(See other recent posts about "evolution".)
One reason why I'm skeptical of the theory of evolution is that it's complex but not very graceful. In Artificial Intelligence we often make use of evolutionary principles in computer simulations and try to encourage the emergence of complex systems by combining a simple starting state, mutations, natural selection, and time. What many computer scientists have come to realize is that this approach is often extremely limiting and rarely yields impressive results. In my research, I've yet to see a single instance of an evolutionary model that produces an output that's more complex than its input.
When asked, the computer scientist will explain that the reason the model isn't working very well is that it isn't complex enough and doesn't account for all the factors that exist in the real universe to facilitate evolution. Real biological systems are incredibly complex, as is the environment in which biological creatures live, and in our simulations we -- by necessity -- make many simplifications.
That explanation makes sense, to a point, but after a while it begs the question of how complex a system actually needs to be for evolution to occur. According to evolutionary theory, all that should be required is a genome that can undergo mutations that affect the genome's survivability, some rules to describe how one genome is selected for reproduction over another (natural selection), and a net input of energy into the system (to overcome the second law of thermodynamics). But experimentally, these components are apparently insufficient, and the excuse is generally that the system in question isn't complex enough.
This apology rubs me the wrong way. It just doesn't feel right, and it strikes me as counterintuitive. Everyone who has taken high school physics knows that physics is rather complicated as well, but when we first learn physics we make all sorts of simplifications that make our models easier to understand. We assume there's no friction. We assume every body is a point mass. We ignore relativity. We ignore electromagnetism and the strong and weak nuclear forces. And so forth. After making all these simplifications the models still work rather well, and the general principles of physics can be clearly taught and understood. The complicating factors can be introduced later and incorporated into the models as desired, and they help us get more accurate answers. It's graceful; that's why Newton's Laws served humanity so well before Einstein came around.
When it comes to evolution, however, there is no such grace. The key argument of evolution is that natural selection leads to increased organization and increased complexity, but no computer simulation has ever crossed that essential cusp. Every software system outputs less complexity than it takes in. The explanation that "it just isn't detailed enough yet" isn't very convincing because there's not even a theory addressing how complex a system has to be for evolution to "work", and the theory of evolution itself makes no such claim about required complexity. Plus, it's illogical. Why should there by some complexity threshold beyond which evolution kicks in? Or is there some other missing ingredient that we just haven't identified yet?
The other side of the argument is to point out that the natural selection rules are also part of the input to the system. In a way, evolution can be seen to "move" complexity from the rules of the universe into a genome (presumably with some significant loss due to inefficiency). The argument then says that evolution in the real universe doesn't actually output more complexity than it takes in, because the complexity of the various physical laws that govern biology are themselves part of the input. Is this something that can be measured? The human brain is the most complex system in the known universe by unit mass, but is the univserse itself more complex than human intelligence? If so, then we'll never be able to comprehend it. If not, then the argument falls apart.
That's rather frustrating.
Michael,
You do believe in artificial selection, don't you? It's how six thousand breeds of dogs were developed from one species of European wolf in the last eight thousand years. Mankind has been tinkering with animals since he domesticated them.
There have been cases where mankind forced an artificial selection by accident. There is a bay in Japan where one in four crabs have the face of a samurai on it. When the Japanese people moved into the area five hundred years ago, they found a crab that had random markings on it that looked something like a fierce face, and the superstitious fishermen threw it back. If you throw back crabs that seem to have a face then crabs with those markings will survive to reproduce. Two processes are in operation: the general struggle to live and reproduce and the fishermen rigging the process to favor those crabs that have faces.
What is natural selection? It is that species must conform to fit their environment; those that don't-- must move elsewhere or die. The environment is always changing, thus, forcing species to adapt. Many of those adaptations are random or cosmetic, and we won't if they are necessary until the environment changes enough to push a species to the brink of extinction. This happens. Scientists believe that Leopards have so little genetic variation that they must have gotten down to one female and her cubs. A case of cosmetic adaptation is the Chimpanzee. There are two species-- if you neglect humans as the third Chimp-- the Chimp and the Bobono. The Bobono got trapped south of the Congo River ten million years ago and slowly drifted genetically from the main Chimp herd. It is smaller and has a different coloring. Better doesn't come into it, since both have similar environments. What happened to make us human was a very harsh environment for about a million years that kept us separate until we could no longer interbreed.
