read about your work.
Genetic Algorithms and Genetic Programming is used fairly extensively in machine learning and AI but I would question how relevant it is to actual evolution.
Doug Futuyma (a well-known evolutionary biologist at Stony Brook University) published a neat little article in Science about this in 1995 (Volume 267: pp 41-42). Many of the points he makes, like the one the OP gives, are examples of microevolution, but there is one area where phylogenetic trees are becoming very useful: pharmaceutical companies that look for new antibiotics, medicines, and otherwise useful compounds in plants and fungi. This has traditionally been a very haphazard process in the past, but companies are becoming increasingly interested in how plant defense chemicals evolve so they can more efficiently search for new, economically useful chemicals. Very few of even the described species in the world have been analyzed for potentially useful compounds, and obviously none of the yet-undiscovered species have been, so many of these companies have an enormous amount of specimens and relatively few researchers. Anything they can do to decide which species are more likely to have useful compounds can save them substantial amounts of time and money, and one very clear way of doing this is by studying the evolution of plant and fungal defense chemicals.
The most obvious example is in animal breeding. You breed big fats pigs with big fat pigs and you get really big fat pigs etc… Just look at all of the breeds of dogs we have managed to artificially breed/evolve in a relatively short period of time. It isn’t hard to see how less severe selection applied over a longer period of time could lead to similar results.
Someone will probably point out that breeding and evolving aren’t exactly the same thing but I bet there are plenty of examples of mutations being bred so that they become commonplace because they are useful to humans. This is just an acceleration and diversion of the natural process in ways that are beneficial to us. If the mutation were favorable to the species in question then it would also happen in the wild without human intervention given enough time and a little luck.
Well, the OP asked for instances of evolutionary theory that are meant to convince creatonists that evolution is correct. There are a number of differences between current GA and real evolution that make the case not so convincing IMHO.
First of all, most GA work in a very tightly bounded seach space whereas evolution works in a very generalised space. It’s not clear that the tenents of macroevolution would hold true when working in such a limited domain.
Secondly, GAs still mostly work on the phenotype level rather than the genotype level. And thirdly, most GAs rely very little, if at all, on mutation and rely on breeding to produce most of the variation.
All of this, to me, seems to present a stronger case for creationism where an intelligent designer (the programmer) laid out all possible forms at some starting date and then only allowed microevolution to occur to settle into some local optima.
In general, evolutionary computing is best used for problems where it’s easy to tell how well a potential solution works, but hard to actually come up with the solutions yourself.
One example I’ve seen is image processing: finding regions of land, or water, or cars in a set of photographs. The user identifies a few parts of a sample image set as water or not-water, and the computer comes up with an algorithm to find water in novel images, which can be refined incrementally by the user agreeing or disagreeing with the results.
What do you mean by working on the phenotype level rather than the genotype level? I have some familiarity with GAs (my employer came out with a generalized GA solver for Excel several years ago) and IME, the terms genotype and phenotype are basically synonyms for input and output. The genotype consists of the values that the GA manipulates, and the phenotype is the process being controlled by those values, which is assigned a fitness value based on some criteria. The GA’s goal is to find the genotype that produces the fittest individuals.
A phenotype describes how an organism behaves. A genotype describes why the organism behaves that way. Most GAs don’t care about the why and just describe the behaviour. Technically, they still are operating at a genotype level, it’s just that theres a magical 1 to 1 correlation between genotype and phenotype.
One tenant of ID is irreducible complexity, ie: you can’t get to there from here. GAs conveniently skip that morass by just assuming you can always get to there from here.
well if we are talking macroevolution obviously there is the problem that it happens over a great period of time. are there any practical applications to ancient geology? (of course some ID people have a problem with geological timeframes too I suppose).
I don’t know if this is the sort of thing that you’re looking for, but I’d guess that medical / economic advances that rely on evolutionary thinking are the most practical applications. As soon as you start looking at more than one organism, you need evolutionary theory to make sense of the differences / similarities you see.
For example, a good way to find genes that might be important in humans (for health, etc) is to look for genes that are important in other organisms that can be more easily studied. Say you find a particular gene in species A, B and C but not in D and E; in order to make a good guess at whether you’ll find it in humans you need to know how A, B, C, D, E and humans are related.
An interesting example that I often use to explain is that if you know how a bunch of organisms are related, and you know some trait about them that you’re interested in, you can tell something about how that trait evolves. Imagine that a new parasite comes into a country that affects dogs, and you’re worried that it might aquire the ability to infect humans. If you take a bunch of related parasites, and work out how they are related (i.e. a phylogenetic tree) and whether they affect dogs or humans, you can make a guess how likely this is to happen. Without going into too much detail, you can work out how often the transition from dogs to humans has happened, and hence how likely it is to happen again.
The trouble with coming up with examples for this type of thing is that, to understand them, you have to be adept at thinking in terms of phylogenetic trees and relationships, by which point you probably don’t need examples. That, and the problems with trying to draw trees in ASCII.
I’ve been having precisely this argument with a creationist on another board recently; his stance is essentially that some questions simply shouldn’t be asked and others just aren’t worth asking.
One way to combat this argument is to discuss some other areas of science that didn’t become particularly useful immediately upon discovery; electricity is (I think) an example of this, having been used for nothing much more than intersting, but ultimately useless parlour tricks until the discovery/invention of electric motors, light bulbs, radio waves, etc.
Or to discuss other areas of science and discovery that currently do not have much practical application to the everyday man on the street; nuclear fusion would probably fall into this category; we’re almost as absolutely sure as it’s possible to be that it actually happens and we understand the results it produces; i.e. heavier elements (and the release of energy), but from the POV of the man in the street (or pew), who cares? Of course he will care (or give thanks to God) if we manage to implement a fusion reactor as a source of cheap electricity.
Unless some immediate harm is implicit (and perhaps not even then), you can’t put a limit on human endeavour and inquiry; we simply don’t know what we don’t know, so we don’t know how useful it will be until we know what it is.
That said, evolution is already producing useful results; it’s useful in predicting that pests and diseases are likely to become resistant to pesticides and medicines; it’s useful in testing of medicine (understanding how closely related is a guinea pig to a human is useful in judging the effectiveness of a medicine under test - of course not everyone feels that animal testing is acceptable, but that’s a different argument) and the understanding of molecular genetics (that is part of the whole package you get when you start investigating evolution/heredity etc) is going to become more and more useful in medicine, agriculture and even manufacturing.
What would the practical difference be between today’s GAs and one that didn’t have a 1:1 correlation between genotype and phenotype?
That depends on the problem you’re trying to solve more than the general algorithm. ID says you can’t get there from here not because it’s impossible to produce some organ with any combination of genes, but because any slight change to the genes that produce that organ would render it totally unusable. The path from here to there allegedly leads downhill (fitness-wise) until you’re a few steps away from there, so any individual who goes halfway down that path will be less likely to survive than one who stays put.
You can run into the same problem with a GA if your best solution is buried in a valley of poor solutions. You can get stuck at a local maximum and ignore a much better solution somewhere else. A good mutation operator will help with both issues, but can’t eliminate them.
I dispute this premise. Creationists believe what they do because they believe (their interpretation of) the Bible trumps any scientific research. It seems unlikely to me that anyone who doesn’t already take a critical look at science would suddenly do so because of a clever demonstration.
At the risk of getting this bumped to GD, I’ll leave it at this: I applaud your efforts, not because I think you will actually convince any creationists, but because it is always good to think about how to present complex data in a clear fashion.