Michael, I don't believe we can neglect evolution. Something is going on here. Great questions are being raised, and the scientific evidence is too much to ignore. Moreover, the evidence in one branch of science dovetails with another-- genetics agree with anthropology and zoology. But, it is too soon to discard God either; God can use evolution to make changes just as man has. Both science and religion should say "that is all very interesting, but we don't know enough yet to discard the other."
I don't think you're comparison to learning physics is very appropriate, if you carry it out a little farther.
To say that physics intuitively "feels right" indicates that you may not have studied much modern physics, especially quantum mechanics.
Certainly, Newton's Laws are easy to grasp with a little mathematical knowledge, but they can really be considered just a subset of Einstein's General and Special Relativities, which are far more complex and far less intuitive.
Quantum Mechanics just messes up your head. Still, while it is entirely counter-intuitive and never quite "feels right", it is incredibly accurate.
mn: On the contrary, QM feels quite right to me, for whatever that's worth. "Feeling right" is certainly not scientific, and I know that, but absent actual repeatable science it's just about all we have to go on.
Lou:
Natural selection doesn't create complexity. It reduces it. The European wolves you mentioned apparently had enormous genetic complexity, as evidenced by the fact that subsequent conditions and breeders were able to draw out from them all the different breeds of dogs today. If natural selection were the only force at work, then a given group of creatures would converge genetically towards a single optimum set of genes for the environment. (Then they would be wiped out by a change in that environment.)
Natural selection is almost the opposite of evolution: The former narrows a group's genetic diversity to suit the environment, while the latter somehow expands that diversity to create new species.
Multiple groups can drift apart genetically, as you say, with the same species in different environments. Previously unexpressed genes can come to the fore while previously dominant genes recede. But that doesn't explain where those unexpressed genes came from.
The environment doesn't create genes, it only selects them. The mystery of evolution vs. creation is where all those genes came from in the first place.
I'm just glad that people are finding these posts interesting. I used to write them and get zero comments.
Ok i would like to know,,is it possible for a chimp and a Human to Interbreed? i mean, the genetic structure between humans and chimps are more simular than lets say Horses and Donkeys, and they can produce an offspring, although steril, but nonetheless alive and kicking,, i searched google for " Humanzee" but found nothing conclusive,, =(
that's a very interesting post you've made there.
A lot of physics problems respond well to simplification, but then we tend to only model something tiny like a pendulum. There's a lot more physics and chemistry going on in the simplest cell, so of course we are so far unable to model one. It's not really that physics is more graceful, it's just that the type of problems you're thinking about are very simple. Try taking the same liberties with partical physics and see what happens. There are plenty of examples of physics problems which fail when we try to simplify them. Add a couple of magnets to the simple pendulum and all hell breaks loose. Try plotting the orbits of three point masses.
Attempts at modelling an artificial life in software have not been entirely unsuccessful. For instance, see tierra project (http://www.his.atr.jp/~ray/tierra/) for a very old project which has reproduced some classic life-like behaviour, such as parasites. If you check out the link, you will notice that the virtual world (tierra) set up is obviously reasonably complex code; but the creatures themselves are, in the first instance, remarkably simple.
I'd say that code to produce images of the mandlebrot set is much simpler than its intricate results.
I don't really understand your closing paragraph. Can you be more specific about what you mean by "complexity"? You talk about complexity "moving" from universe to genome, and compare the universe's "complexity" with the intelligence of the human brain, but it's all lost on me. My comments, as far as i understand your points, are: the universe doesn't become less complex because genomes are developing; the genome takes advantage of any complexity in chemistry and physics available (what would stop it?); what does intelligence have to do with complexity? Intelligence is the ability to identify patterns, so we might judge an intellect on the basis of the complexity of the patterns it can recognise. Is this what you're talking about?
Evolution can occur wherever there is variation and selection. But what sort of variation have you been playing with? In the tierra project, one problem that was overcome was the brittleness of machine code. That is, if you vary machine code a bit, the program tends to break. They invented a virtual machine, with an instruction set which wasn't so brittle, so that minor changes could occur without wrecking the program.
Also, for a very long time on earth, life was extremely simple. Not until the cambrian explosion did complexity arise. It might be that if we start with simple "creatures" and populate the "world" with barely a few species, we'd have to wait millions of years for there to be enough variety around for competition to kick in and real circus-grade evolution to take place. That's a little theory of mine.
a chimp and a human cannot interbrede, different number of chromosomes.
jez: Read up bit on information theory, because that's basically what I'm talking about when it comes to "complexity". It's information density.
You said: "I'd say that code to produce images of the mandlebrot set is much simpler than its intricate results."
But that's not true. All the complexity is contained in the code itself, just in a more compressed format than in the images.
As for just running simulations for millions of years, you're missing the point. Time isn't what's important, generations are what's important. We can already run simulations for millions of generations in a relatively short amount of time, and no significant emergent properties are observed.
I wasn't sure that you were using the technical meaning of complexity. I toyed with the idea that perhaps you were talking about entropy, since you brought up 2nd law of thermodynamics.
But now I know that you are, I reckon the complexity of the universe is infinite, since i believe it is in part random. (see qm, bell's inequality etc.) same for my brain. From the way complexity is defined, I expect they're both countably infinite, but i haven't really thought about it.
You do have to run the simulation for enough generations, and more generations on a given simulation / hardware => more cpu time.
What do you mean by "significant" emergent properties? Did you check out the tierran project? Those emergent properties pass my criteria for significance.
what simulations are you talking about? You can't very well expect a simulated biology to evolve to solve tasks they aren't selected for; nor can they ever do things which are a "physical" impossibility in their (simulated) environment. They obey the laws of physics, but they tend to exploit any niches physics has available.
Ben: as well as natural selection, there is "variation". genes collide and recombine through sex, and mutation gradually alters genes, can create new genes given enough time. The code in DNA is not "brittle" -- that is it can tolerate lots of variations before it breaks completely. Our simulations must mimic this property.
I see, because qm "feels right" to you, that it is not resiliance against simplification which upsets you about evolution. Can you think what else it might be?
On a vaguely related note, how do you feel about relativity?
j: Information and entropy are two sides of the same coin. And entropy always wins, in the end.
Plus, there's no way to prove whether the universe contains random elements or not, but there are certainly some things that appear random. That has nothing to do with complexity however, since randomness doesn't create information or contain infinite information.
I'm somewhat familiar with TIERRA and it's pretty interesting work, but I'm not convinced it's anything other than optimization, which isn't a great accomplishment. Plus, my understanding of it is that the system prevents fatal mutations, but I could be wrong about that. Contrary to what you wrote, DNA is actually pretty "brittle", and the vast majority of mutations in nature (other than in introns, arguably) are harmful or fatal.
Random strings tend not to be significantly compressable. And to describe something like the universe, unless it is entirely mechanistic, requires an infinitely long string, with some random element. (bell's inequality strongly suggests that the universe is in part random). Hence the infinite information.
A random string does create information, just not very much (if you already know the relevent frequency statistics). Information is to do with how surprised you are at a message; so if a message conforms with the frequencies you expect a priori, you get the minimum information you could from a message of that length, but it's still positive. And an infinite, random message transmit infinite information.
Fatal mutations aren't prevented in tierra, but code which produces an error advances towards the head of the sinister-sounding "reaper queue". It isn't quite the same as optomisation, because most of the features that arise are the result of competition. I guess the difference is just that the target is moving (just like in life) rather than the standard convergence on some fixed solution, which has been around since since antiquity.
I think DNA is reasonably brittle. As a trivial examples, the code is redundancy (more than one way to represent most amino acids); and the vast majority of the code does not encode proteins (probably more to do with regulation). Which mutations are you thinking of, and what leads you to consider them the vast majority?